Programme Grants for Applied Research

Background Long-term (previously termed chronic) depression is associated with significant personal, social and economic consequences often resulting in a poor quality of life for individuals. Effective treatments that improve quality of life for individuals with chronic depression remain limited. DIALOG+, a solution-focused and person-centred intervention, guided by a tablet-based application, was developed to address these challenges. Initially proven to be effective in handling psychotic disorders, this study adapted and evaluated DIALOG+ for individuals with chronic depression. Objectives The research aimed to adapt and assess the feasibility, clinical and cost-effectiveness of DIALOG+ for improving quality of life, reducing depression severity and enhancing treatment satisfaction among individuals with chronic depression. Setting The study was conducted across nine National Health Service sites in England, encompassing both urban and rural settings. Participants were recruited from outpatient Community Mental Health Teams, including primary care services, such as increasing access to psychological therapies, ensuring diverse sociodemographic characteristics and clinical contexts. Clinicians from varied professional backgrounds, including mental health nurses, psychiatrists and social workers, were trained to deliver the intervention. Design The study comprised multiple interlinked work packages: Systematic review: existing recruitment and retention strategies for trials with individuals with depression. Focus groups and pilot study: to adapt the intervention. Small-scale cluster randomised controlled trial: testing feasibility of the intervention and trial processes. Large multisite pragmatic cluster randomised controlled trial: including a nested qualitative process evaluation and economic evaluation. Training and implementation study: evaluating feasibility and applicability within different clinical services. A total of 366 participants with chronic depression were included. Patients were clustered by clinician and clusters randomised 1 : 1 to receive DIALOG+ or an active control (treatment as usual with DIALOG scale completion). Participants were assessed at baseline, 3, 6 and 12 months. The primary outcome was subjective quality of life, measured by the Manchester Short Assessment of Quality of Life at 12 months. Secondary outcomes included depression severity (Montgomery–Åsberg Depression Rating Scale, Beck Depression Inventory-II), treatment satisfaction (Client Satisfaction Questionnaire-8) and health-related quality of life (EuroQol-5 Dimensions, five-level version). Economic evaluation was conducted from National Health Service and Personal Social Services perspectives. Process evaluation involved implementation data and interviews with participants and clinicians to assess intervention experience. Results The systematic review highlighted the overall poor quality of reporting for existing trials of psychosocial intervention, with many trials failing to outline information on recruitment and retention. It was demonstrated that the intervention was feasible and acceptable to individuals with chronic depression during the feasibility and pilot work. Although the intervention remained unchanged, feedback from the initial phases of the project, alongside lived experience input led to changes in intervention training, which included the addition of case vignettes specific to long-term depression. Within the cluster randomised controlled trial, at 12 months, the DIALOG+ group exhibited a borderline significant improvement in Manchester Short Assessment of Quality of Life scores compared to the control group (adjusted mean difference: 0.18, p = 0.05). This equated to an average improvement of 1 point on 2 of the 12 Manchester Short Assessment of Quality of Life domains. There were no significant differences in any other outcome at 6 and 12 months. However, longitudinal analysis demonstrated significant reductions in Montgomery–Åsberg Depression Rating Scale and Beck Depression Inventory-II scores in the DIALOG+ group at 12 months (p < 0.01), and participants reported better coping strategies and resilience in managing depressive symptoms. The intervention was associated with an incremental cost-effectiveness ratio of £4749 per quality-adjusted life-year gained. Probabilities of cost-effectiveness were 79.8% at a willingness-to-pay threshold of £20,000/quality-adjusted life-year and 87% at £30,000/quality-adjusted life-year. Despite high retention rates (73% at 12 months), challenges included attrition due to COVID-19 disruptions, with whole clusters lost during early lockdown phases. Fidelity to the intervention protocol varied, with just under a third of participants receiving fewer than three sessions (the minimum recommended dose). Nonetheless, the observed treatment effects were robust across varying levels of engagement. Findings from the process evaluation were increased satisfaction with care, improved therapeutic relationships and a greater sense of empowerment. However, some participants noted challenges with the structured nature of the intervention and the use of technology. This was mirrored by the clinician experience, which found DIALOG+ easy to integrate into routine care and appreciated its structured, patient-centred approach. Barriers included variability in implementation fidelity and technological challenges, particularly during remote delivery. Limitations Key challenges across the programme included: (1) disruption from the COVID-19 pandemic, leading to adaptations in delivery methods and pauses in recruitment; (2) implementation variability, with some clinicians requiring additional training to maintain fidelity within the cluster randomised controlled trial; and (3) attrition and missing data, particularly in measures for economic evaluation, which necessitated the use of multiple imputation for some analyses. Conclusions The DIALOG+ demonstrated small but clinically meaningful improvements in quality of life and reductions in depression severity for individuals with chronic depression. While the improvement in Manchester Short Assessment of Quality of Life scores was very modest, it represents a potential gain for a population with long-term difficulties and low baseline quality of life. The probabilities of the intervention being cost-effective in comparison to active control at the National Institute for Health and Care Excellence recommended willingness-to-pay thresholds were high. The intervention was also associated with high patient and clinician satisfaction. These findings support the wider implementation of DIALOG+ in routine mental health care, with adaptations for remote delivery and ongoing support for clinicians. Future directions Further research is needed to explore the long-term impacts of DIALOG+ on quality of life and depression severity. Peer-delivered models and strategies to enhance engagement with the intervention may strengthen its effectiveness and scalability. Trial registration This trial is registered as ISRCTN16864442. The definitive cRCT was registered as ISRCTN11301686. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0615-20010) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 20. See the NIHR Funding and Awards website for further award information. Plain language summary A common mental illness is ‘chronic depression’, which involves long-lasting low mood. Around one in four people who experience depression go on to be diagnosed with chronic depression. People with chronic depression are treated by a range of teams, where they have regular one-to-one meetings with a mental health professional. A solution-focused intervention called ‘DIALOG+’, which structures communication within these meetings improved quality of life, reduced symptoms and saved the National Health Service money when used with people with another serious mental illness called psychosis. The research team were interested to discover if using DIALOG+ with people with chronic depression would have the same benefits. This programme involved six studies which aimed to adapt and test DIALOG+ for people with chronic depression. Throughout, it was important to include the perspectives of individuals with lived experience of chronic depression, so we set up a Lived Experience Advisory Panel. First, we did some small-scale testing with the original DIALOG+ to see what changes might be needed. Findings indicated that the same life areas and treatment aspects were relevant. This meant that very little needed to be changed about the intervention. However, the training for mental health professionals needed adapting. This led to a feasibility trial, which tested the trial processes in preparation for a large randomised controlled trial. Finally, a randomised controlled trial was conducted, which involved 356 people with chronic depression across multiple National Health Service sites in England. We found that receiving DIALOG+ over 12 months marginally improved the quality of life compared to a comparison group. This means that people who received DIALOG+ were more satisfied with their life than people who did not receive the intervention. We analysed how much it costs to deliver DIALOG+ and found that, although DIALOG+ was slightly more expensive, these costs were minimal. The experiences of service users and clinicians using DIALOG+ were largely positive, and many wanted to continue using DIALOG+. Further research is needed to look at how DIALOG+ works over a longer period of time when used in the National Health Service. Research could also look further into the reasons why some health professional and patients did not want to use DIALOG+ or only used it for a short period of time. However, taken together, this shows that DIALOG+ is a worthwhile approach to use for people with chronic depression across different settings in the National Health Service. Scientific summary Background Depression is one of the most common serious mental disorders globally. There is widespread recognition of the negative impact it can cause not only to affected individuals but also on healthcare systems and wider society. Despite the advances in and improved availability of treatments over the last few decades, depression continues to be prevalent within the general population. A substantial proportion of people who experience an acute episode of depression go on to develop a chronic disorder, with many experiencing recurrences or long illness periods throughout their lives. It has been estimated that 20–30% of individuals diagnosed with major depressive disorder, and up to 47% of those treated in mental healthcare services, suffer from chronic-recurrent forms. Long-term or chronic depressive disorders are associated with poorer clinical and social outcomes, including more severe depressive episodes, an increased suicide risk, physical comorbidity, reduced social networks and significant functional impairment. The available evidence rarely distinguishes between initial onset and chronic-recurrent forms, with a resulting lack of understanding of which therapeutic models are most effective, or appropriate, for long-term forms of depression. Additionally, psychotherapeutic approaches are often expensive and require referrals to specialist services or professionals. There is a need, therefore, to develop and adapt evidence-based interventions that can be implemented across different clinical settings to make routine care for people with chronic depression more effective and accessible. The DIALOG+, a technology-assisted and resource-oriented intervention, might be one potential solution. The intervention is supported by an app and is based on the principles of quality-of-life (QoL) research, patient-centred communication and solution-focused therapy. It was purposefully developed to be flexible in delivery and applicable to a variety of settings. The aim of DIALOG+ is to improve clinical outcomes through the use of routine meetings with mental health professionals by facilitating more structured conversations based on the therapeutic principles of solution-focused therapy. The intervention asks individuals to complete a structured QoL scale called the DIALOG scale. This includes 11 items, 8 related to overall subjective QoL and 3 with treatment satisfaction. The items are rated from 1 to 7 on a Likert scale. The DIALOG scale is loosely based on the Manchester short assessment of quality of life (MANSA), but it includes a different number of items and different scoring scale. Crucially following each item on the DIALOG scale, the patient is asked whether they would like help with the area. Following an overview of the 11 items, up to 3 are selected to take forward in the remainder of the consultation. A four-step brief solution-focused approach is then applied to help identify the patient’s resources and what actions can be taken to address the concerns raised in the DIALOG scale. The DIALOG+ has been shown to be effective for service users with psychosis receiving treatment within the NHS. A cluster randomised controlled trial (cRCT), of DIALOG+, compared to an active control, demonstrated that regular use resulted in better subjective quality of life (SQoL), better social outcomes as well as lower treatment costs (£1500 per person) over a period of 1 year. The TACKling chronic depression (TACK) programme aimed to assess whether similar benefits could be established for service users with long-term depression when using DIALOG+. Objectives The overall objective of the TACK programme was to adapt DIALOG+ for service users with long-term depression and then test the adapted intervention to evaluate clinical effectiveness and cost-effectiveness. This was undertaken across six inter-related work packages (WP), with each WP having its own key objective. Work package 1 (systematic review) aimed to synthesise existing knowledge regarding successful recruitment and retention into depression trials. The objectives were to: (1) identify factors which facilitate or impede recruitment and retention of participants into trials investigating psychosocial interventions for depression and (2) identify previous strategies that have successfully improved recruitment and retention into such trials. Work package 2 (exploratory study) aimed to establish the applicability of the existing DIALOG+ software/intervention and had three specific aims: (1) to explore the perspective of different stakeholders in relation to the existing DIALOG+ intervention and identify required adaptations for long-term depression; (2) to exploratorily test the adapted intervention with 5 clinicians and 15 patients, in situ; and (3) to adapt existing online training materials. Work package 3 (feasibility trial) aimed to establish the feasibility of conducting a large-scale trial and had defined feasibility objectives. These were to: (1) assess the feasibility of recruitment and retention strategies from Community Mental Health Teams (CMHTs) and additional NHS services that treat service users with long-term depression; (2) assess the feasibility of the proposed eligibility criteria; (3) assess the feasibility of the selected outcome measures and (4) assess the acceptability of the adapted DIALOG+ intervention for this patient population. Work package 4 (definitive cRCT) had a primary objective to establish whether the use of DIALOG+ over a 12-month period, in various clinical settings, could improve QoL in patients with long-term depression, compared with an active control. Secondary objectives were to evaluate whether DIALOG+ improves secondary clinical outcomes and assess the cost-effectiveness of the intervention. Work package 5 aimed to further understand the trial findings through a mixed-method, embedded process evaluation focusing on the experience of delivery or receiving the intervention, and implementation barriers and facilitators at different levels (individual, team and organisation). Work package 6 aimed to trial the developed DIALOG+ training materials with non-CMHT staff to assess the feasibility and applicability of the training across different clinical settings where individuals with depression may access care. Methods The TACK WPs used a variety of methods, collecting both quantitative and qualitative data, to address the research objectives. Work package 1 comprised a systematic review and narrative synthesis. Papers were identified and extracted from the reference list of the draft National Institute for Health and Care Excellence (NICE) guidelines for the management of depression, published in 2017. Two reviewers independently screened the list of papers and decided upon included papers before key data were extracted. Findings from included papers were narratively synthesised to generate findings. Work package 2 was an exploratory pilot utilising qualitative methods to test the original DIALOG+ software in different clinical settings to establish its applicability and relevance to service users with long-term depression. DIALOG+ was tested in a convenience sample in routine community care appointments for 3 months across three different NHS Trusts in England. Of these, 15 clinicians and 19 service users were individually interviewed about their experiences. Thematic analysis was used to analyse the interview transcripts by an analytic team which included a service-user researcher. Work package 3 was a mixed-methods feasibility study involving a two-arm cRCT and a qualitative evaluation. It was conducted in three CMHTs in South-East England. Eligible participants, and their clinicians, were cluster-randomised to either the intervention (DIALOG+) or an active control group [treatment as usual (TAU) + completion of the DIALOG scale with no clinical input]. Participants were assessed at baseline and 6 months post randomisation. Qualitative interviews were conducted with the participants from the DIALOG+ arm about experiences of receiving and/or delivering the intervention. Work package 4 was a large-scale, multisite two-arm cRCT, that was hosted in nine NHS sites across England. Eligible participants were identified by their treating clinician and screened by a researcher. Once randomised, all participants within the same cluster (clinician acted as the unit of randomisation) received regular sessions of DIALOG+ over a 12-month intervention period. The active control arm used the DIALOG scale at the end of their routine sessions. Participants completed outcome measures at baseline, 3, 6 and 12 months post randomisation, facilitated by blinded researchers. Primary outcome was SQoL as measured by the MANSA at 12 months. Work package 5 used a mix of quantitative and qualitative methods as part of the process evaluation. One-to-one qualitative interviews were conducted with trial-arm service-user participants who were purposively sampled based on key criteria. Additionally, routine (clinical) DIALOG+ data that were collected during intervention delivery were extracted from the trial tablets and analysed descriptively. Finally, audio and video recordings of DIALOG+ being delivered were analysed to check for clinician fidelity to the therapeutic manual. Work package 6 piloted the use of the DIALOG+ training package across different services. Post-training surveys and in-depth qualitative interviews were used to explore views about the training delivered and any additional needs. Survey data were analysed descriptively, and framework analysis was used to generate findings from interview transcripts. Results The systematic review synthesised data from 90 manuscripts. Publication dates ranged from 1996 to 2016, and studies came from 17 different countries. Findings were very mixed, highlighting that equal numbers of included studies either over or under-recruited individuals within their trials. Few studies offered incentives to increase participation, and the modality of delivery of the intervention seemed to have a small but significant impact on retention. Recruiting directly from mental health services was the most popular approach, compared to allowing self-referral, but the impact on retention was unclear. The overall quality of depression trials was low, and the reporting of recruitment and retention strategies was poor despite the Consolidated Standards of Reporting Trials statements. For WP2, the analysis indicated that DIALOG+ was viewed as acceptable by both service users with long-term depression and their clinicians, albeit with some caveats. Clinician training required significant improvements to address perceived issues, particularly around support with using technology and integration of DIALOG+ into their working practices. The content of the intervention itself needed no adaptation and was found to be directly applicable to the patient population. For WP3, 36 service users and 11 clinicians were cluster-randomised into the trial. This represented a 75% recruitment rate, explained by organisational-level difficulties at one potential trial site. Eligibility criteria were therefore modified ahead of the definitive trial launch to ensure adequate inclusiveness and reflection of the clinical reality relating to the treatment of long-term depression. The selected outcome measures were feasible except for the clinician-rated ones, which were replaced in the main trial case report form. There was a trend towards better SQoL and a decrease in depressive symptoms at 6 months in the intervention arm. Qualitative interviews were conducted with seven service users and six clinicians, with data indicating that DIALOG+ was highly acceptable to those that used it. Within WP4, n = 366 service-user participants were recruited to the trial, representing a 97.3% recruitment rate, with a 7.4% screening failure rate. One hundred and twenty-eight clusters were randomised, with an average cluster size of 2.9, ranging from –1 to 6. The primary outcome was assessed in 79% of the sample at 3 months, 76.5% at 6 months and 73.4% at 12 months. The difference between groups on the primary outcome (MANSA score at 12 months) was significant at p = 0.05, with the direction of effect indicating an average improvement in MANSA-assessed QoL by one point on approximately 2–3 of the 12 composite questions. No statistically significant effects were observed for SQoL at 3 and 6 months, nor any of the other secondary outcomes at any time point. The longitudinal analysis indicated a significant but small positive effect for the intervention compared to the active control at all time points. The majority of participants were satisfied with their treatment while using DIALOG+, with 78% of respondents saying they were a little or totally satisfied with their care after 6 months. When asked, at 12 months, if they preferred DIALOG+ to the treatment they had received previously, 85% had either no preference or preferred DIALOG+. Economic evaluations showed that the overall costs involved in providing the intervention arm were relatively low (£16.86 per participant in the DIALOG+ arm vs. £12 per participant in the active control arm). Base-case analysis suggested that DIALOG+ is more effective over the 12-month trial period than TAU, justifying the marginally increased costs. The probabilities of DIALOG+ being cost-effective in comparison to TAU were high at the NICE-recommended willingness-to-pay thresholds [79.8% at £20,000/quality-adjusted life-year (QALY) and 87% at £30,000/QALY]. Within WP5, 43 service users and 25 clinicians were interviewed. Analysis of data resulted in six themes that described the derived and experienced benefits of using DIALOG+; these were: (1) structure and focus; (2) getting a bigger picture; (3) mapping the recovery journey; (4) focusing on strengths; (5) working on solutions and (6) the impact on therapeutic relationship. Regarding the extracted DIALOG+ data, in total, there were 496 individual DIALOG+ sessions recorded. Number of sessions delivered per participant ranged from 1 to 12, with the mean number of sessions being 3.64; 49.2% of included patient participants had between three and six sessions, representing an adequate intervention dose. Participants rated their mental health as the lowest life domain (mean score = 3.45), and relatedly, the most frequent request for additional help was with their mental health (in 62.7% of sessions). Fidelity checks were completed on six observational recordings and resulted in generally high adherence scores, showing that clinicians, in general, tended to be able to apply the therapeutic principles they were taught in training. The global adherence score ranged from 13 to 17 (out of a possible total of 20). In total, 86 clinicians responded to the post-training survey in WP6. Overall, clinicians found the training to be sufficient, with 90.7% of respondents agreeing or strongly agreeing that the training seminar met its stated aims and objectives. However, confidence was lacking in using the DIALOG+ app, with only 69.8% agreeing that they felt confident using the technology, and only 66.3% either agreeing or strongly agreeing that they felt confident in delivering DIALOG+ to their caseload. Fourteen clinicians agreed to be interviewed following their attendance at training and the application of DIALOG+ as part of their clinical roles. The success of the training session was linked to several elements, primarily the relevance of the training content to their specific service setting, coattendees and the skill of the training facilitator. Factors to improve training were most often related to technical aspects of using the technology. Overall, clinicians from non-CMHT settings were content with the generic training they received and felt it was applicable to their routine practice. Conclusions The combined findings of the TACK programme demonstrate that DIALOG+ is a feasible and effective intervention for use with people with long-term depression in community care and other clinical contexts. Although the effect between DIALOG+ and DIALOG observed in the present study were small, this merged with the findings of the other definitive cluster RCT (cRCT) testing DIALOG+ for people with psychotic disorders [European Prediction of Psychosis Study (EPOS)], which suggests that DIALOG+ can be used as a generic, care planning tool across a range of serious mental illnesses. Prior to the programme, the evidence base for use of DIALOG+ was restricted to service users with psychotic disorders, there are now two large definitive trials with positive results, demonstrating that DIALOG+ can improve QoL for service users with a range of mental health symptoms and disorders. This is encouraging for services such as CMHTs who treat diagnostically heterogeneous populations and often need tools and resources that need to be applied across a diverse range of presentations. Findings from the training study, however, suggested that more development work is needed to ensure that clinical training in DIALOG+ is reflective of the realities and specificities encountered by individual clinicians or teams. Although the current general training resources are adequate, some elements of the training, such as clinical vignettes, are required to be context-specific. Future research should also investigate how ongoing supervision can support the implementation of DIALOG+ and which training models lead to the highest fidelity and highest patient satisfaction. Finally, in-depth economic evaluation showed that the costs involved in the delivery of DIALOG+ were relatively low, both in absolute terms and in comparison to the control arm. DIALOG+ has high probabilities of being cost-effective compared to active control from both the NHS and Personal Social Services perspectives and the societal perspective. Therefore, DIALOG+ may be a clinically effective and cost-effective intervention for people with chronic depression (including those with complex presentations), which is relatively simple to use and integrate into exiting systems and practices within the NHS. Future research should focus on how DIALOG+ can be sustainably implemented and embedded within mental health services, including integration within established electronic patient record systems. Such research could be completed through naturalistic experiments or ethnographic observational methods, considering that DIALOG+ is increasingly used within select NHS Trusts. Trial registration This trial is registered as ISRCTN16864442. The definitive cRCT was registered as ISRCTN11301686. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0615-20010) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 20. See the NIHR Funding and Awards website for further award information.

Background Dementia affects nearly 1 million people in the United Kingdom and their unpaid carers physically, psychologically, socially and economically, highlighting that both need support. Post-diagnostic support is recognised as inadequate globally. In England, it is fragmented or absent apart from primary care – leaving many feeling unsupported as the condition progresses. The Dementia PersonAlised Care Team research programme began in 2018 to address evidence gaps on the effectiveness of dementia support roles and what constitutes ‘good’ support. Objective Dementia PersonAlised Care Team aimed to develop and evaluate an intervention comprising dementia support workers based in primary care, supporting people with moderate to severe dementia and their carers. Phase 1 focused on intervention development and feasibility testing. Phase 2 evaluated the dementia support workers’ value and impact. Our Peer Research Group contributed to the design, analysis and interpretation of data in both phases. Design and methods Phase 1 employed a realist approach to develop theory, the intervention and practitioner support package. This was informed by literature, consulting experts by experience, academics and practitioners and a formative evaluation. A pilot cluster randomised controlled trial tested the feasibility of recruitment, outcome measures and eligibility criteria. Phase 2 included a realist longitudinal mixed-methods evaluation to determine what worked, for whom and in what circumstances. High-volume qualitative data included realist interviews with participants with dementia and carers at two time points, interviews with dementia support workers and general practitioner staff/other practitioners, observations, dementia support worker case notes and reflections, and medical records. Quantitative measures of health- and quality of life-related outcomes, experience of care, carer well-being and support, and resource use were collected at three time points. A mixed-methods integration was conducted following separate qualitative, quantitative and economic analyses. Setting and participants Phase 1 recruited participants with dementia (some without capacity) and their carers (if present) from general practitioner practices in rural Devon and Greater Manchester. In phase 2, participants were recruited from practices in rural and coastal areas in Devon (high deprivation) and urban Greater Manchester (including areas with higher representation of South Asian communities). Intervention The Dementia PersonAlised Care Team intervention, developed in phase 1, delivers personalised, proactive and integrated dementia care via trained dementia support workers embedded in primary care. Drawing on coaching principles, dementia support workers empower decision-making and address psychosocial, health and practical needs through flexible, ongoing support. Support is tailored to individuals’ and carers’ needs. Consented participants were offered 12 months of support, which is flexibly stepped up or down based on need. Findings In phase 1, the Dementia PersonAlised Care Team intervention was tested in 10 general practitioner practices. COVID-19 necessitated switching to remote intervention delivery, recruitment and data collection. Intervention and recruitment strategy (including capacity assessment) were acceptable. However, only 51% (56/110) of the recruitment target was met, making a fully powered cluster randomised controlled trial unfeasible within the funding envelope. A mixed-methods longitudinal realist approach was therefore adopted for the evaluation (phase 2). In phase 2, 126 PwD and 121 carers were recruited from 13 practices – reaching 70% (126/180) of the recruitment target. Retention in the intervention was high (79%). However, participation in follow-up data collection was poor for all quantitative measures, for example, only 22% (28/126) of participants completed the Engagement and Independence in Dementia Questionnaire at 9–12 months. The longitudinal realist analysis identified four areas of positive impact linked to dementia support worker actions: living well, proactive care, reactive care and care transitions. Participants spoke about improved social engagement, medication management and holistic care outcomes despite the inevitable functional decline. Supervision and peer support for dementia support workers, along with joint work with practice teams, were key to delivering this care. Quantitative data showed either stable or declining trends except for carer perceptions of support, which improved over time. Realist-informed economic analysis identified 23 cases where dementia support workers likely prevented high-cost care escalation through early intervention, access to general practitioner records and ability to prompt general practitioners’ care. Resource use shifted towards increased primary/community care and reduced secondary care costs. At a caseload of 53 dyads at any one time, mean annual cost per dyad was estimated at £1222.48. Limitations Low data completion rates, individual variability and the non-randomised design significantly limited utility of quantitative findings. Conclusions This work demonstrated how gaps in post-diagnostic support for people with dementia and carers can be addressed through tailored, proactive support and dementia support workers being fully embedded in primary care. A comprehensive training and ongoing support package is essential for enabling dementia support workers to carry out their role. Future work Further implementation research is needed to test delivery in changing National Health Service contexts, including whether different roles can deliver personalised dementia support. Study registration This study is registered as ISRCTN90828574. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0217-20004) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 17. See the NIHR Funding and Awards website for further award information. Plain language summary Support for people with dementia after diagnosis remains very limited and uneven nationally. The Dementia PersonAlised Care Team study examined how a dementia support worker based in general practitioner surgeries can best support people with dementia and their carers, and the value of such support. In phase 1, we consulted people with dementia, carers and experts, and examined the literature. We developed a dementia support worker role and training package, and worked out how to recruit and collect data from people with dementia, who may have lost capacity. In phase 2, we recruited people with dementia and their carers from general practitioner practices in Southwest and Northwest England. Practices were in rural, coastal, urban areas, some with higher deprivation levels or South Asian communities. One hundred and twenty-six people with dementia and 121 carers from 13 general practitioner practices participated. Participants received 12 months’ dementia support worker support. Multiple types of qualitative and quantitative data allowed us to examine value in different ways and how Dementia PersonAlised Care Team worked. Eighty per cent of participants stayed engaged with the support. Qualitative data showed how positive outcomes were associated with dementia support worker actions in four areas: living well, proactive care, timely reactive care, and care transitions. Additionally, there was some evidence that dementia support workers help avoid expensive negative outcomes. How dementia support workers were supported was also crucial. Quantitative data were hard to interpret due to low completion rates (22%) and a lack of controls. Only carers’ feelings of support improved over time, while other measures remained unchanged or declined slightly. The cost to support a person with dementia and their carer was £1222 per year. The study highlighted how positive outcomes are possible for many and how they came about. It demonstrated how the dementia support worker model can integrate with primary care to deliver support and provided guidance for organisations to start delivering dementia support worker care. Scientific summary Background Dementia currently affects almost 1 million people in the UK. The condition has physical, psychological, social and economic impacts that extend to unpaid carers, who may experience the emotional, physical and financial strain of caring. Support for both is, therefore, important. Unfortunately, post-diagnostic support remains inadequate globally and in England, or absent except for primary care. Many feel unsupported especially once beyond diagnosis and as the disease progresses. When the Dementia PersonAlised Care Team (D-PACT) research programme launched in 2018, evidence on the effectiveness of dementia support roles – variously known as dementia support workers (DSWs), advisors or care navigators – was limited and varied in models. Nor was it known what counts as ‘good’ post-diagnostic support. Although recent studies have begun addressing these gaps, questions remain. Aims and objectives The D-PACT study was funded by the National Institute for Health and Care Research (NIHR; PGfAR RP-PG-0217-20004; 2018–24) to develop and evaluate a personalised post-diagnostic support intervention delivered by DSWs based in primary care. Intended for people with moderate to severe dementia – who often miss out on care – and their unpaid carers, the intervention also aimed to respond to calls for a lead health or social care professional to act as a single point of contact as well as recommendations for a task-shifted approach to dementia care where primary care’s role is expanded. To achieve its aim, the programme carried out two phases of work. Phase 1 had two work packages (WPs). WP1 aimed to develop and iteratively test the programme theory (an explanation of how the D-PACT intervention could have its intended effects). An ‘intervention delivery platform’, that is a manual, training and supervision guide, was also developed. WP2 tested the feasibility of a cluster randomised controlled trial (cRCT); it demonstrated that the intervention was acceptable but that a RCT design was not feasible for the phase 2 evaluation. Phase 2 aimed to: (1) test and refine the D-PACT programme theory, (2) assess the intervention’s value (importance, usefulness) and impact (effects, positive or negative) in a range of settings (rural, urban, coastal, high deprivation, South Asian communities) and (3) advance methodology for community-based dementia studies, focusing on inclusive recruitment and meaningful outcome measurement for individuals with variable capacity, building on phase 1. Public and patient partners [‘Peer Research Group’ (PRG)] and a professional reference group (‘Expert Reference Group’) contributed to both phases. Methods Phase 1 (work package 1 and work package 2) A realist-informed approach to theory building was employed in WP1. A prospective phase developed initial programme theories (IPTs) and the intervention delivery platform. We: consulted stakeholders comprising people with dementia (PwD), academics and practitioners conducted a pragmatic review of existing literature and existing primary care interventions identified key principles underpinning personalised care (to form domain headings for IPTs and ensure the intervention was guided by these principles). In the feasibility study: IPTs were elaborated. The prototype intervention was tested via a realist-informed formative process evaluation that took place in general practitioner (GP) practices in Southwest (SW) (Devon) and Northwest (NW) (Greater Manchester) England, allocated to intervention or treatment as usual at a ratio of 3 : 1. Theory and intervention were refined based on qualitative data (semistructured realist interviews and observations) with those receiving the intervention (PwD and their carers) and those providing the intervention (DSWs and their supervisors). In both the prospective and feasibility phase, an ‘if-then’ heuristic was used to guide theory development. However, as discussed later (Phase 1: initial programme theory development), a decision was made to switch to using a Context-(intervention) Component-Mechanism-Outcome heuristic to configure programme theories. In WP2, trial methodology was developed by testing outcome measures, recruitment and eligibility criteria in recruited GP practices. Candidate primary [Dementia Quality Of Life Questionnaire (DEMQOL)] and secondary outcome measures [Psychological Outcome Profiles (PSYCHLOPS); Engagement and Independence in Dementia Questionnaire (EID-Q); modified Carer Wellbeing and Support (mCWS)] were selected on the basis of literature and consultation with the PRG. A pilot cRCT trial tested the feasibility of GP practice and participant eligibility criteria, recruitment, randomisation and retention. Phase 2 We conducted a longitudinal mixed-methods realist evaluation, using high-volume qualitative data from multiple sources, quantitative outcome and experience measures and resource use data collected at regular time points over the intervention period. People with a diagnosis of dementia and unpaid carers, as well as DSWs, their supervisors and staff from participating surgeries and the community were recruited. A minimum target of 90 (maximum 180) PwD and 80–160 carers (to account for PwD without carers) was set. Qualitative data comprised: realist interviews with individuals, carers (at two time points – 3–4 months and 9–12 months) and DSWs, DSW case notes, reflective logs, researcher observations and extraction of electronic health records. Quantitative measures for PwD included were the EID-Q and the Person-centred Community Care Inventory (PERCCI); carer support and well-being were assessed using the mCWS. In addition to resource use and EuroQol-5 Dimensions, five-level version (EQ-5D-5L), timesheets that recorded time on all activities were completed by DSWs. Quantitative measures, resource use and EQ-5D-5L were recorded at baseline [time (T) 0], 4–6 months (T1) and 9–12 months (T2). Three analyses were conducted: Case study analysis of longitudinal high-volume multiple qualitative data. Statistical analysis of quantitative outcome and experience measures collected at three time points. An exploratory realist-informed economic analysis, combining conventional health economics (HE) tools, cost the delivery of the intervention with realist approaches to understand the impact of the DSW on relatively low frequency but high-cost events like unplanned hospital admissions, respectively. Data were then integrated into a joint display matrix. Results Phase 1 work package 1 The D-PACT intervention, and related theory as to how it is intended to work and be supported, was the key output. DSWs, based in primary care, provide personalised and integrated support to PwD and their carers. Support in the D-PACT model is ongoing (i.e. people are not discharged); frequency and intensity depend on need. DSWs are trained to use a coaching-inspired approach to empower individuals in decisions about their health and care, as well as support in a range of areas depending on what matters to the PwD and carer. This includes psychosocial well-being, physical health, future planning, practical help, signposting to support elsewhere and joint work with primary care staff. DSWs are supported in turn via supervision, peer support and refresher training. Key learning from WP1 included: The intervention was highly acceptable to participants. The D-PACT model was more easily articulated as two tiers: Delivery tier: theories explained how outcomes (e.g. engagement, shared understanding) were generated for PwD and carers within specific contexts. Facilitation tier: theories explained how outcomes (e.g. readiness for the role) for the DSW were generated through training and supervision. Identified gaps in theory for phase 2, for example, theory relating to collaborative working within primary care, support with transition points, and how varying levels of readiness among participants affect outcomes and intervention delivery. Some outcomes may be observed only over time, after DSWs themselves had received sufficient ongoing support and developed confidence to deliver the intervention as intended. Remote delivery of the intervention, necessitated by COVID-19 lockdowns, could influence which mechanisms were triggered and needed to be addressed through training. Phase 1 work package 2 Participant retention and measures: By 12 months, 60.7% (34/56) of participants with dementia and 35.7% (20/56) of carers were lost to follow-up. Completion rates across measures were also low at follow-up: 10.7% (27/56) for PSYCHLOPS, 23.2% (13/56) for EID-Q and 28.6% (16/56) for DEMQOL. Carer completion rate of the mCWS was 42.9% (24/56). The EID-Q was selected for phase 2 because it mapped most closely onto programme theory outcomes. A measure for experience of care, the PERCCI, was added. The mCWS was retained in phase 2. The Montreal Cognitive Assessment (MoCA) was included to measure cognitive status of participants with dementia at baseline. Recruitment and eligibility A four-step approach to recruitment was developed (letters, phone call and visit) along with detailed procedures for consent with varied capacity, in person and remotely. Recruitment occurred during two active periods due to COVID-19: in-person (September 2019–March 2020) and remote (September 2020–March 2021). Ten out of 33 (30.3%) GP practices approached participated, with recruitment rates of 50% in the SW and 24% in the NW. Eligibility criteria for moderate to severe dementia were based on Addenbrooke’s cognitive examination III (> 85) or the modified Telephone Interview of Cognitive Status (> 29). This criterion was removed in phase 2. Ultimately, 56 individuals (13% of 421 approached) with dementia and their carers were enrolled, meeting 50.9% of the recruitment target. Phase 2 Recruitment, participants and retention Seventy per cent of the recruitment target was achieved: 126 participants with dementia (14.7% of those approached) and 121 carers were recruited from 13 GP surgeries (6 in Southwest Devon and 7 in Greater Manchester). Sixty-one per cent of participants with dementia were female, 5% were South Asian. Mean age was 68 years [standard deviation (SD) = 13]. Severe cognitive impairment was indicated by a mean MoCA score of 4 (SD = 3). Intervention engagement remained strong at 79.3%. However, data completion rates were low: 53.9% of participants and 57% of carers contributed at T1, and 23% (both groups) at T2. Not all returned usable questionnaires, further reducing sample size to be analysed. Realist case study findings (qualitative) This analysis revealed that DSWs delivered holistic and personalised support to both PwD (including those with other long-term conditions) and carers. A broad range of care and support needs were addressed. Achieved outcomes linked to DSW actions were seen across the four care domains, and included support to live well, improved physical and mental health, planning for the future and easing transitions in care. Mechanisms that led to ongoing disclosure, engagement and enactment of agreed actions (immediate outcomes) were triggered by early relationship-building by DSWs. These immediate outcomes created the context that facilitated DSWs’ delivery of personalised care. In order for the intended outcomes from personalised care to be realised, people needed to feel more ready to take next steps towards increasing their independence, health and quality of life. Challenges to achieving outcomes included limited community resources, insufficient buy-in from professionals and delays in accessing patient records. Some gaps were identified in the practitioner support package, such as vague role explanations, which sometimes delayed participants’ engagement. Insufficient DSW hours for those on part-time contracts [0.4 full-time equivalent (FTE) or lower] also affected delivery for a small number of cases. Over time, DSWs developed professional autonomy through supervision, training and collaboration with other professionals, enabling them to adapt and personalise care more effectively. The findings highlighted the importance of refining the support package and promoting professional collaboration to ensure consistent delivery of personalised care. Statistical analysis findings (quantitative measures) Median scores for EID-Q, PERCCI and mCWS measures were reported across time points. Median EID-Q scores declined from 73 (51–90) at T0 (N = 97/126; 76.9%) to 60 (44–78) at T1 (N = 51; 40.5%), before rising slightly to 67 (37–79) at T2 (N = 28; 22.2%). PERCCI scores for care experience remained stable. mCWS well-being scores for carers showed little change but declined at T2, while mCWS support scores improved. Findings should be interpreted cautiously due to the non-randomised design and data limitations. Health economics Six DSWs, equivalent to 3.1 FTEs, delivered the intervention with 1446 direct participant contact events over 12 months, averaging 11.5 events per dyad/PwD. Mean contact duration was 45.47 minutes, though support varied greatly, with 15% receiving under 1 hour annually and 9% exceeding 15 hours. DSWs also performed related activities like liaising with GPs, training and updating files. On average, DSWs spent 31.3 hours on each dyad/individual per year. Assuming 44 working weeks per year and a 37.5-hour week, we estimate a full-time DSW could support an average of 52.7 dyads/individuals at any one time per year at an estimated cost of £1222.48. An exploratory realist-informed economic analysis of qualitative data identified 24 cases where DSW actions (e.g. flagging an emerging infection, risk of falling or urgent need for medication review to the GP) helped mitigate high-cost events like unplanned hospital admissions. Access to GP records, ability to task GPs digitally and an ongoing relationship where trust/knowledge had been built with dyads were critical in achieving this outcome. Four instances of unprevented escalation highlighted challenges like stretched services, delayed DSW action due to uncertainty about solutions and dyads who were difficult to contact. These findings suggest DSWs may reduce high-cost inappropriate care in some cases, supporting the role of personalised, proactive intervention. Discussion Synthesis Qualitative and quantitative (including HEs) findings were examined in a joint matrix display. This revealed some discrepancies between quantitative and qualitative findings. EID-Q scores declined, while qualitative interviews suggested improvements in social engagement, medication management and other care aspects. Similarly, PERCCI scores on experience of care remained relatively stable but did not reflect the enhanced care described in interviews. mCWS measures of well-being showed an initial increase then decline, but did not capture the subjective experience reported in qualitative interviews. Only the mCWS measure of perception of support, which increased over time, was consistent with qualitative data. Overall, despite increasing needs as indicated by the EQ-5D-5L, DSWs might have helped mitigate or delay deterioration. Strengths Dementia PersonAlised Care Team is the first study to conduct a realist evaluation using high-volume longitudinal data. This enabled case-by-case analyses assessment of the impact of the intervention. It also supported further elaboration of an evidence-based model detailing how to achieve personalised dementia care based in primary care. Other methodological contributions include innovations around GP practice recruitment that drew on embedded researcher models and digital capabilities in order to reduce administrative burden and encourage participation. Limitations The key limitation was the poor completion rates at follow-up and non-randomised design, making interpretation of quantitative analyses problematic. Despite concerted efforts to recruit GP practices in a range of settings, the participant group was not as diverse as we had hoped. Conclusion The D-PACT programme theory outlines personalised support strategies for PwD and their carers, demonstrating impact in living well, addressing health needs via proactive and timely reactive care, future planning, and care transitions. Its manual, support package and detailed evaluation serve as key resources for providing dementia support in primary care. Study registration This study is registered as ISRCTN90828574. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0217-20004) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 17. See the NIHR Funding and Awards website for further award information.

Background Inflammatory bowel disease (IBD) adversely impacts the quality of life. Healthcare professionals target treatment at controlling gut inflammation. However, patients report troublesome symptoms of fatigue, abdominal pain and urgency or faecal incontinence in the absence of inflammation. Objectives To address the symptoms of fatigue, pain and urgency/faecal incontinence in people living with inflammatory bowel disease. Research questions: Identifying the burden of symptoms: 1a. What proportion of people with inflammatory bowel disease have fatigue, pain or urgency/faecal incontinence and want help for them? 1b. Which symptoms could benefit from optimisation of medical treatment? Addressing symptoms: 2a. To develop and test a supported tailored online self-management programme (IBD-BOOST) for these symptoms in inflammatory bowel disease to improve the quality of life and reduce symptoms. 2b. Does better symptom control reduce healthcare and personal costs and is the intervention cost-effective? Design and methods Qualitative interviews with patients and nurses. Questionnaire to identify the inter-relationship of symptoms and the desire for intervention. Feasibility study of a checklist and algorithm (‘Optimise’) to manage medically reversible causes of symptoms. Multicentre, two-arm, parallel-group, randomised controlled trial of the IBD-BOOST intervention with qualitative and quantitative process evaluation. Target population: 740 people with inflammatory bowel disease recruited from the questionnaire survey. Six months of access to IBD-BOOST was compared with care as usual. Setting and participants United Kingdom inflammatory bowel disease clinics and online. Interventions Randomised controlled trial – a 12-session online self-management IBD-BOOST intervention, developed jointly with patients. Main outcome measures Primary outcomes for randomised controlled trial: United Kingdom Inflammatory Bowel Disease Questionnaire (inflammatory bowel disease-related quality of life) and Global Rating of Symptom Relief at 6 months. Cost-effectiveness of IBD-BOOST was assessed: incremental cost per quality-adjusted life-year 1 year following randomisation from healthcare service and societal perspectives. Results Interviews People with inflammatory bowel disease told us that they wanted help to self-manage their symptoms. Inflammatory bowel disease nurses wanted to be supportive but reported limited capacity to help. Survey There were 8486 responses: 3281 men and 4883 women. In the previous week, 2550 (30%) reported fatigue; 1766 (21%) pain and 4565 (54%) reported faecal incontinence; 925 (10.9%) reported all three; 56% of all respondents ‘definitely’ wanted help for fatigue; 42% wanted help for pain; 53% wanted help for incontinence and 29% ‘definitely’ wanted help for all three symptoms. Feasibility study of the optimise checklist and algorithm The 201/515 (39%) individuals reporting symptoms consented and 194/201 (97%) returned the symptom checklist; 157 (81%) returned a postal faecal calprotectin sample. The algorithm suggested at least one clinical test or intervention for fatigue, pain or incontinence in 67/157 (43%) participants, of whom 25 (37%) declined intervention for reasons unknown. Of those for whom clinical actions were indicated, 66% completed 3-month follow-up. Randomised controlled trial of the IBD-BOOST intervention Participants (N = 780) were randomised to the intervention (n = 391) or care as usual (n = 389). At 6 months, there were no statistically significant between-arm differences for United Kingdom Inflammatory Bowel Disease Questionnaire [mean difference = −1.67 (95% confidence interval = −4.17 to 0.83), p = 0.19] and Global Rating of Symptom Relief [mean difference = 0.44 (95% confidence interval = −0.56 to 1.43), p = 0.39]. Complier-averaged causal effects analysis demonstrated that participants who complied with the intervention reported better United Kingdom-Inflammatory Bowel Disease Questionnaire scores than ‘would-be’ compliers receiving care as usual (p = 0.03). Subgroup analyses did not identify any statistically significant treatment effect modifiers; 57% completed four sessions or more. Serious adverse events were similar between groups. Process evaluation suggested high facilitator fidelity but low overall participant compliance with the programme. Cost-effectiveness of the IBD-BOOST intervention Per participant cost for the development and delivery of intervention was £151.19. Allocation to inflammatory bowel disease-BOOST intervention resulted in additional 0.016 quality-adjusted life-years (95% confidence interval = 0.002 to 0.030) per participant over 12 months and cost savings of −£304.66 (−803.51; 194.18) for healthcare costs and −£39.48 (−388.09; 309.12) for out-of-pocket costs and time off work over months 3–6 and 9–12. Over the 12-month follow-up, there were cost savings of −£28,633 (95% confidence interval = −51,555 to 18,764) and −£33,568 (−64,421; 26,198) per quality-adjusted life-year gained with inflammatory bowel disease-BOOST from health services and societal perspectives, respectively. Limitations Our samples were predominantly White and English-speaking. Conclusions People with inflammatory bowel disease want help with their symptoms, but health professionals lack capacity. Over half of people with inflammatory bowel disease have faecal incontinence, and 1 in 10 have fatigue, pain and incontinence. Nearly a third want help for all three. Screening using a checklist and algorithm was feasible, and we identified around 40% who could benefit from optimisation of their medical treatment. Inflammatory bowel disease-BOOST had low compliance and did not improve disease-specific quality of life or Global Rating of Symptom Relief in inflammatory bowel disease patients with fatigue, pain and/or incontinence compared with care as usual, but it may help symptoms in people who comply with the programme. The economic evaluation indicated that the inflammatory bowel disease-BOOST intervention improves quality-adjusted life-years, and could be cost-effective, but economic results did not reach statistical significance. Future work Research should explore targeting the intervention to specific subgroups within inflammatory bowel disease and improving adherence to digital interventions. Study registration This study is registered as Optimising medical management ISRCTN15380317 and the RCT as ISRCTN71618461. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0216-20001) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 19. See the NIHR Funding and Awards website for further award information. Plain language summary Our programme addresses the common problems of fatigue, pain and urgent need for a toilet in inflammatory bowel disease. We worked closely with people with inflammatory bowel disease throughout our research. People we interviewed told us that they wanted to try online self-management and to avoid more clinic appointments. Inflammatory bowel disease nurses told us they wanted to help but had limited time. We conducted a survey in 8486 people with inflammatory bowel disease. Over half reported experiencing bowel incontinence in the past week, one in four reported experiencing fatigue and one in five reported pain. Nearly one in three (29%) ‘definitely’ wanted help for all three symptoms. There is a large unmet need for help with fatigue, pain and incontinence. We also developed a checklist and a flow chart to help inflammatory bowel disease nurses manage these symptoms (‘Optimise’). For two in five patients (43%), a test or intervention was indicated that might improve symptoms. Optimise was easy to use, and we are training inflammatory bowel disease nurses to use it. We designed a 12-session online-supported self-management programme (inflammatory bowel disease-BOOST) that people could follow at home, with one phone call from a ‘facilitator’ and online messaging. We recruited and randomly allocated (e.g. tossing a coin) 391 people to try inflammatory bowel disease-BOOST and 389 who had their usual care. After 6 and 12 months, inflammatory bowel disease quality of life, rating of symptom relief, fatigue and pain were not different between these groups. Incontinence and general health-related quality of life were a little better in the inflammatory bowel disease-BOOST group. Most people did not complete as many sessions as we had planned. So, we cannot conclude that inflammatory bowel disease-BOOST is generally helpful in its current format. However, economic evaluation suggested possible cost savings with the intervention and together with the improvement in general health-related quality of life, and it indicated that the inflammatory bowel disease-BOOST intervention could be cost-effective for the self-management of pain, fatigue and faecal incontinence in people with inflammatory bowel disease. We hope that our programme raises awareness of these troublesome symptoms and encourages future work to help fatigue, pain and urgency in inflammatory bowel disease. Scientific summary Background Inflammatory bowel disease (IBD), including Crohn’s disease (CD) and ulcerative colitis (UC), is a lifelong illness. Prevalence is approaching 1% of the population and increasing worldwide. The primary aim of medical management is control of inflammation using medications. However, patients are burdened by symptoms, including fatigue, pain, urgency/faecal incontinence (FI), even when inflammation appears to be under control. These symptoms limit people’s quality of life (QoL) and ability to work and socialise. Patients report that these symptoms are not taken seriously by health professionals and that little help is given. Objectives/research questions Identifying the burden of medically treatable symptoms: 1a. What proportion of people with IBD have troublesome fatigue, pain or urgency/FI? 1b. Of people with these symptoms, what proportion have identifiable pathophysiological contributors which are potentially medically treatable? Addressing symptoms which persist after optimal medical treatment: 2a. Does a nurse-supported, tailored, online self-management programme for fatigue, pain and urgency/FI in IBD (IBD-BOOST) improves the QoL and rating of symptom relief 6 months after randomisation when compared with care as usual (CAU)? 2b. Does IBD-BOOST reduce healthcare costs and personal costs and is the intervention cost-effective? 2c. What factors impact on the completion and effect of the intervention? Design and methods Focus groups and individual interviews: with 40 people living with IBD and focus groups with 45 IBD nurse specialists. Interviews were recorded, transcribed and analysed thematically, with the input of eight people with IBD. A national online and postal questionnaire survey: asking about symptoms of fatigue, pain and urgency/FI and desire for help with these symptoms. Using validated Patient Reported Outcomes Measurement Information System (PROMIS) tools, we defined presence of symptoms as: fatigue: PROMIS fatigue T-score of ≥ 60; pain: PROMIS pain intensity ≥ T-score of 60; PROMIS bowel incontinence: raw score of ≥ 50. Participants also reported disease activity using the relevant PRO-2 score, IBD-Control, anxiety (Generalised Anxiety Disorder-7), depression [Patient Health Questionnaire-9 items (PHQ-9)] and generic QoL [EuroQol-5 Dimension, five-level version (EQ-5D-5L)]. We continued to invite participants until the randomised controlled trial (RCT) (below) was fully recruited. From the survey responses, we conducted a separate analysis to assess the impact of fatigue, pain and FI on generic health-related QoL (measured by the EQ-5D-5L questionnaire), with QoL utility calculated on a scale ranging from 1 (perfect health) to −0.594 (worst health). Linear regression models assessed the associations of these symptoms with QoL controlling for IBD type, sociodemographic characteristics, comorbidities and, in further analysis, for IBD activity and IBD control. We developed a checklist and algorithm (‘Optimise’) to manage medically reversible causes of fatigue, pain and urgency/FI in IBD and conducted a multisite, non-randomised feasibility study of its use in NHS clinics and interviewed nurses implementing the algorithm. We developed IBD-BOOST, a digital interactive facilitator-supported self-management intervention, to address fatigue, pain and urgency/FI in IBD, using theory, evidence and extensive stakeholder input. A trans-symptomatic cognitive behavioural (CB) framework of symptom perpetuation was developed, and a logic model was used to define the intervention techniques. Intervention facilitators (IBD nurse specialists and psychology graduates) were trained and supervised to deliver a 30-minute phone call and 12 weeks of in-site messaging during the intervention. A RCT comparing IBD-BOOST with CAU in relieving symptoms and improving QoL in people with IBD experiencing fatigue, pain and/or urgency/FI: a multicentre, two-arm, parallel-group, RCT; 4449 patients with IBD who rated the impact of fatigue, pain and/or FI as ≥ 5/10 on our previous survey were invited. The intervention was 6 months access to the online IBD-BOOST programme and the control group received CAU. Primary outcomes were UK Inflammatory Bowel Disease Questionnaire (UK-IBDQ) and Global Rating of Symptom Relief (GRSR) at 6 months post randomisation. Subgroup and complier-averaged causal effects (CACE) analyses were pre-specified. Cost-effectiveness analysis of the RCT: healthcare and other resource use and QoL data were collected at randomisation and at 6- and 12-month follow-ups. A cost–utility analysis from health and social care services and societal perspectives was conducted over the 1-year RCT follow-up. Total costs and quality-adjusted life-years (QALYs) were estimated and compared between trial arms using mixed-effects models, adjusting for pre-specified baseline factors. Process evaluation of the RCT: we conducted a mixed-methods process evaluation incorporating quantitative data for patient engagement and intervention fidelity, and we collected qualitative interview data from patient and intervention facilitator perspectives. A blinded randomised study within a trial in the RCT comparing a standard-length participant information leaflet (PIL) with a shortened one. Potential RCT participants were randomised to receive a standard length or shortened PIL electronically for recruitment to the RCT. We evaluated a suicide risk assessment protocol for people who report suicidal ideation on the PHQ-9 depression scale. Latent profile analysis was conducted on pre-randomisation RCT baseline measures of fatigue, pain and incontinence. Multinominal logistic regression, adjusted for inverse probability sample weights, examined associations between profile membership and clinical including inflammatory [faecal calprotectin (FCP)], demographic and psychological factors. Results Interviews People with IBD told us that they wanted help to self-manage their symptoms of fatigue, pain and urgency/FI. The persistent, often stark impact of multiple coexisting symptoms on physical and emotional well-being could force unwanted adjustments and limitations in working, social and intimate arenas of life. Unpredictable symptoms were challenging and impacted each other in negative vicious cycles. People told us what was helpful in self-managing symptoms and about their ways of coping and desired intervention functionality and content. Participants attempt to manage all three symptoms simultaneously. They wanted an accessible online intervention, with symptom management and activity-tracking features. Inflammatory bowel disease nurses told us that they had very limited capacity to assist patients with these symptoms. Most were positive about facilitating our intervention, but they were concerned about finding the time within already overloaded clinical workloads. They were keen to receive adequate training and support if undertaking facilitation. Survey Of 8486 useable responses (7716 online, 770 postal), 4176 reported CD, 4255 UC or other form of IBD; 3281 identified as male and 4883 as female. Median age was 51 years (range 18–92 years); 2550 (30%) reported fatigue, 1766 (21%) pain and 4565 (54%) FI; 925 (10.9%) reported having all three symptoms in the past week. Participants scored severity and impact a mean between 3.3 and 4.8 on a 0–10 scale, with a wide variation; 56% of all respondents ‘definitely’ wanted help for fatigue; 42% wanted help for pain; 53% wanted help for FI and 29% reported ‘definitely’ wanting help for all three symptoms. The mean QoL across all participants was 0.76 (standard deviation 0.23). From the three symptoms, pain was associated with the largest QoL decrement (−0.159), then fatigue (−0.140) and incontinence (−0.048). Co-occurrence of pain and fatigue further reduced QoL. Clear, graded associations were observed between symptom severity and QoL decrements (p < 0.001). Depression and anxiety were associated with further significant QoL decrements (−0.102 and −0.110 for moderate-to-severe anxiety and moderately severe depression, respectively). Worse IBD control and higher IBD activity were associated with lower QoL. Feasibility of Optimise checklist and algorithm About 515 individuals reporting IBD-related symptoms were invited to participate, and 201 (39%) consented; 194/201 (97%) returned the symptom checklist, of whom 157 (81%) returned a postal FCP sample. Five (3%) participants reported ‘red flags’, and 31/157 (20%) participants had a FCP result ≥ 200 μg/g, of whom 12 (8%) were judged to have likely active inflammation when clinical symptoms and disease history were reviewed. The algorithm suggested at least one clinical test or intervention for fatigue, pain or urgency/FI in 67 (43%) participants, of whom 25 (37%) declined. Among 87 participants for whom clinical actions were indicated, 57 (66%) completed follow-up outcomes 3 months after algorithm implementation. Three nurses interviewed found the Optimise algorithm easy to administer. Co-design of the intervention Patient feedback and qualitative interviews refined the website content and functionalities, including the use of visual aids, e-mail reminders and graphical tracking of symptoms. The IBD-BOOST intervention comprised 12 45- to 60-minute sessions. Accessibility was scored as 9.43/10 and ease of use as 8.07/10. Randomised controlled trial of the IBD-BOOST intervention Participants (N = 780) were randomised to the IBD-BOOST intervention (n = 391) or CAU (n = 389). At 6 months, there were no statistically significant between-arm differences for UK-IBDQ (mean difference = −1.67 [95% confidence interval (CI) = −4.17 to 0.83), p = 0.19] and GRSR [mean difference = 0.44 (95% CI = −0.56 to 1.43), p = 0.39]. CACE analysis demonstrated that participants who complied with the intervention reported better UK-IBDQ scores than ‘would-be’ compliers receiving CAU [mean difference = −2.36 (95% CI = −4.44 to −0.28), p = 0.03]. Subgroup analyses did not identify any statistically significant treatment effect modifiers. Serious adverse events were similar between the intervention (n = 55, 14%) and CAU (n = 79, 20%). Economic evaluation of IBD-BOOST intervention Development, maintenance, facilitators’ training and delivery of the IBD-BOOST intervention over the 12 months in the trial were estimated to cost £59,114 or £151.19 per participant allocated to the intervention. The IBD-BOOST intervention resulted in additional per participant 0.016 QALYs (95% CI 0.002 to 0.030) over the 12 months in the study and cost savings of −£304.66 (−803.51; 194.18) for healthcare costs and −£39.48 (−388.09; 309.12) for out-of-pocket costs and time off work over months 3–6 and 9–12. The cost-effectiveness analysis indicated that allocation to IBD-BOOST intervention resulted in cost savings of −£28,633 (95% CI −51,555 to 18,764) per QALY gained from the health services’ perspective and −£33,568 (95% CI −64,421 to 26,198) per QALY gained from the wider societal perspective. The IBD-BOOST intervention had a probability of being cost-effective compared to CAU above 74% at the threshold of £0/QALY and above 95% across willingness-to-pay thresholds of £20,000–30,000 per QALY from both healthcare and societal perspectives. Process evaluation Quantitative analysis found that participants completed a variable number of sessions with a mean of 5/12 sessions completed; 57% completed our pre-defined ‘dose’ of four or more sessions. Time spent was generally less than that recommended. Before the intervention, 30 patient participants were interviewed; 28 patient participants were interviewed after the intervention. They felt that the IBD-BOOST intervention significantly enhanced their symptom management and QoL. They expressed high satisfaction with the programme’s content and structure and hoped that the intervention will be freely available to all IBD patients. Facilitator fidelity was high (87–100% across elements of adherence to in-site messaging and 85.5–92.8% for CB content of messages). Nineteen facilitators were interviewed before and/or after the intervention. They found the training and the manual to be helpful. All were positive about IBD-BOOST, as they felt the content and format addressed patients’ needs. This process evaluation demonstrated that, although the intervention trial outcomes were unsuccessful overall, this was not due to a lack of facilitator intervention fidelity or their level of support for the programme. Study within a trial About 4201 participants were randomised to the standard length (n = 2099) and shortened (n = 2102) PIL arms; 34 e-mail queries were received about the PILs – 18 from those who received the standard and 16 from those receiving the shortened; 708 study within a trial (SWAT) participants were recruited to the RCT – 333 (15.86%) who received the standard length PIL and 375 (17.84%) who received the shortened PIL [odds ratio 1.15, (95% CI 0.98 to 1.35), p = 0.09]. Retention rates in the RCT were not statistically different between groups. Patient Health Questionnaire-9 items suicide risk assessment Overall, there were 58 risk alerts during the trial, 41 of which required risk assessment. All were assessed as low risk or medium risk; none were high risk. We found that the protocol was feasible to use. A refined version is available via our website. Additional randomised controlled trial baseline data analysis A three-profile model was the most appropriate, with moderate symptoms (n = 366, 50%), high symptoms (n = 296, 40.4%) and severe symptoms (n = 70, 9.6%). IBD diagnosis (UC or CD) and inflammation (FCP level) were not associated with profile membership, but comorbidity, time since diagnosis and irritable bowel syndrome were. Men were less at risk of severe symptoms than women. Depression, anxiety, negative illness perceptions, all-or-nothing behaviours and avoidance, all significantly increased the relative risk of being in the high and severe symptoms’ profiles compared with moderate symptoms. Higher self-efficacy was associated with risk reduction. Conclusions Fatigue, pain and urgency/FI are common, troublesome and coexist. People want help for these symptoms. Some people have underlying medical issues which have not been adequately previously addressed. The IBD-BOOST intervention did not improve disease-specific QoL or GRSR in IBD patients with fatigue, pain and/or urgency/FI compared to CAU. People who complied with the intervention appeared to derive benefit and future work should target improving engagement with the intervention. Only 57% completed a ‘dose’ of four or more sessions. However, health economic evaluations found that the IBD-BOOST intervention improves generic health-related QoL and is likely to be cost-effective in IBD patients with fatigue, pain and/or urgency/FI compared to CAU. Implications for health care It is important for healthcare professionals managing patients with IBD to recognise the huge unmet need presented by entangled symptoms of fatigue, pain, urgency and FI, as well as by anxiety and depression. These significantly influence QoL and are not fully explained by disease activity. These symptoms should be proactively enquired about and systematically assessed (and periodically re-assessed) routinely. There is a need for increased capacity, training and access to specialist nurses for people living with IBD to enable addressing symptoms early and effectively. Early intervention may have a role in preventing symptoms becoming chronic and vicious cycles of symptoms becoming established. Clinical systems need an auditable way of flagging the need for assessment and the resulting outcomes and actions. Staff need training and support to achieve this. Research questions to be prioritised How can the online IBD-BOOST self-management programme be optimised to encourage engagement and adherence? What are the biopsychosocial predictors of the presence and progression of symptoms of fatigue, pain and FI? Can these be prevented at an earlier stage after IBD disease onset? Test Optimise for clinical effectiveness. Further research to develop evidence-based information for IBD patients to self-manage symptoms. Is there a role for Artificial Intelligence-operated programmes aimed at individual educational level and learning styles? Study the causes and subtypes of FI and urgency which affect the majority of people with IBD yet very little is known about causes, subtypes and how to manage these symptoms. Study registration This study is registered as Optimising medical management ISRCTN15380317 and the RCT as ISRCTN71618461. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0216-20001) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 19. See the NIHR Funding and Awards website for further award information.

Background About 14.9% of people (9.9 million) registered with primary care practices in England and Wales are prescribed medication for hypertension. However, many do not take their medication as prescribed. To address this problem, we need scalable interventions. Objective To develop a scalable low-cost intervention to support medication adherence in people prescribed medication for hypertension in primary care, and to obtain precise and robust estimates of the effectiveness and cost-effectiveness of the intervention compared with usual care. Design Systematic reviews and meta-analyses; qualitative meta-synthesis; interviews and focus groups; expert consultations; pre-testing study; randomised feasibility trial; randomised controlled trial of effectiveness and cost-effectiveness; economic modelling. Setting and participants Primary care practices in England and Wales. Patients prescribed medication for hypertension with poorly controlled blood pressure. Interventions Very brief intervention delivered by a practice nurse or healthcare assistant followed by a digital intervention (text messaging programme or smartphone app). Main outcome measures Acceptability, feasibility, fidelity and cost of the interventions. Systolic blood pressure. Biochemical and self-reported measures of medication adherence. Results Our systematic reviews showed that both app-based and face-to-face interventions in patients with long-term conditions have a positive effect on medication adherence. The meta-synthesis of published qualitative studies showed that: digital interventions to support medication use were perceived as acceptable and useful; a digital intervention would be more effective if it was personalised and tailored; barriers to using digital interventions included lack of interest, lack of confidence and lack of proficiency and experience in using the technology; digital interventions should be simple, easy to navigate and age-appropriate; patients wanted accurate information on their health condition, potential side effects of medication and health consequences of non-adherence; reminder notifications and a self-monitoring feature were perceived as helpful by some patients; some patients suggested that a digital intervention should enable them to communicate with pharmacies, but practitioners were concerned that this would increase their workload. The interview and focus group study identified several barriers to adherence, including forgetting, unpleasant side effects and reluctance to medicate. A digital intervention to support medication adherence was acceptable to patients if it was user-friendly, the content was tailored to the user, and the privacy of user data was protected. Simple reminder messages for taking medication and reordering prescriptions were considered more useful by patients than those providing information on the benefits of medication or the consequences of non-adherence. Patients preferred to receive feedback on their adherence levels in the form of a simple graph, percentage score or statistic. Practitioners thought that it would be feasible to introduce a digital intervention to patients in a very brief face-to-face consultation. In the pre-testing study, participants reported that the interventions we developed were easy to use and that they would recommend them to others. The feasibility trial showed that the combined intervention was acceptable and that a large cost-effectiveness trial was feasible. The main trial showed no difference between arms in systolic blood pressure or medication adherence at 12-month follow-up. The estimate (95% confidence interval) for the difference in means between arms in self-measured systolic blood pressure at 12 months was −0.61 mmHg (−3.05 to 1.82), p = 0.62 [for intervention vs. control (reference group)]. In the base case analysis, the intervention had a mean incremental cost-effectiveness ratio below the usual willingness-to-pay thresholds in the National Health Service in the United Kingdom. The probability of cost-effectiveness was between 77% and 80% at willingness-to-pay thresholds of £15,000, £20,000 and £30,000 per quality-adjusted life-year, but the confidence intervals are wide and cross zero, indicating some chance that the intervention could be less effective and more costly. Limitations The effectiveness trial was conducted during the COVID-19 pandemic. To reduce the risk of infection, the very brief intervention was delivered by telephone instead of face-to-face, and all study measurements were conducted remotely. This may have led to lower response rates and data quality and lower effectiveness of the intervention. A significant proportion (24%) of participants did not have raised blood pressure at baseline, and self-reported medication adherence was high at baseline, which reduced the possible scope for an intervention effect. Conclusions The findings on effectiveness do not support the commissioning of the intervention in United Kingdom primary care. The cost-effectiveness findings are more equivocal, showing a high probability of being cost-effective at standard United Kingdom willingness-to-pay thresholds, but with some uncertainty. Future work Future research should address the challenge of identifying and recruiting people who are poorly adherent and have raised blood pressure and test the intervention in this group. Variants such as face-to-face delivery, adding a follow-up consultation or a purely digital version could also be investigated. Study registration This study is registered as CRD42017080150; CRD42020164049; ISRCTN12805654; ISRCTN74504989; ISRCTN82013652. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0615-20013) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 18. See the NIHR Funding and Awards website for further award information. Plain language summary About 15% of people (10 million) registered with primary care practices in England and Wales are prescribed medication for hypertension (high blood pressure). However, many do not take their medication as prescribed, which may harm their health and increases health service costs. To address this problem, we need low-cost interventions that can reach all the people who need them. This research programme aimed to develop a new intervention to support people with poorly controlled blood pressure to take their hypertension medication as prescribed and to assess how well it worked and how much it cost compared with usual care only. The findings would inform a decision on whether to introduce the intervention in primary care practices in the United Kingdom. We used a range of different research methods, including literature reviews, interviews and focus groups, and randomised controlled trials. The intervention we developed has two parts: a very brief intervention delivered by a practice nurse or healthcare assistant by telephone, followed by a digital intervention (individually tailored text messages for up to 420 days or a smartphone app). In the main trial, 537 participants received either the intervention or just usual care and were followed up 12 months later. The intervention was low cost, but it was not effective: the results showed no difference in blood pressure or medication adherence between the two groups at 12 months. Not all participants had raised blood pressure at the start of the study, and many were already taking their medication as prescribed; this may explain why the intervention did not work. The findings suggest that the intervention should not be introduced in United Kingdom primary care practices before showing that it works in patients who do not take their medication as prescribed and have raised blood pressure. Scientific summary Background Hypertension is a global health challenge accounting for 8.5 million deaths worldwide despite the availability of low-cost pharmaceutical treatment. About 14.9% of people (9.9 million) registered with primary care practices in England and Wales are prescribed medication for hypertension. However, many patients with hypertension do not take their medications as prescribed – 42% in the UK according to one study. Non-adherence to antihypertensive medication is associated with increased risk of suboptimal blood pressure (BP) control, complications and all-cause mortality, and increased healthcare costs. Primary care practitioners have an important role in supporting patients to adhere to their prescribed medication. However, they lack time to provide ongoing support for adherence, and their time is expensive. A potential solution is for a practitioner such as a practice nurse to deliver a very brief intervention (VBI) during a consultation and to use a digital intervention such as text messaging or a smartphone app to support subsequent adherence. About 96% of UK adults use a mobile phone, and in 93% of cases, this is a smartphone; the corresponding figures for those aged 65 and above are 88% and 77%. This suggests that digital interventions have the potential to reach the majority of this population. Digital interventions have several other advantages over traditional interventions: they can be fully automated; provide information that is highly tailored to the individual; be interactive; be available at any time; deliver support in real time; deliver support with high fidelity; and be easily updated. Recent meta-analyses have reported promising findings for the effectiveness of nurse-led and digital interventions to improve medication adherence and reduce BP in people with hypertension. The interventions examined in these reviews varied widely in content and delivery, and the digital interventions that have been evaluated to date have not made full use of individual tailoring, interactivity and other features that may increase user engagement and potential effectiveness. Objectives The PAM programme (Programme on Adherence to Medication) aimed to develop and evaluate an intervention to support medication adherence that combines a VBI from a practice nurse or healthcare assistant with a digital intervention (text messaging programme or smartphone app). Such an intervention would be inexpensive to deliver, scalable and potentially cost-effective. The objectives were: To develop a scalable low-cost intervention to support medication adherence in people prescribed treatment for hypertension in primary care. To evaluate the acceptability and feasibility of the intervention and the feasibility of conducting a (cost-)effectiveness trial. To provide precise and robust estimates of the effectiveness and cost-effectiveness of the intervention compared with usual care (UC). To develop an economic model of the cost-effectiveness of medication adherence interventions. To inform a decision on whether to implement the intervention in primary care. Methods The target group is patients in primary care practices in England and Wales who do not take their antihypertensive medication as prescribed and have raised BP. Methods used were: Systematic reviews of randomised controlled trials of app-based (9 trials) and face-to-face interventions (20 trials) to support medication adherence, with random-effects meta-analyses. A meta-synthesis of 30 published qualitative studies of adults taking medication for cardiovascular-related long-term health conditions (e.g. type 2 diabetes, hypertension) and/or healthcare practitioners who treat patients with cardiovascular conditions who were asked about their views and experiences of digital interventions to support medication adherence. Interviews with 11 healthcare practitioners [6 practice nurses, 2 healthcare assistants, 2 practice pharmacists, 1 general practitioner (GP)] and 6 patients, and 4 focus groups, with a total of 14 patients, to gather views on the acceptability and content of digital interventions for medication adherence. Expert consultations with 2 commissioners, 2 academics, 2 nurses, 2 patients and 10 GPs. Participants were e-mailed a description of the proposed intervention, a description of the proposed design of the randomised feasibility study and a link to an online questionnaire that asked their views on the delivery mode and content of the intervention and on the proposed feasibility study. Commissioners were asked about the evidence needed to inform a decision whether to commission the intervention. Pre-testing study of early versions of the digital interventions to assess acceptability. The text messaging intervention was used by 22 patients with hypertension for 28 days, and four of them also used the smartphone app for an additional 28 days. Data were collected by weekly telephone interviews, questionnaires and log files showing how they had used the interventions. Randomised feasibility trial to assess the feasibility and acceptability of the PAM intervention and the feasibility of conducting a large cost-effectiveness trial. Patients with hypertension who had raised BP and were non-adherent to their prescribed medication as indicated by their practice records and practice GP assessment were eligible for the study. One hundred and one eligible patients from nine general practices in the East of England and London were randomised to receiving the PAM intervention (N = 61) or UC only (N = 40). Randomised controlled trial to estimate the effectiveness and cost-effectiveness of the PAM intervention to improve medication adherence and reduce BP compared with UC only, to inform a decision on whether to implement the intervention in primary care (‘main trial’). A total of 573 eligible patients from 57 practices in England and Wales were individually randomised, stratified by practitioner, to the PAM intervention or control (UC only) and followed up at 12 months. The primary outcome was systolic blood pressure (SBP). The analysis was based on 537 participants. Within-trial economic analysis of the cost-effectiveness of the PAM intervention compared with UC alone. The main cost-effectiveness measure was the incremental cost per quality-adjusted life-year (QALY) gained, and the analysis included extensive deterministic and probabilistic sensitivity analyses. Results Systematic reviews The findings from the meta-analysis of app-based interventions showed that, at follow-up, patients in the intervention groups were more likely to self-report adherence to medication than those in the comparator groups [odds ratio 2.12, 95% confidence interval (CI) 1.64 to 2.75, n = 988, p < 0.0005]. None of the behaviour change techniques (BCTs) used in the interventions was significantly associated with intervention effect size. In the meta-analysis of face-to-face interventions, statistically significant pooled effects were found favouring the intervention arm over the control arm for several Medication Event Monitoring System measures of adherence, for example, percentage of prescribed doses taken on time over a period of 3 weeks to 2 months [mean difference (MD) 9.34, 95% CI 4.36 to 14.33, n = 3,667, p = 0.0002]. We also found significant between-arm effects for a self-report measure of adherence (Morisky scale). The impact of BCTs on intervention effectiveness could not be estimated as the analyses were underpowered. Taken together, these reviews supported our proposal to use face-to-face and digital components in the PAM intervention. However, we were unable to identify promising BCTs for potential inclusion in the proposed intervention. Meta-synthesis of previous qualitative studies The main findings from the meta-synthesis of published qualitative studies were: digital interventions to support medication use were perceived as acceptable and useful; a digital intervention would be more effective if it was personalised and tailored; barriers to using digital interventions included lack of interest, lack of confidence and lack of proficiency and experience in using the technology; digital interventions should be simple, easy to navigate and age-appropriate; patients wanted accurate information on their health condition, potential side effects of medication and health consequences of non-adherence; reminder notifications and a self-monitoring feature were perceived as helpful by some patients but unnecessary by others; some patients suggested that a digital intervention should enable them to communicate with pharmacies, but practitioners were concerned that this would increase their workload. Interviews and focus groups with practitioners and patients This study identified several barriers to adherence, including forgetting, unpleasant side effects and reluctance to medicate. A digital intervention to support medication adherence, either via text messages or smartphone app, was acceptable to patients, provided that it was user-friendly, the content was tailored to the user and the privacy of user data was protected. Simple reminder messages for taking medication and reordering prescriptions were considered more useful by patients than those providing information on the benefits of medication or the consequences of non-adherence, which were favoured by practitioners. Rather than messages of encouragement, patients preferred to receive feedback on their adherence levels in the form of a simple graph, percentage score or statistic. All the practitioners thought that it would be feasible to introduce a digital intervention to patients in a very brief face-to-face discussion during a primary care consultation. Expert consultations There was substantial similarity of views between the different stakeholders. They found the concept of a VBI delivered face-to-face by a healthcare practitioner acceptable. However, they felt that it was not feasible to address possible reasons for medication non-adherence in a VBI, and that the intervention should be limited to emphasising the importance of taking medication as prescribed and signposting the patient to a digital intervention. Pre-testing study Participants reported that the interventions were easy to use and that they would recommend them to other people. They were satisfied with the frequency of the messages and the content of the daily reminder and weekly query messages, but they were somewhat less satisfied with the content of the daily non-reminder (advice) messages. The response rate to the query messages was 100%, indicating a high degree of engagement. Randomised feasibility trial All 101 participants had their BP measured at baseline, and the vast majority provided a urine sample for chemical adherence testing. Participants were on average 65.8 years of age, 54% male, with a substantial minority (35%) from the most deprived areas, based on practice postcode. Baseline characteristics were similar in the two arms. At 3-month follow-up, 83% of participants had their BP measured and provided a urine sample, and the percentage was similar in the two arms. Ninety-two per cent of participants randomised to the intervention arm opted to receive text messages, and 8% opted to use the app. Ninety per cent responded to the tailoring questions which were administered digitally. Four intervention participants actively disengaged from the digital intervention by sending a STOP message. Seventy-two per cent continued to use the digital intervention for at least 1 month. The post-trial interviews showed that intervention participants found the intervention to be acceptable. Participants were satisfied with the baseline and follow-up consultations and the study procedures, and there were no concerns among control participants about being randomised to this arm. Practitioners also confirmed that the study procedures and intervention were acceptable. From baseline to follow-up, mean SBP reduced from 146.9 mmHg to 136.9 mmHg in the intervention arm compared with no change in the control arm (adjusted MD 9.2 mmHg, 95% CI 5.7 to 12.6), and biochemically measured adherence increased to a greater extent in the intervention arm than in the control arm, suggesting that the intervention was potentially effective. The findings from this trial showed that the intervention was acceptable to participants and that most offered the digital intervention used it, at least in the short term. The trial procedures were demonstrated to be practicable. Together with the findings on trial uptake and retention rates, this suggested that a large cost-effectiveness trial was feasible. Main trial Baseline characteristics were similar in the two arms. The majority of participants were recruited from practices in the East of England. Similar to the feasibility trial, 56% were male and mean age was 66.5 years. The vast majority categorised themselves as being of White ethnicity, but there was a range of deprivation levels, based on participant home postcode. Mean BP, obtained from practice records before randomisation, was 145/82 mmHg, again similar to the sample in the feasibility trial. Of participants, 75.8% had a BP reading above the accepted cut-off of 140/90 mmHg and 71.7% had a SBP reading above 140. We were, therefore, partially successful in recruiting a sample of participants who had a raised BP even though they were prescribed antihypertensive medication. The estimate (95% CI) for the difference in means between arms in the primary outcome of self-measured SBP at 12 months was −0.61 mmHg (−3.05 to 1.82), p = 0.62 [for intervention vs. control (reference group)]. Thus, the estimated effect was very small, and the detectable effect size of 5 mmHg did not fall within the CI. The estimate for the difference in means between arms for SBP obtained from practice records was 1.04 mmHg (−1.56 to 3.65), p = 0.43. Thus, the intervention appeared to have no effect on SBP. There was also no effect of the intervention on biochemically measured adherence. Of the 392 participants who provided a urine sample at follow-up, 388 (99.0%) were found to have at least one antihypertensive medication (or metabolite) in their urine. Based on the urinalysis, 94.3% of participants were categorised as ‘fully adherent’, 4.8% as ‘partially adherent’ and only 0.9% as ‘non-adherent’. There was no difference in these percentages between trial arms. Self-reported adherence at 12 months was also high, with no difference between trial arms. The mean score on the five-item Medication Adherence Report Scale questionnaire was 23.8 [standard deviation 1.7] out of a maximum score of 25 (based on 387 participants who provided 12-month data on this scale). Of the intervention participants, 212 (77.9%) opted to receive text messages; 60 (22.1%) opted to use the app, and 39 of these became active users. On average, participants were satisfied with the combined intervention (VBI plus digital intervention) and thought that it was acceptable and effective. However, 61.0% used the digital intervention for less than 3 months, with the main reasons being not needing any further support and finding the messages annoying. Economic analysis The total mean intervention cost per patient was £30. In the base case analysis, the intervention was found to be cost-effective compared with UC, with a mean estimated incremental cost-effectiveness ratio (ICER) of £1231 per QALY gained (95% CI −£13,156 to £19,535) and mean incremental net monetary benefit (INMB) of £289 (95% CI −£498 to £1026) at a willingness-to-pay (WTP) threshold of £15,000/QALY. The INMB rose to £400 (95% CI −£643 to £1383) and £621 (95% CI −£933 to £2104) for £20,000/QALY and £30,000/QALY, respectively. The probability that the intervention is cost-effective was between 77% and 80% for these WTP thresholds. Limitations The effectiveness trial was conducted during the COVID-19 pandemic. To reduce the risk of infection, the VBI was delivered by telephone instead of face-to-face, and all study measurements were conducted remotely. This may have led to lower response rates and data quality and lower effectiveness of the intervention. A significant proportion (24%) of participants did not have raised BP at baseline, and self-reported medication adherence was high at baseline, which reduced the possible scope for an intervention effect. Conclusions The combination of a VBI delivered by a practice nurse or healthcare assistant and a digital intervention was acceptable to both patients and practitioners. However, although the feasibility trial showed promising results, the main trial showed no effect of the intervention on medication adherence or SBP at 12 months. The cost-effectiveness findings showed a mean ICER below the usual WTP thresholds, but the CIs are wide and cross zero, indicating some chance that the intervention could be less effective and more costly compared with UC only. The effectiveness findings do not support the commissioning of the intervention in UK primary care. Future research should address the challenge of identifying and recruiting people who are poorly adherent. If it is feasible to recruit patients who are non-adherent to their prescribed antihypertensive medication and have raised BP, the intervention could be tested with the VBI delivered remotely or face-to-face. Variants such as adding a follow-up consultation or testing a purely digital version could also be investigated. The economic model developed for this programme can provide the basis for future economic evaluations of similar interventions across a range of different conditions. Study registration This study is registered as CRD42017080150; CRD42020164049; ISRCTN12805654; ISRCTN74504989; ISRCTN82013652. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0615-20013) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 18. See the NIHR Funding and Awards website for further award information.

Background Hazardous prescribing is an important and expensive preventable cause of patient harm, which can be identified using electronic healthcare records. In our Programme Grant we investigated the large-scale roll-out of a type of clinical decision support called OptimiseRx® (First Databank, Hearst Health, Exeter, United Kingdom) (which had not been independently evaluated previously) and two large-scale roll-outs of pharmacist-led IT-based intervention (an intervention that had been shown to reduce hazardous prescribing in a cluster randomised trial). Objectives To estimate harm and economic impact of specific hazardous prescribing events. To investigate whether a clinical decision support intervention (OptimiseRx) was associated with reduction in hazardous prescribing. To investigate whether a pharmacist-led IT-based intervention was associated with reduction in hazardous prescribing. To undertake economic and process evaluations of both interventions. Design and methods The harm and economic impact of hazardous prescribing events were investigated using data from published sources to generate estimates of patient outcomes and cost associated with 11 prescribing safety indicators. An incomplete (not randomised) stepped-wedge design study with control groups was used to investigate the effectiveness of OptimiseRx (409 general practices). Incremental cost-effectiveness ratios were generated for OptimiseRx compared with standard care (cost per hazardous prescribing event prevented), subgroup analysis: cost per quality-adjusted life-year. Incomplete stepped-wedge design studies [East Midlands study (343 and 115 general practices); whole of England study (725 and 179 general practices)] investigated the effectiveness of pharmacist-led IT-based intervention. Incremental cost-effectiveness ratios were generated for pharmacist-led IT-based intervention compared with standard care (cost per hazardous prescribing event prevented, cost per quality-adjusted life-year). Data from interviews, focus groups, a survey and stakeholder events were analysed qualitatively to identify factors influencing the sustained implementation and operation of OptimiseRx or pharmacist-led IT-based intervention. Setting and participants The interventions took place in general practices in England. The process evaluation involved clinicians, patients and other stakeholders in England. Interventions Clinical decision support intervention (OptimiseRx). Pharmacist-led IT-based intervention. Main outcome measures Changes in hazardous prescribing rates. Cost per quality-adjusted life-year. Data sources Clinical Practice Research Datalink. ResearchOne. General practice clinical records. Process evaluation with clinicians, patients and other stakeholders. Results All the hazardous prescribing events studied were associated with loss of quality-adjusted life-years. The costliest for the National Health Service were prescriptions of antipsychotics to patients with dementia (£1168 per patient) and anticoagulants in combination with nonsteroidal anti-inflammatory drugs (£1020 per patient). Of 409 practices included in the clinical decision support study, 227 (56%) were intervention practices. Deployment of OptimiseRx was associated with a 10% reduction in exposure to hazardous prescribing after 2 years (odds ratio 0.90, 95% confidence interval 0.88 to 0.92). OptimiseRx costs £14–48 per hazardous prescribing event prevented, depending on time point and patient volume. Based on two indicators only, OptimiseRx costs £236 per quality-adjusted life-year gained at 6 months and becomes dominant (quality-adjusted life-year increasing; cost saving) at 24 months. Of 343 general practices included in the East Midlands pharmacist-led IT-based intervention evaluation, we completed further data extraction for 115 (33.5%) of these. Pharmacist-led IT-based intervention was associated with a decrease in hazardous prescribing of 18% (adjusted odds ratio 0.82, 95% confidence interval 0.78 to 0.86) at 6 months, and 22% (adjusted odds ratio 0.78, 95% confidence interval 0.73 to 0.84) at 24 months post intervention. Pharmacist-led IT-based intervention generates approximately 0.15 quality-adjusted life-years per practice extra compared with no intervention. Expected savings do not fully offset delivery costs, so pharmacist-led IT-based intervention costs just over £1000 per practice. Pharmacist-led IT-based intervention was estimated to be £7449. With 12- and 24-month intervention effects, health gains rise to over 0.2 quality-adjusted life-years and net expenditure falls to < £800 per practice; around £3500 per quality-adjusted life-year gained. Of 725 practices that implemented pharmacist-led IT-based intervention nationally, we had usable data for only 179 because of the COVID-19 pandemic. For these practices, pharmacist-led IT-based intervention was associated with a reduction in hazardous prescribing of 11% (adjusted odds ratio 0.89, 95% confidence interval 0.86 to 0.92) at 6 months post intervention. The process evaluation studies demonstrated the importance of collaboration and system-level support for the clinical decision support and pharmacist-led IT-based interventions. Limitations The main limitation was that we could not robustly evaluate the impact of OptimiseRx on serious harm outcomes and the analysis was limited for pharmacist-led IT-based intervention because of the COVID-19 pandemic. Future work Further research is needed to determine whether interventions that reduce hazardous prescribing, also reduce serious harm to patients. Conclusions Large-scale implementation of OptimiseRx was associated with reduced hazardous prescribing, at less than £50 per hazardous prescribing event avoided. Large-scale implementation of pharmacist-led IT-based intervention was also associated with reduced hazardous prescribing with a cost per quality-adjusted life-year of £7449. It should be noted, however, that these were quasi-experimental studies using observational data and so the findings may have been affected by unknown confounding. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research programme (NIHR award ref: RP-PG-1214-20012) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 15. See the NIHR Funding and Awards website for further award information. Plain language summary There is usually some risk of harm from prescribed medicines, but there are ways of reducing this risk. The purpose of this research was to test two new ways of doing this in general practices. We did some background research to estimate the harm (and costs) associated with certain types of hazardous prescribing. We then looked at computer software called OptimiseRx® which provides ‘alerts’ on the computer screen to warn general practitioners if they are about to prescribe a medicine that may be harmful to the patient. The software has been used in many parts of England since 2014 but had not previously been studied independently. We looked at changes in potentially harmful prescribing from 2011 to 2019 in 409 general practices, including 227 using OptimiseRx. After 2 years, there was a 10% reduction in hazardous prescribing in practices using the software. In another two studies, we looked at the roll-out of an approach called pharmacist-led IT-based intervention, which involves searching general practitioner computer records to find patients who have received medicines which might harm them, and a pharmacist working with the general practice to correct the problems found. In a previous study (a randomised trial) we showed that pharmacist-led IT-based intervention reduced potentially harmful prescribing. The first study looked at changes in potentially harmful prescribing from 2013 to 2019 in relation to the roll-out of pharmacist-led IT-based intervention to 115 general practices in the East Midlands; we found that potentially harmful prescribing went down by 18% after 6 months and 22% after 2 years. The second study looked at changes in potentially harmful prescribing from 2018 to 2020 in relation to the roll-out of pharmacist-led IT-based intervention nationally; for 179 general practices included in the study there was an 11% reduction in hazardous prescribing after 6 months. We talked to patients, health professionals and healthcare leaders about these ways to improve prescribing safety, finding it was important for long-term success, that people worked together and senior leaders gave support. Our studies appear to show that OptimiseRx and pharmacist-led IT-based intervention can reduce potentially harmful prescribing in general practices, but we cannot be sure because these were ‘observational’ studies, and the results may have been influenced by things we were not able to control for. Scientific summary This section contains material reproduced from Penner LS, Gavan SP, Ashcroft DM, Peek N, Elliott RA. Does coprescribing nonsteroidal anti-inflammatory drugs and oral anticoagulants increase the risk of major bleeding, stroke and systemic embolism? Br J Clin Pharmacol 2022;88:4789–811. https://doi.org/10.1111/bcp.15371. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text. Background Hazardous prescribing is an important preventable cause of harm to patients. Prescribing safety indicators (PSIs) describe potentially hazardous prescribing events (HPEs) and many of these can be identified using electronic healthcare records. There are two ways in which PSIs have been used in primary care to reduce potentially hazardous prescribing. The first is to incorporate them into computerised clinical decision support (CDS) alerts used at the point of prescribing decision-making; systematic reviews have shown benefits but there have been few studies in primary care. The second is to run computer searches to identify patients with hazardous prescribing and to intervene to reduce risk; we have previously shown the benefits of such an approach in a cluster randomised trial using pharmacists [the pharmacist-led IT-based intervention (PINCER) intervention]. In our Programme Grant we investigated the large-scale roll-out of a type of CDS called OptimiseRx® (First Databank, Hearst Health, Exeter, UK) (which had not been independently evaluated previously) and two large-scale roll-outs of PINCER. Work package 1: Estimating harm and economic impact of hazardous prescribing; validating codes and algorithms Objectives To estimate harm associated with specific HPEs and economic impact of the HPEs in the PINCER intervention. To validate code lists and algorithms for identifying serious harm outcomes (SHOs) associated with the HPEs. Methods Project 1 We conducted two population-based cohort studies using linked primary care [Clinical Practice Research Datalink (CPRD)] and Hospital Episode Statistics (HES) records to examine the risk of serious harm associated with (a) coprescribing nonsteroidal anti-inflammatory drugs (NSAIDs) and oral anticoagulants (OAC) and (b) antipsychotic prescribing in people with dementia. Project 2 Cohort-level state transition (Markov) models were developed to generate estimates of patient outcomes [quality-adjusted life-years (QALYs)] and cost to the NHS in England associated with each of 11 HPEs. Project 3 We validated computer code lists and algorithms used to define five SHOs associated with the HPEs. Four general practices ran computer searches for these SHOs. Clinicians reviewed patient records to identify confirmatory evidence. From this we calculated the positive predictive value (PPV) for each SHO. We compared the number of SHOs identified within primary and secondary care data to estimate percentage of SHOs missed when using either data-source alone. Results Project 1 Compared with OAC therapy alone, concomitant prescription of NSAIDs with OACs was associated with increased risk of gastrointestinal (GI) bleeding [hazard ratio (HR) 3.01, 95% confidence interval (CI) 1.63 to 5.55], stroke (HR 2.71, 95% CI 1.48 to 4.96) and major bleeding (HR 2.77, 95% CI 1.84 to 4.19). Compared with non-exposure, antipsychotic exposure among people with dementia was associated with elevated risks for all outcomes investigated except ventricular arrhythmia. Project 2 All HPEs were associated with QALY loss. The costliest HPEs were prescribing antipsychotics to patients with dementia (£1168) and anticoagulants with NSAIDs (£1020). Project 3 The PPV for the code lists and algorithms for identifying SHOs was at least 70% when the clinical reviewer had ‘moderate certainty’ and between 79% and 95% when they had ‘high certainty’. There were considerable differences between primary care records (CPRD) and secondary care records (HES) in the proportion of SHO cases ‘missed’ when using each type of record alone. Limitations Project 1 Residual confounding cannot be excluded from our studies using observational data. Project 2 A key uncertainty in the economic analysis is how long people subjected to HPEs are exposed to the risk. Project 3 We used a relatively small number of general practices to undertake validation of the SHO codes. Conclusions Coprescription of NSAIDs with OACs was associated with increased risk of GI bleeding. Antipsychotic prescribing in people with dementia was associated with a range of serious adverse outcomes. Hazardous prescribing events are associated with loss of QALYs. The algorithm and codes used to identify the SHOs associated with the PINCER indicators are reasonably specific. The comparison of SHO codes from primary care (CPRD) and secondary care (HES) showed that many serious harms were not identified by either data set alone. Work package 2: Evaluating OptimiseRx clinical decision support Objectives To describe the nature and incidence of prescribing safety alerts generated by OptimiseRx CDS and frequency of alert overrides. To evaluate whether OptimiseRx was associated with reduced hazardous prescribing rates. To assess the cost-effectiveness of OptimiseRx compared with standard practice, and cost per extra QALY in a subgroup analysis of prescribing indicators. Methods Project 1 We conducted a retrospective observational study. We analysed data from general practices contributing to The Phoenix Partnership ResearchOne database (1 January 2014–31 December 2019), some of which deployed OptimiseRx during that period. We estimated the trends in numbers of alerts per practice per month (pppm) over time, and trends in overriding of alerts. Project 2 We analysed 48 PSIs in OptimiseRx. We used patient-level, de-identified primary care data from ResearchOne database from all practices that contributed data between 2011 and 2019. Study design was an incomplete (not randomised) stepped-wedge design with control groups. Intervention practices implemented OptimiseRx between 2014 and 2018 and contributed at least 12 months of post-intervention data; some practices were censored (see synopsis for details) and all others were counted as controls. We derived composite outcomes for all prescribing HPEs, all monitoring HPEs, and three types of prescribing HPE at quarterly intervals. Outcomes were analysed using binomial mixed regression analysis with a random intercept for each practice, adjusting for secular trends and seasonal effects. Assessment points were 6, 12 and 24 months after intervention start. Project 3 We used similar study design, data, indicators and outcomes as work package 2 (WP 2), Project 2. The unit of analysis for Project 3 was the HPE. We estimated the cost of the prescribing safety component of OptimiseRx by apportioning the annual licence-fee according to the proportion of active indicators that relate to safety. We estimated cost per HPE prevented, compared with standard care, by dividing costs for an average practice by expected HPE reduction. We derived a partial estimate of cost-per-QALY-gained for OptimiseRx, compared with standard care, for the two indicators overlapping with PINCER. Results Project 1 In 258 general practices that deployed OptimiseRx, 146,993 interruptive alerts were generated during the study period, of which 109,687 alerts (74.6%) were overridden. On average, 11.1 interruptive alerts were generated by OptimiseRx pppm. In the first 24 months of deployment, the average number of alerts pppm decreased from 12.21 to 10.07, while the average override rate did not change. Project 2 From a total of 409 practices, 31 of the 258 that deployed OptimiseRx were censored, leaving 227 (56%) intervention practices. The percentage of people exposed to at least one prescribing HPE decreased over time, and this trend was accelerated after deployment of OptimiseRx. Two years after the intervention began, OptimiseRx was associated with an additional 10% reduction in exposure to prescribing HPEs [odds ratio (OR) 0.90, 95% CI 0.88 to 0.92] and a 6% reduction in exposure to monitoring HPEs (OR 0.94, 95% CI 0.93 to 0.95). Project 3 We estimate that OptimiseRx costs £44–48, £14–16 and £22–24 per HPE prevented, at 6, 12 and 24 months, respectively. In a subgroup analysis assessing the only two indicators for which we could estimate long-term cost and QALY impacts, OptimiseRx has an incremental cost-effectiveness ratio (ICER) of £236 per QALY gained at 6 months and becomes dominant (QALY-increasing; cost saving) at 24 months. Limitations While our evaluation of OptimiseRx used a quasi-experimental design, it was an observational study and so the reductions in hazardous prescribing could have been due to unknown confounding factors. We did not achieve linkage of primary care data to HES and deaths data and so were not able to evaluate the impact of OptimiseRx on SHOs. Conclusions The majority of OptimiseRx alerts were overridden, in common with most other studies of CDS implementations. OptimiseRx was associated with a reduction in hazardous prescribing. Based solely on the two indicators for which we could estimate long-term impacts, OptimiseRx has an ICER of £236 per QALY gained at 6 months and was QALY-increasing and cost saving at 24 months. Work package 3: Evaluating pharmacist-led IT-based intervention Objectives To evaluate the effectiveness of a PINCER when widely implemented in general practices to: Reduce the prevalence of patient exposure to hazardous prescribing. Reduce the incidence of serious harm in patients at risk of hazardous prescribing. To assess the cost-effectiveness of PINCER compared with standard care, and cost per extra QALY. Methods Project 1 This was an incomplete (not randomised) stepped-wedge design without controls. For consenting practices from our previous East Midlands PINCER roll-out, we collected quarterly data for 11 HPEs between 2013 and 2019, and composite indicators were calculated. To model the time series, we represented the time post intervention as a categorical variable coding each post-intervention quarter; all pre-intervention quarters were assigned a single reference level for the treatment effect. Calendar time (to account for secular trends) was included as a covariate; random intercept terms for general practice allowed for within-practice correlations, seasonality included. We used a mixed effects logistical regression; effect sizes are presented as (adjusted) OR (aOR) compared with the pre-intervention period. We determined the cost-effectiveness of the PINCER intervention compared with standard care by combining the effectiveness in HPE reduction and intervention costs with the effect of the individual HPEs on patient outcomes and healthcare costs, to estimate the overall effect on costs and QALYs. The economic analysis generated ICERs; cost per HPE avoided and QALYs generated using PINCER compared with standard practice, at practice level, from the perspective of the NHS. Project 2 We obtained data from CPRD for general practices implementing PINCER as part of the national roll-out between 2018 and 2021, along with implementation date. We used a similar design and approach to analysis as WP 3, Project 1, but also examined changes in SHOs associated with the HPEs. Results Project 1 We extracted data from 115 (33.5%) of the 343 general practices included in our East Midlands PINCER roll-out. For the composite indicator, PINCER was associated with a decrease in the rate of hazardous prescribing of 18% [aOR 0.82, 95% CI 0.78 to 0.86] at 6 months, 23% (aOR 0.77, 95% CI 0.73 to 0.81) at 12 months and 22% (aOR 0.78, 95% CI 0.73 to 0.84) at 24 months post intervention. The base-case analysis, using 6-month observed effects, shows that PINCER generates approximately 0.15 QALYs per practice extra compared with no intervention, with a net cost of just over £1000 per practice, giving an ICER of approximately £7500 per QALY gained compared with no intervention. The probability that PINCER represents good value for money when QALYs are valued at £20,000 each is 0.947. Project 2 Of the 2817 practices that implemented PINCER as part of the national roll-out, 725 were included in the CPRD data with HES linkage. To minimise the effects of the COVID-19 pandemic on the analysis, we used pre-lockdown data only, up until March 2020, and only 179 practices had at least 6 months’ follow-up data. These 179 practices had an 11% reduction in hazardous prescribing overall (aOR 0.89, 95% CI 0.86 to 0.92) and an 18% reduction in hazardous prescribing associated with risk of GI bleeding (aOR 0.82, 95% CI 0.78 to 0.86). There was no change in SHOs when using primary and secondary care data combined, but when using secondary care data (HES) alone there was a 5% reduction (aOR 0.95, 95% CI 0.91 to 0.99). There was no impact on hospitalisations or deaths. Limitations While our evaluations of PINCER used quasi-experimental designs, these were observational studies and so the reductions in hazardous prescribing and SHOs could have been due to unknown confounding factors. Only a third of eligible general practices provided data for our more detailed evaluation of the roll-out of PINCER across the East Midlands. As a result of the COVID-19 pandemic, only a quarter of potentially eligible practices in the CPRD database had sufficient usable follow-up data for our analysis. Conclusions The implementation of PINCER across the East Midlands, and subsequently across England was associated with a reduction in hazardous prescribing. The implementation of PINCER across the East Midlands had an estimated ICER of approximately £7500 per QALY gained compared with no intervention. The implementation of PINCER across England was associated with a small reduction in SHOs, but only when examining secondary care data alone. Work package 4: Process evaluation Objectives To develop learning about the key technical, organisational, contextual and policy factors that influence the sustained implementation and operation of CDS and PINCER (or similar approaches) in primary care. To generate overarching recommendations for optimal reach, implementation and sustainable use. Methods Project 1 (clinical decision support) We conducted 39 semistructured interviews with stakeholders involved in the implementation and use of CDS across 2 English regions, with 11 follow-up interviews. We used all the interviews for a thematic analysis using Normalisation Process Theory (NPT) and 23 interviews for interpretive data analysis using socio-technical theory. Project 2 (pharmacist-led IT-based intervention) A longitudinal process evaluation incorporated 50 semistructured interviews across 27 organisations, and 18 follow-up interviews. A national electronic survey accompanied the interviews, involving 81 general practices and 3 Clinical Commissioning Groups, across 11 NIHR Clinical Research Networks. Data analysis used an inductive approach. Findings were mapped to the four NPT constructs. Project 3 We undertook a consolidated learning exercise including: A documentary review of findings from other WPs within the Avoiding patient harm through the application of prescribing safety indicators in English general practices (PRoTeCT) study alongside relevant policies and studies. Semistructured interviews with 27 key stakeholders, with key themes mapped to the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework. Two sequential online consensus-building workshops involving 20 stakeholders. Results Project 1 (clinical decision support) Participants understood the CDS as having value and a role in both medication safety and cost saving. Most of the ‘work’ required to embed and sustain the impact of CDS was required at system level in supporting general practices to use the tool. General practitioners (GPs) reported ignoring CDS alerts where they were perceived as inaccurate, overwhelming or irrelevant for decision-making. Project 2 (pharmacist-led IT-based intervention) Participants reported that ‘top-down’ approaches were used to raise awareness of PINCER. Facilitators for uptake included collaboration across health organisations using recent impact and uptake data. Inclusion of PINCER indicators in national policy helped to raise awareness and influenced uptake and receiving appropriate training and support also influenced ongoing engagement. Overall, public perceptions of PINCER were positive with participants seeing potential benefits for medication safety and recognising the value of utilising pharmacists to deliver PINCER. Project 3 The RE-AIM framework was used to synthesise findings into five main themes: fitting into current context; engaging hearts and minds; building resilience; achieving engagement with secondary care to align prescribing safety guidelines and emphasising complementary use of PSI-based interventions. A clear strategy emerged to align services and work in teams across the healthcare system for a consistent focus on prescribing safety, emphasising the role of organisational support at multiple levels. Limitations There were recruitment challenges that affected our sampling, and while we observed a wide spectrum of responses, some viewpoints may have been missed. Conclusions Successful implementation requires appropriate engagement with, and support for, general practices. Overall conclusions In keeping with systematic review evidence for the effectiveness of CDS, large-scale implementation of OptimiseRx CDS was associated with reduced hazardous prescribing, at less than £50 per HPE avoided. Having previously demonstrated the effectiveness of PINCER in a cluster randomised trial, large-scale implementation of PINCER was also associated with reduced hazardous prescribing with a cost per QALY of £7449. It should be noted, however, that these were quasi-experimental studies using observational data and so the findings may have been due to unknown confounding. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research programme (NIHR award ref: RP-PG-1214-20012) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 15. See the NIHR Funding and Awards website for further award information.

Background Sleep problems are common in young people with epilepsy and neurodevelopmental disorders, impacting learning, behaviour and family function. Sleep disturbances can trigger seizures, which in turn disrupt sleep, creating a vicious cycle. Available behavioural interventions have been designed and positively evaluated for typically developing children or selected neurodevelopmental disorder populations, but these do not address epilepsy-specific concerns. Most relevant interventions are delivered face to face by psychologists, limiting accessibility and scalability in the National Health Service. A digital intervention based on behavioural change techniques, adapted for parents of children with epilepsy, is a potential solution. Objectives The overall aim was to broaden the perspective of the paediatric team managing young people with epilepsy beyond a narrow focus on seizures towards issues of importance to young people with epilepsy and their caregivers. To enable this, we (1) developed a core outcome set for childhood epilepsy research that was meaningful to children and their families and recommended candidate epilepsy-specific patient-reported outcome measures of children’s health-related quality of life and (2) produced an online behavioural intervention for sleep problems in young people with epilepsy, which we (3) evaluated using mixed-methods in a randomised controlled trial in the United Kingdom. Design A systematic review and Delphi survey to identify items of the core outcome measure set, with steps toward identifying suitable core outcome measures (work package 1) and a randomised controlled trial with internal pilot (work package 4) of an online behavioural sleep intervention cocreated with parents and young people (work package 3). The trial was evaluated using both quantitative and qualitative methods (work package 2) and with a cost-effectiveness analysis (work package 4). Setting Online and telephone survey of healthcare professionals. Online and in-person meetings of Delphi panel. Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention was created in a university psychology lab. The clinical trial took place in 26 National Health Service paediatric outpatient clinics across the United Kingdom. Participants Eighty-eight healthcare professionals completed a practice survey. One hundred and two stakeholders contributed to the core outcome set. Eighty-five children aged 4–12 years with clinician confirmed epilepsy and parent-reported sleep problems took part in the clinical trial. Interventions Parental access to the Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention online behavioural intervention plus standard care. The intervention arm was compared against standard care alone. Main outcome measures The primary end point was total score on the Child Sleep Habits Questionnaire, 19-item version and incremental cost per quality-adjusted life-year gained. Results Parent and clinician interviews before we designed the program highlighted a mismatch in patient expectations and healthcare providers’ capacity to meet specific health-related needs, including sleep problems. We developed a core outcome set for childhood epilepsy research, with multiple stakeholder input, comprising 10 domains and identified 2 leading candidate epilepsy-specific, patient-reported outcome measures of children’s health-related quality of life (work package 1). We also created an epilepsy-specific, online, parent-based behavioural intervention for sleep problems (Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention) (work package 3). The internal pilot feasibility study demonstrated that parents’ treatment preferences would render a drug arm unfeasible and therefore a factorial drug/behaviour design was simplified to behaviour intervention versus standard care work package 4. Eighty-five children were randomised (42 standard care, 43 standard care + Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention). The mean Child Sleep Habits Questionnaire score at 3 months did not differ significantly between standard care and standard care + Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention groups [mean difference = 3.00 (1.46), 95% confidence interval 0.06 to 5.93, p = 0.05)]. There were no adverse effects in either group. We note that 20/43 families given access to Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention did not engage with the core material. Reasons for this were explored in the qualitative assessments (work package 2) conducted at 3 and 6 months after randomisation. Twenty-two families participated in interviews. While Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention content was described as ‘high quality’, one father reported being overwhelmed by the amount of information and noted that some of it conflicted with other sources. Engagement was further affected by time constraints and distractions. Four Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention users also reported problems with parental sleep disruption and tiredness, including falling asleep, staying asleep, waking up or restless sleep. The economic analysis (work package 4) yielded an incremental cost-effectiveness ratio of £433,167 per quality-adjusted life-year gained and a probability of 0.01 for Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention being cost-effective at a threshold of £20,000 per quality-adjusted life-year. The results of the full trial showed that the Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention behavioural intervention plus standard care was not superior to standard care alone in improving parent report of child sleep problems in young people with epilepsy. However, there was a significant decrease of 16.5 minutes in child sleep onset latency between baseline and follow-up in the Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention arm compared to standard care. Parental knowledge of child sleep also significantly increased in the intervention arm. Limitations A significant proportion of participants randomised to the intervention did not engage with the online material. The health economics and secondary outcomes analyses were limited by large amounts of missing data in the final follow-up. Conclusions These results suggest that an exclusive digital approach to solve sleep problems for children with epilepsy, despite objective efficacy, has limited effectiveness and cost-effectiveness as a public health strategy in the National Health Service. Future work Future studies should evaluate a blended intervention of behaviour change techniques coached by paediatric epilepsy specialist nurses. Training in sleep management could be incorporated into secondary care mental health assessment and treatment. Trial registration This trial is registered as ISRCTN13202325. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0615-20007) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 16. See the NIHR Funding and Awards website for further award information. Plain language summary Sleep problems are very common in children with epilepsy, significantly impacting their behaviour, learning, seizure activity as well as family functioning. We created a framework to enable researchers to capture these important effects for future studies in childhood epilepsy, called as a core outcome set. Current approaches to managing sleep problems rely on highly trained psychologists who are in short supply across the National Health Service. To address this gap, we produced an e-learning resource (Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention), which has sleep strategies for parents to use with their child with epilepsy. These strategies address a wide range of difficulties, including falling asleep and staying asleep. This study aimed to find out if Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention improved parents’ view of sleep problems compared to families not using Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention, and whether Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention was value for money. A patient and public involvement group advised the research team throughout the study. We recruited children with epilepsy aged 4–12 years who had sleep problems, and their parents, from 26 paediatric epilepsy clinics in the United Kingdom. Families were randomly assigned to either access Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention in addition to standard care or receive only standard care. Parents reported on their children’s sleep problems at the start of the study and again after 3 months using the Children’s Sleep Habits Questionnaire. Eighty-five families participated: 42 had standard care, 43 parents were given access to Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention, although of these only 23 parents looked at the Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention sections on how to change their children’s sleep. We did not find a big difference between the sleep problem scores of children whose parents used Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention and those who had not. While Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention did not demonstrate value for money, it also did not have any negative effects from using it. Although the Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention did not significantly improve the reported sleep problems, children did fall asleep 15 minutes earlier. Project oversight and monitoring were undertaken by an independent Trial Steering Committee. Scientific summary Background Sleep problems are very common among young people with epilepsy and neurodevelopmental disorders and exert a major impact on learning, behaviour and family function among other things. Additionally, sleep disturbance can trigger seizures, which in turn disrupt sleep, initiating a vicious cycle. Available behavioural interventions for sleep problems have been designed and positively evaluated for typically developing or selected neurodevelopmental disorder populations, but none are available that account for the particular concerns that young people with epilepsy and their families share. Most relevant interventions are administered face to face by trained psychologists, but this route is neither fully accessible nor scalable across the NHS. A digital intervention based on behavioural change techniques, adapted for parents of children with epilepsy (CWE), is a potential solution. Objectives The overall aim of the programme was to broaden the perspective of the paediatric team managing young people with epilepsy beyond a narrow focus on seizures towards issues of importance to young people with epilepsy and their caregivers. To enable this, we (1) identified a core set of outcome measures reflecting health and quality of life that were meaningful to children and their families and (2) produced an online behavioural intervention for sleep problems in young people with epilepsy, which we (3) evaluated using mixed-methods in a randomised controlled trial (RCT) in the UK (see Report Supplementary Material 2 for the trial protocol). Methods Design A systematic review and Delphi survey to develop a core outcome set (COS) for childhood epilepsy research and identification and appraisal of epilepsy-specific patient-reported outcome measures of children’s health-related quality of life (work package 1) and a RCT with internal pilot (work package 4) of an online behavioural sleep intervention coproduced with parents and young people (work package 3). The trial was evaluated using both quantitative and qualitative methods (work package 2) and with a cost-effectiveness analysis (work package 4). Setting Online and telephone survey of healthcare professionals was performed besides online and in-person meetings of Delphi panel. Changing Agendas on Sleep, Treatment and Learning in Epilepsy Online Sleep Intervention (COSI) was created in a university psychology lab. The clinical trial took place in 26 NHS paediatric outpatient clinics across the UK. Sample size The trial aimed to recruit 110 participants, sufficient to detect a clinically meaningful difference of 6.25, with a standard deviation (SD) of 8.9 in 3-month Child Sleep Habits Questionnaire (CSHQ) scores with 90% power at a 5% significance level, accounting for potential attrition of 10%. However, owing to slow recruitment, 85 participants were randomised, which provided 80% power to detect change in the primary end point. Participants Eighty-eight healthcare professionals completed a practice survey. One hundred and two stakeholders contributed to the COS. Eighty-five children aged 4–12 years with clinician-confirmed epilepsy and parent-reported sleep problems were included. For the clinical trial, 85 children aged 4–12 years, with clinician-confirmed epilepsy and parent-reported sleep problems plus their parents, were included. Inclusion criteria: patients eligible for the trial must comply with all of the following at randomisation: children with clinician-confirmed diagnosis of epilepsy aged ≥ 4 years and < 13 years at the time of randomisation parent-/carer-reported child sleep problem as defined by mild, moderate or severe score over past 2 weeks on Hiscock Australian global sleep question (Hiscock H, Canterford L, Ukoumunne OC, Wake M. Adverse associations of sleep problems in Australian preschoolers: national population study. Pediatrics 2007;119:86–93) documented informed consent to be received from a person with parental responsibility family to have an e-mail address and mobile phone parent and child are to have a good enough understanding of the English language to read and answer study questionnaires. Exclusion criterion: children with moderate/severe learning disabilities. Interventions Access for primary caregivers to the COSI online behavioural intervention plus standard care (SC). The intervention arm was compared against SC alone. Primary outcome measures To determine if COSI is superior to SC with respect to the total score on the CSHQ, 19-item version. Primary economic outcome To determine whether SC + COSI is cost-effective compared with SC alone from the perspective of the NHS and Personal Social Services. Secondary outcomes To determine if COSI is superior to SC with respect to the CSHQ score at 6 months. To compare COSI to SC with respect to time to first seizure. To compare COSI to SC with respect to time to 6-month seizure remission. To determine if there is a change in parental sleep-related knowledge with respect to the Score of Knowledge About Sleep in Childhood (KASC) scale at 3 months. To determine if COSI is superior to SC with respect to parental anxiety assessed by the Hospital Anxiety and Depression Scale score at 3 and 6 months. To determine if COSI is superior to SC with respect to parental sleep problems assessed by the Insomnia Severity Index score at 3 and 6 months. (Buysse DJ, Reynolds 3rd CF, Monk TH, Berman SR, Kupfer DJ. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res 1989;28:193–213). To compare measures of child’s sleep related reaction time and executive function between COSI and SC measured via SleepSuite assessment (iPad game) at 3 months. To compare health-related quality of life between COSI and SC assessed by the Health-Related Quality of Life in Children with Epilepsy score change in children and WHO-5 Well-Being Index score change in primary carers at 6 months. To compare measures of children’s behaviour between COSI and SC assessed by the total score on Strengths and Difficulties Questionnaire at 3 and 6 months. To compare parenting self-efficacy between COSI and SC via score changes in Parenting Self Agency Measure at 3 and 6 months (Dumka LE, Stoerzinger HD, Jackson KM, Roosa MW. Examination of the cross-cultural and cross-language equivalence of the parenting self-agency measure. Fam Relat 1996;45:216–22). To compare sickness-related school absences between COSI and SC. To determine if objective sleep of parents and children, measured by actigraphy (sleep onset latency, sleep efficiency and total sleep time), changes at 3 months in the COSI and SC group. Statistical analysis Outcome data were analysed using intention-to-treat principles. As the 19-item CSHQ tool used in the trial had not undergone prior validation, the overall Cronbach’s alpha was calculated to assess its internal consistency. For the primary analysis, individual items (specifically items 16, 19 and 20) were excluded from the total score, because their deletion increased Cronbach’s alpha and they showed a very low ‘corrected item–total correlation’ (< 0.20). Sensitivity analyses were conducted to examine the effect of alternative scoring approaches and to make comparisons with the validated 33-item CSHQ questionnaire. The item-level mean, SD and n were tabulated for each treatment group and overall, at each time point (TP), to allow for comparison with other studies (see Appendix 1, Table 4). Continuous outcomes, including CSHQ score, were compared at relevant TPs (3 or 6 months) using linear mixed-effects regression models, with fixed effects included for intervention, baseline measure, baseline sleep medication status (Y/N) and random effects for centre. Time-to-event outcomes were analysed using a Cox proportional hazards model, including intervention, baseline sleep medication status (Y/N) and centre effects (see Appendix 1, Table 6). The adjusted treatment effect estimates (mean difference for continuous outcomes and hazard ratio for time to event outcomes) comparing SC + COSI against SC are presented with a 95% confidence interval (CI), with statistical tests conducted at the 5% two-sided significance level. Unadjusted analyses were examined for completeness. Analyses were based on complete cases, which were unbiased when data were missing completely at random (Gelman A, Hill J. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press; 2007). Analyses included patients with follow-up data captured within ± 1 month of the TP of interest. Due to the revised study design, a subset of participants did not achieve a full 6 months of follow-up, and sensitivity analyses were conducted that included participants with data collected at ≥ 4 months post randomisation. Normality assumption was examined by histograms or Q–Q plots of the residuals. The assumption of homoscedasticity was examined by plotting residuals against fitted values or predictor variables. Residuals should be independent of each other in a mixed-effects model, and this also includes the independence of random effects check. Assumptions regarding the distribution of random effects were considered. Random effects were assumed to follow a normal distribution with mean zero and constant variance. All analyses were reported in accordance with the Consolidated Standards of Reporting Trials checklist (Schulz KF, Altman DG, Moher D, Group C. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. BMJ 2010;340:c332) and regardless of statistical significance. Cost-effectiveness analysis The within-trial analysis used individual, patient-level data to compare the costs incurred and utility experienced by participants within each of the treatment groups over the first 6 months post randomisation. The primary outcome was the incremental cost-effectiveness ratio (ICER), expressed as costs per quality-adjusted life-year (QALY) gained. Participants’ use of primary and community care services, as well as their prescribed medicines, were measured using a resource use questionnaire. Hospital inpatient, outpatient and emergency care use was based on routine Patient Level Information and Costing System data. Resource use was valued in monetary terms (GBP) based on 2021–2 prices. Unit costs for NHS and Personal Social Services were taken from standard references. QALYs were generated from utility data measured using the Child Health Utility 9 Dimension Index. The economic analysis involved multiple imputation to account for missing data and multivariate Bayesian regression models to adjust for any baseline differences and to increase the precision of estimated costs and QALYs. The ICER was calculated as the difference between mean total costs of SC + COSI and SC alone divided by the difference in mean QALYs. Uncertainty in the costs and QALYs were evaluated using draws from the joint posterior distribution of the estimates from the regression analyses. A range of sensitivity and scenario analyses was undertaken to account for uncertainties in the data besides modelling assumptions. Results Parent and clinician interviews and surveys highlighted a mismatch in patient expectations and healthcare providers’ capacity to meet specific health-related needs, including sleep problems. The internal pilot feasibility study demonstrated that parents’ treatment preferences would render a drug arm unfeasible and therefore a factorial drug/behaviour design was simplified to behaviour intervention versus SC. Eighty-five families were randomised (42 SC, 43 SC + COSI). The mean CSHQ score at 3 months did not differ significantly between SC and SC + COSI groups [mean difference = 3.00 (1.46), 95% CI 0.06 to 5.93, p = 0.05]. The ICER was £433,167 per QALY gained, based on costs of £2021 (95% credibility interval £1352 to £3455) and £790 (£409, £1628) for SC + COSI and SC, respectively, and QALYs of 0.42 (0.39 to 0.45) and 0. 41 (0.38 to 0.45). The probabilities of COSI being cost-effective at thresholds of £20,000 and £30,000 per QALY were 0.01 and 0.04, respectively. No significant differences were observed in seizure control or other secondary outcomes. However, children in the SC + COSI group fell asleep on average − 16.35 minutes faster (95% CI −30.63 to −2.07, p = 0.03) and parental sleep knowledge on the KASC questionnaire improved significantly in the SC + COSI group, 2.03 (0.48) (95% CI 1.01 to 3.05, p ≤ 0.001). There were no adverse effects in either group. It is noteworthy that 20/43 families given access to COSI did not access the core behaviour change material. Reasons for this were explored in the qualitative assessments (work package 2) conducted at 3 and 6 months after randomisation. Twenty-two families participated in interviews. While COSI content was described as ‘high quality’, one father reported being overwhelmed by the amount of information and noted that some of it conflicted with other sources. Engagement was further affected by time constraints and distractions. Four COSI users also reported problems with parental sleep disruption and tiredness, including falling asleep, staying asleep, waking up or restless sleep. The results of the full trial showed that the COSI behavioural intervention plus SC was not superior to SC alone in improving parent report of child sleep problems in young people with epilepsy. The economic analysis did not support the case for the cost-effectiveness of COSI in the primary economic analysis or any of the sensitivity analyses conducted (see Appendix 3 and Report Supplementary Material 1). Limitations A significant proportion of participants randomised to the intervention did not engage with the online material. The health economics and secondary outcomes analyses were limited by large amounts of missing data in the final follow-up. Conclusions These results suggest that an exclusive digital approach to solve sleep problems for CWE, despite objective efficacy, has limited effectiveness as a public health strategy in the NHS. Future work Future studies should evaluate a blended intervention of behaviour change techniques coached by a trusted healthcare professional, such as a paediatric epilepsy specialist nurse. Training in sleep management could be incorporated into secondary care mental health assessment and treatment. Trial registration This trial is registered as ISRCTN13202325. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0615-20007) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 16. See the NIHR Funding and Awards website for further award information.

Background There is increasing evidence that sedentary behaviour has a detrimental effect on health and well-being. At least 1.2 million people living in England have had a stroke, and longer-term outcomes are poor for many. Stroke survivors are particularly sedentary compared to age-matched controls. Objective(s) We sought to enhance health outcomes for stroke survivors through the development and evaluation of strategies to reduce sedentary behaviour. Research questions related to identification of existing evidence; exploration of capabilities, opportunities and motivation related to sedentary behaviour after stroke; determining whether an intervention developed using principles of coproduction was feasible and whether trial design issues could be addressed before assessing the clinical and cost-effectiveness of the intervention. A pragmatic definitive trial to assess the developed intervention was planned; due to the COVID-19 pandemic, the trial was reduced to an external pilot trial. Design and methods Five overlapping workstreams were undertaken: Systematic review to collate the quantitative and qualitative evidence relating to sedentary behaviour in adults including survivors of stroke. Observations and semistructured interviews in two stroke services to inform the coproduction process and explore the capability, opportunity and motivation to address sedentary behaviours. Development of an intervention to reduce sedentary behaviour through use of coproduction principles in Yorkshire and Scotland. A single-arm feasibility trial in three services to refine the developed intervention and clarify trial procedures. A multicentre external pilot cluster randomised controlled trial evaluating the intervention incorporating embedded process and economic evaluations. Setting and participants Stroke services (inclusive of primary, secondary and community care provision) across England and Scotland. Participants were stroke survivors and their carers, and healthcare professionals in the included stroke services. Intervention An intervention to reduce sedentary behaviour was coproduced in workstream 3, informed by information obtained in workstream 1 and workstream 2. In workstream 4, groups (healthcare professionals, stroke survivors and researchers) were established in three stroke services that led implementation in their service and contributed to the iterative refinement of the intervention. The intervention (called Get Set Go) is a whole-service intervention designed to be implemented and embedded within routine practice. Delivery commences in the inpatient stroke unit setting and continues into the community for at least 12 weeks post discharge. The intervention focuses on: (1) educating staff and stroke survivors; (2) preparing and enabling staff to support and encourage stroke survivors to stand and move more in everyday stroke care; (3) encouraging stroke survivors to monitor their own standing and moving, Main outcome measures The primary outcome was stroke survivor self-reported Nottingham Extended Activities of Daily Living scale at 12 months post registration. The key secondary outcome was mean daily sedentary time (minutes) at 12 months post registration using activity monitor (activPAL) data. Data sources Literature reviews; observations and semistructured interviews in stroke services; feasibility and trial outcomes. Results Workstream 1: two quantitative (n = 85 studies and n = 30, respectively) and two qualitative systematic reviews (n = 30, n = 17) were undertaken and published. Workstream 2: observational (in inpatient and community settings) and interview work was undertaken in two sites (Yorkshire and Scotland). Workstream 3: five coproduction workshops were conducted concurrently in Edinburgh and Yorkshire (n = 43 participants), and a provisional intervention was developed. Workstream 4: implementation groups in three services, who operationalised and implemented the intervention. Trial procedures were refined and a concurrent qualitative data collection guided adaptation of the intervention. Workstream 5: 15 English stroke services were randomised (n = 8 treatment, n = 7 control) and recruited 334 stroke survivors (n = 249 retained at 12-month follow-up). Results did not indicate potential effectiveness of Get Set Go compared with usual care [adjusted mean difference in Nottingham Extended Activities of Daily Living scale at 12 months (95% confidence interval) p = −1.76 (−7.63 to 4.25), p = 0.530]. No significant differences between arms were found in secondary outcomes. The exploratory findings from this external pilot study suggest that the Get Set Go intervention is unlikely to be cost-effective. Limitations Due to coronavirus disease discovered in 2019, workstream 5 was reduced to become an external pilot trial with reduced 24-month follow-up. This trial was therefore not powered to detect definitive outcomes. Conclusions Participants appreciated the relevance of the work and components of the intervention developed. Implementing service change in the inpatient setting is challenging. Future work Work is ongoing to tailor the intervention to different patient types based on a secondary analysis of the patient interview data using ideal-type analysis methods. Development of a consistent messaging strategy relating to sedentary behaviour and survivors of stroke is required. Trial registration This trial is registered as ISRCTN 12246326, registered on 7 August 2019 and ISRCTN82280581, registered on 1 April 2020. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0615-20019) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 14. See the NIHR Funding and Awards website for further award information. Plain language summary Too much time spent in sedentary behaviours (sitting/lying down) is linked with ill health and mortality. Stroke survivors are particularly sedentary, spending up to 80% of their waking day sitting or lying down. We developed and evaluated an intervention to reduce sedentary behaviour for stroke survivors through five workstreams. We updated reviews of previous and current research in workstream 1. In workstream 2, we interviewed staff and observed what happens on stroke units and during therapy sessions in patients’ homes. We also interviewed stroke survivors and their carers, to explore their thoughts on sedentary behaviour, and what might help or hinder this. In workstream 3, through a series of meetings, using information from workstream 1 and workstream 2, ideas and action plans (an intervention) for reducing time spent sitting/lying were produced. The intervention (called Get Set Go) was refined by working with three stroke services in workstream 4. The intervention included training and materials for staff and stroke survivors to encourage stroke survivors to undertake and monitor their own standing and moving in the stroke unit and after discharge. We undertook a preliminary exploration of the developed intervention in 15 stroke services; 8 were randomly allocated to deliver Get Set Go and the other 7 continued providing their usual care to all stroke survivors. We aimed to explore whether Get Set Go was likely to improve stroke survivors’ ability to complete extended activities of daily living in the first year after stroke. A total of 334 stroke survivors were included in the trial. In this exploratory work, no significant differences were found between groups in any of the measured outcomes and Get Set Go seems unlikely to be cost-effective, but our work was limited by the pandemic and did not include sufficient participants to detect a difference. Participants and staff appreciated the relevance of the work and components of the intervention. Scientific summary Background There is increasing evidence that sedentary behaviour defined as any waking behaviour characterised by low energy expenditure ≤ 1.5 metabolic equivalent of tasks while in a sitting or reclining posture has a detrimental effect on health and well-being. Stroke survivors are particularly sedentary. The aim of this programme was to enhance health and disability outcomes for stroke survivors through the development and evaluation of strategies to both reduce overall time spent sedentary and break up long bouts of sedentary behaviour. The objectives were to: Ensure that intervention development was informed by up-to-date review of the quantitative and qualitative evidence. Inform development of an intervention by observations and qualitative investigation of sedentary behaviour with stroke survivors, their carers and health professionals, including exploring their capability, opportunities and motivation to address this behaviour. Use coproduction principles to develop intervention strategies to reduce sedentary behaviour in people after stroke. Test the implementation of the intervention and clarify trial design components in a feasibility study in three stroke services. Conduct a large pragmatic cluster randomised trial evaluation of the developed intervention with internal pilot, process evaluation and cost-effectiveness analysis. Due to the unprecedented effects of the coronavirus disease discovered in 2019 (COVID-19) pandemic on the NHS, this trial was reduced in size and scope to become an external pilot trial with time-limited follow-up of participants. Methods Stroke survivors and their carers participated in the development of this research programme and contributed throughout. We engaged with local stroke groups and co-applicant Gill Carter assisted in all workstreams (WSs). In addition, our methods were purposively inclusive of stroke survivors’ input through qualitative reviews (WS1), interviews (WS2) and directly in our coproduction work (WS3), intervention refinement (WS4) and review of the delivered intervention (WS5). Workstream 1 Four systematic reviews of the quantitative and qualitative evidence relating to sedentary behaviour were undertaken to ensure our work was founded on strong evidence which included identifying relevant behaviour change techniques and barriers and facilitators to intervention implementation. Workstream 2 Observations (in hospital and in the community) and semistructured interviews in two stroke services were undertaken. Interviews with stroke survivors and their carers, at 6 and 9 months post stroke, and with members of the multidisciplinary team explored the capability, opportunity and motivation to address sedentary behaviours from their perspectives. Thematic analysis was undertaken for all observational data; interviews were analysed using Framework Analysis. Workstream 3 Informed by information gathered in the previous two WSs and behaviour change approaches, coproduction principles were utilised to develop a tailored intervention to reduce sedentary behaviour after stroke. Workstream 4 A single-arm feasibility trial was undertaken in three stroke services to: clarify content and methods of delivery of the intervention; assess fidelity; and capture learning from, and acceptability to, staff and participating stroke survivors. Iterative refinement of the intervention was undertaken by service-led implementation groups and captured through Table of Change methodology. Trial procedures relating to patient eligibility criteria and recruitment processes were clarified. Outcome measures were administered including a provisional Client Service Receipt Inventory to capture resource use data for the health economic evaluation. Workstream 5 A multicentre pilot cluster randomised controlled trial with embedded process and economic evaluations was undertaken in 15 stroke services to provide a preliminary exploration of the effectiveness of Get Set Go (GSG) in improving the ability to complete extended activities of daily living (ADL) and reduce sedentary behaviour in the first year after stroke. The process evaluation, including observations of training and practice; interviews with staff, patients and carers; and documentary analysis, was undertaken to explore implementation of the intervention, how it was experienced and understood by recipients and providers and identify any potential moderators and mediators of the intervention effect. The health economic analysis explored core resource use and costs associated with delivering the intervention including impacts on wider care. Results Workstream 1 Two quantitative (n = 85 studies and n = 30, respectively) and qualitative systematic reviews (n = 30, n = 17) were undertaken and published. An evidence gap was confirmed. The findings indicated that interventions might be effective in reducing time spent sedentary, but intervention strategies to encourage sustainability of effects are needed. Interventions targeting a reduction in sedentary behaviour need to be flexible with context awareness. Workstream 2 Over 132 hours of observations, semistructured interviews with stroke survivors (n = 31), carers (n = 12) and staff (n = 30) across the two services were undertaken. The physical and social environment, perceptions of stroke survivors’ physical and psychological capability to move, and routinised practices enacted by staff facilitate an expectation to be sedentary in the inpatient setting. This is often carried over when stroke survivors leave hospital. Staff, stroke survivors and carers reported they recognised the value of reducing sedentary behaviour and wanted to learn about safe and appropriate methods for doing so. Workstream 3 Five face-to-face coproduction workshops comprising key stakeholders (stroke survivors, carers, inpatient and community healthcare professionals, and public health practitioners n = 43 overall) were convened and undertaken concurrently in Yorkshire and Scotland. Informed by findings from earlier WSs, a prototype tailored intervention with delivery strategies was developed, called Get, Set, Go, standing and moving more after stroke. A logic model and Template for Intervention Description and Replication checklist were also completed. Workstream 4 Three stroke services participated in this feasibility work; the intervention was delivered as a service-wide initiative, supported by purposely convened implementation groups. Thirty-nine stroke survivors were recruited to complete outcome measures; follow-up was low due to the start of the COVID-19 pandemic. There were some missing data in outcome assessment and some components of the health economic data collection tool were poorly completed, resulting in refinement of these measures prior to WS5. The activPAL (activity monitor) was shown to be feasible for use, although need for training refinement was highlighted. There were no safety concerns. Qualitative observations and interviews (n = 28 staff; n = 12 patients) informed changes to the fidelity and adherence data collection methods. Refinement of the intervention was undertaken on review of the Table of Change. Intervention The intervention (GSG) comprised four main components: Education aimed at staff, stroke survivors and carers. Staff guidance (assessment tools) and materials (posters, prompt magnets) to support stroke survivors integrating standing and moving more into daily routines. A guide (with monitoring forms) to support stroke survivors to stand and move more after stroke, including information for their family, friends and informal carers. Environmental adaptation. A key component of the intervention was therapy assessment of ability to stand by way of a traffic light system (red, amber, green) with a linked ‘prescription’ for standing and moving provided, to inform and encourage survivors of stroke to stand and move more. Stroke survivors are reassessed regularly and the prescription amended as required with detailed information and self-complete monitoring forms in the accompanying guide. Workstream 5 A total of 15 clusters were randomised; 8 were randomised to deliver GSG plus usual care (UC) and 7 were randomised to deliver UC. Intervention training took place over a 24-month period primarily online due to COVID-19 restrictions supported by in-person visits when allowed. In total, 103 training sessions were delivered to 381 staff. A total of 5984 patients were screened: 1029 (17.2% of screened) were approached, 969 (94.2% of approached, 16.2% of screened) were eligible and 334 (34.6% of those eligible; 5.6% of those screened) were registered into the trial; 181 participants were registered to the GSG arm and 153 to the UC arm. Participants across the treatment arms were similar in terms of age, gender and living arrangements. Stroke survivors had an average age of 69 years (range 23–98) and 60.4% were female. However, there were some differences between arms in ethnicity of participants; 94.4% were White British in the GSG arm compared to 85.4% in the UC arm. Ethnicity differences could have been due to regional variations between the recruiting sites. Participants in each arm had similar levels of disability, stroke severity and cognitive impairment. Of the 331 participants, 96 (29.0%) participants withdrew from at least one element of the study; 56 (31.1%) in the GSG arm and 40 (26.5%) in the UC arm. Baseline characteristics were very similar between participants who withdrew and did not withdraw. The mean Nottingham Extended Activities of Daily Living (NEADL) score pre-stroke was 57.4, with slightly greater independence in the UC arm (58.9 compared to 56.1). As a preliminary exploration of effectiveness, we report at 6 months after stroke, the mean NEADL score was approximately 45.3 with a median score of 50.0 and a wide interquartile range from 31 to 61 points. Mean NEADL score at 12 months was 43.3 (18.87) in the GSG arm versus 47.4 (17.11) in the control arm. The mean difference in the cluster-level adjusted residuals for NEADL score between the two arms at 12 months was −1.76 [95% confidence interval (CI) −7.63 to 4.25, p-value 0.530]. Independence increased in both arms for the subset of participants followed up at 24 month. At both 6 and 12 months, the UC arm had slightly greater independence. The mean difference in the cluster-level adjusted residuals for mean daily sedentary time in minutes between the two arms at 12 months was −29.6 (95% CI −33.71 to 92.81, p-value 0.322). The difference was not statistically significant for mean daily sedentary time. No significant differences between arms were found in secondary outcomes. This pilot trial was not powered to detect a difference. Process evaluation Observations (11 training, 44 intervention, 17 control) and interviews [n = 52 staff (40 intervention, 12 control); and n = 27 patients (17 intervention, 10 control)] were undertaken across four time points each separated by approximately 3 months, alongside documentary analysis. Key findings Many patients and staff valued learning and emphasised that the core message of movement after stroke is important. Intervention implementation and awareness occurred inconsistently among patients and staff. Inpatients were largely inactive outside of therapy sessions, many felt their movement was restricted, and some viewed GSG as giving permission to move. While the intervention provided initial patient motivation, it was not suited for all and its use decreased over time as patient priorities shifted or they felt they no longer needed it. Staff supported implementation using additional local setup and ‘Champion’ roles, but felt the message was sometimes lost in trial paperwork. Movement occurs in the context of safety and comorbidity, within a complex system and pressured environment. Exploratory economic evaluation Resource use patterns at baseline were similar between groups for formal services but the GSG group used more unpaid care (24 vs. 4 hours on average); this greater use of unpaid care was also evident at follow-ups (122 vs. 78 hours on average at 12 months), coupled with higher use of hospital services. Consequently, total costs were higher in the GSG group across baseline (£2710 vs. £1452) and all follow-ups (£11,140 vs. £8989 at 12 months). The GSG group showed higher variability which means there was wider cost dispersion and implies a highly skewed cost distribution with a standard deviation twice as the mean at 12 months follow-up. Utility estimates also suggested lower quality-adjusted life-years (QALYs) for the GSG group over the follow-up period. In combination, from both NHS/Personal Social Services and societal perspectives, costs were higher in the GSG group, while outcomes measured in QALYs were worse, suggesting in this exploratory work that the GSG intervention is unlikely to be cost-effective. This is further evidenced by the bootstrapped replications, which show that almost all replications lie in the northwest quadrant of the cost-effectiveness plane. Conclusions Awareness of the detrimental effects of sedentary behaviour has increased over the lifetime of this grant. Our research question remains valid, as no effective interventions have yet emerged for this client group. Our extensive literature reviews and multisite qualitative work informed our programme and will assist others in this area; similarly, the detailed operationalisation of the process of coproduction will assist others using this methodology. We successfully developed a robust meaningful intervention (GSG), using coproduction involving a range of stakeholders, which was compatible with routine practice. Unfortunately, our work was considerably impacted by COVID-19, and we were unable to complete the planned definitive trial. The exploratory findings from this pilot study must be viewed with caution but suggest that the GSG intervention is unlikely to be cost-effective. Important learning has emerged however; stroke survivors and staff appreciated and understood the relevance of the work and components of the intervention developed. Survivors of stroke and healthcare professionals did not always understand the concepts of sedentary behaviour and the importance of breaking up sedentariness throughout the day. This is an important public health message for everyone and requires clear dissemination. Despite best efforts, take-up was patchy with other members of the multidisciplinary team perceiving encouraging movement to be the role of the therapy team rather than a service initiative. Efforts to reduce sedentary behaviour may conflict with other initiatives aimed to enhance patient safety such as falls risk policies. There is a challenge when to start such an intervention, either in busy inpatient settings to ensure behaviour change messages are provided early or later when the situation is less frantic, but habits may already be formed. The overlay of paperwork to capture implementation and costs added complexity (and some confusion) and perhaps reduced implementation. Activity monitors proved acceptable and this work provides one of the largest ever data sets in this population. Patient-completed outcome measures are burdensome for participants, clinical and research staff. Within the economic evaluation, a comparison between self-reported resource use data and routine secondary care records highlights a need for careful strategies for data collection. It could be helpful to evaluate interventions using hybrid data collection approaches that make use of routine data and Sentinel Stroke National Audit Programme. Recommendations for research While this external pilot trial was successfully concluded, take-up of the intervention was patchy and participants’ dropout greater than we would wish. Refinement of the intervention implementation and development of a more efficient trial design are required prior to a definitive trial. The GSG intervention remains a potentially useful and important tool. Further work is being conducted based on a secondary analysis of the patient interview data using ideal-type analysis methods, to produce recommendations for how GSG can be tailored to different types of survivors of stroke. Future analysis could explore trajectories and clustering to determine whether GSG works for some patient groups and not for others. The timing of delivery of this and other self-management interventions remains problematic. Further work should explore clinical staff’s views of assessing stroke survivors’ ability to self-manage aspects of their life after stroke. Trial registration This trial is registered as ISRCTN 12246326, registered on 7 August 2019 and ISRCTN82280581, registered on 1 April 2020. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0615-20019) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 14. See the NIHR Funding and Awards website for further award information.

Background Cognitive–behavioural therapy is an effective treatment for depression. A key question is how to increase access. Engagement with cognitive–behavioural therapy-based computerised interventions is poor, and programmes are inflexible and impersonal. Innovative use of technology and integration of online materials could increase engagement and widen access. Objectives To develop and evaluate a novel approach to delivering cognitive–behavioural therapy for depression integrating therapist-led sessions and online cognitive–behavioural therapy materials. Design and methods The INTEgrated theRApist and online CbT for depression research programme comprised four work packages. The first developed the online therapy platform and materials, and training for therapists. This comprised a series of studies: a systematic review and network meta-analysis to compare the effectiveness of different types and components of cognitive–behavioural therapy; a Delphi study focused on the effective components of cognitive–behavioural therapy; a decision model to evaluate the cost-effectiveness of different formats of delivering cognitive–behavioural therapy; a survey of accredited cognitive–behavioural therapy therapists asking about their use and views of different resources used in cognitive–behavioural therapy; and iterative design work aimed at understanding the design requirements for the platform. The prototype platform was then evaluated in a pilot study, with subsequent final refinement. The second and third work packages evaluated the clinical and cost-effectiveness of the intervention compared with usual general practitioner care in a multicentre randomised controlled trial (with a parallel economic evaluation) over 12 months in primary care patients with depression. The fourth examined the intervention’s acceptability through a nested qualitative study of patients, therapists and supervisors. Setting The randomised controlled trial was based in United Kingdom primary care in Bristol, London and York. Participants Patients aged ≥ 18 years experiencing depressive symptoms in primary care were eligible for the randomised controlled trial. Interventions In the randomised controlled trial, participants were individually randomised to: (1) integrated cognitive–behavioural therapy (in addition to usual general practitioner care); or (2) to continue with usual general practitioner care. Main outcome measures The primary outcome for the randomised controlled trial was depressive symptoms measured using the Beck Depression Inventory, version 2 at 6 months post randomisation. Secondary outcomes included response and remission (based on Beck Depression Inventory, version 2 score), depressive symptoms (Patient Health Questionnaire-9), anxiety symptoms (Generalised Anxiety Disorder-7), function (Work and Social Adjustment Scale), quality of life (EuroQol-5 Dimensions, five-level version), and costs of interventions and wider services. Results Work package 1: the network meta-analysis found no evidence of effect for any content components or combinations of components. There was uncertainty around estimates of cost-effectiveness for different treatment modalities and intensities. Effective components of cognitive–behavioural therapy were identified through the Delphi study, and resources used by therapists identified through a survey of United Kingdom cognitive–behavioural therapy practitioners. Key requirements of an online platform identified through iterative design work were: (1) overcoming depression-related barriers; (2) supporting engagement; (3) reinforcing learning and skill acquisition. In a pilot study with 18 primary care patients, patients said that the integrated approach made therapy more accessible. Therapists commented on the flexibility of the approach. Not all participants engaged with between-session tasks and some technical issues were experienced. Platform refinements were made prior to the randomised controlled trial. Work package 2: overall, 451 patients were recruited to the INTEgrated theRApist and online CbT for depression randomised controlled trial. Participants were predominantly female (n = 313, 69%) and, on average, aged 39 years. The mean Beck Depression Inventory, version 2 score at baseline was 32.8, indicative of severe depression. Most had a history of depression (nearly half having had five or more prior episodes of depression), were taking antidepressants (70%) and the duration of the current episode of depression was ≥ 2 years for 51% of participants. In the intervention group, 14 individuals (6.2%) had no therapy sessions and 137 participants (60.9%) completed therapy (received at least 9 sessions or reached an agreed end of therapy with their therapist in fewer than 9 sessions). Including the above 14 individuals, 88 (39.1%) either withdrew from therapy (n = 50) or were discharged for non-attendance (n = 38). In total 334 patients (171 integrated cognitive–behavioural therapy; 163 usual care) were included in the primary analysis. The intervention group had a Beck Depression Inventory, version 2 score that was, on average, 4.4 points lower (less depressed) than the usual care group at 6 months [difference in means: −4.4 (95% confidence interval −7.0 to −1.9); p = 0.001]. In repeated-measures analyses using data from 6 and 12 months, individuals in the intervention group had a Beck Depression Inventory, version 2 score that was, on average, 3.8 points lower than those in the usual care group (95% confidence interval −6.1 to −1.5; p = 0.001). The intervention group had a twofold increased odds of response and remission, fewer symptoms of depression (Patient Health Questionnaire-9) and anxiety (Generalised Anxiety Disorder-7), and improved functioning (Work and Social Adjustment Scale). Work package 3: the mean costs of integrated cognitive–behavioural therapy were £987 (standard error £14) per participant. The mean costs of usual care were estimated at £382 (standard error £57) in usual care group and £193 (standard error £42) in intervention group. In the primary analyses, costs from the National Health Service/Personal Social Services perspective were £753 (standard error £77) per participant in the usual care group and £1754 (standard error £127) in the intervention group, with adjusted incremental costs of £1009 (95% confidence interval £737 to £1286). The mean quality-adjusted life-years were 0.554 (standard error 0.017) in the usual care group and 0.597 (standard error 0.017) in the intervention group, with adjusted incremental quality-adjusted life-years at 0.033 (95% confidence interval −0.002 to 0.059). The incremental cost-effectiveness ratio was £30,576 per quality-adjusted life-year gain. Complete-case analysis from the National Health Service/Personal Social Services perspective showed a slightly more favourable picture with an incremental cost-effectiveness ratio at £22,421. Work package 4: through interviews with trial participants and therapists, we found that the integrated approach helped patients manage their depression. Platform benefits included the opportunity to review transcripts and to support homework tasks. Typing allowed reflection and a focused discussion. Less could be covered than during an in-person session. Patients who did not complete therapy struggled with typing and found cognitive–behavioural therapy too demanding. Limitations In the trial, the 6-month follow-up rate was slightly below the original target. Conclusions Integrated cognitive–behavioural therapy is an effective and acceptable treatment for patients with depression. There was uncertainty around the cost-effectiveness of the intervention. This novel mode of delivery could increase the availability of cognitive–behavioural therapy and access for those who find it difficult to attend appointments in person. Future work To examine whether effects are sustained long term and to understand which aspects of the platform lead to improvements. Study registration This study is registered as ISRCTN14850613 (phase 2 pilot study) and ISRCTN13112900 (RCT). Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0514-20012) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 13. See the NIHR Funding and Awards website for further award information. Plain language summary There is a high demand for talking therapies such as cognitive–behavioural therapy for depression. Cognitive–behavioural therapy can be delivered by computers or online as written guidance that can be worked through with or without support. These computerised therapy packages are inexpensive and convenient but are not as effective or as engaging as having a therapist. They do not allow treatment to be tailored for the individual. We built an online therapy platform. This combined live therapist sessions with online materials to help patients practise outside the sessions. In the first session, patients and therapists met by videocall. Thereafter, they communicated by typing during live online therapy sessions. We worked with stakeholders to develop integrated cognitive–behavioural therapy. Patients could receive between 9 and 12 sessions of therapy. We evaluated integrated cognitive–behavioural therapy in three ways: We recruited 451 patients with depression and randomly allocated them to either integrated cognitive–behavioural therapy or to continue with usual general practitioner care. Those allocated to therapy were less depressed and more likely to have recovered after 6 and 12 months. This means we can be confident that this is a clinically effective treatment. We assessed whether integrated cognitive–behavioural therapy was good value for money. Although patients felt better, the treatment just failed to meet criteria for value for money. However, the average cost of integrated cognitive–behavioural therapy was similar to the cost of therapy in National Health Service talking therapy services. In-depth interview study: we asked patients and therapists about the treatment. They found it acceptable. Patients who completed the therapy valued being able to talk to a therapist and had learnt skills to manage their depression. Reviewing the record of therapy sessions helped them complete homework tasks and track progress. While less could be covered in a session compared with in-person therapy, the slower pace allowed room for reflection. This also meant therapists used more focused questions. Some patients found it difficult to express themselves through typing. Scientific summary Some text in this section is reproduced from Tallon D, Thomas L, Brabyn S, Ching BCF, Hahn JS, Jude B, et al. Integrated therapist and online CBT for depression in primary care (INTERACT): study protocol for a multi-centre randomised controlled trial. Trials 2023;24:421. https://doi.org/10.1186/s13063-023-07396-9). This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text. Background Cognitive–behavioural therapy (CBT) is an effective treatment for depression and recommended by the National Institute for Health and Care Excellence (NICE). A key question for commissioners and healthcare providers is how to increase access. Cognitive–behavioural therapy-based computerised CBT interventions form part of the stepped care pathway for depression but are not an alternative to high-intensity CBT as they lack flexibility and are impersonal. Cognitive–behavioural therapy delivered online using instant messaging is clinically and cost-effective. Developing materials that are integrated with modern technologies yet permit the therapist to tailor treatment to the individual is critical. Ready access to such materials could facilitate engagement with tasks that take place outside therapy sessions and increase effectiveness. In 2020–1, 90% of UK households had a home computer and 84% of over 16-year-olds private use of a smartphone. Innovative use of technological developments and integration of online materials into therapy offers the potential to increase engagement and widen access to populations that are difficult to reach (e.g. those who are disabled or have difficulty attending appointments for other reasons). Our intervention integrates therapist-led sessions and online CBT materials in a novel approach to the treatment of depression. Aims and objectives The aim of the INTEgrated theRApist and online CbT for depression (INTERACT) programme was to develop [work package (WP) 1] and evaluate (WPs2–4) an integrated approach to delivering CBT for depression in primary care (integrated CBT). The specific aims of the WPs are listed below. Work package 1: intervention development To identify clinical and cost-effective components of CBT to inform the development of the intervention. To develop an online platform to support the delivery of integrated CBT. To develop the online CBT materials. To develop a training package for therapists. Work package 2: randomised controlled trial To examine the clinical effectiveness of an integrated approach to delivering CBT for depression over 12 months’ follow-up. Work package 3: economic evaluation To examine the cost-effectiveness of the integrated CBT intervention. Work package 4: qualitative evaluation To explore patients’, therapists’ and supervisors’ views and experiences of using an integrated approach to delivering CBT for depression. To understand patients’ reasons for completing or not completing integrated therapy. To assess patients’, therapists’ and supervisors’ views on how this novel approach affects the therapist-patient relationship. Methods and results Work package 1: intervention development Systematic review and network meta-analysis of cognitive–behavioural therapy components We conducted a systematic review of randomised controlled trials (RCTs) in adults with depression, which included a CBT intervention, to compare the effectiveness of different types of therapy, different components and combinations of components and aspects of delivery used in CBT for depression. Outcomes were pooled using standard and component-level network meta-analysis (NMA). Among 91 studies included, there was strong evidence that CBT interventions resulted in a larger short-term decrease in depressive symptoms compared with treatment as usual (TAU). The standardised difference in mean change for face-to-face (F2F) CBT compared with TAU was −1.11 [95% credible interval −1.62 to −0.60]; for hybrid CBT was −1.06 (−2.05 to −0.08); and for multimedia CBT was −0.59 (−1.20 to 0.02). A wait list control was detrimental compared with TAU [0.72 (0.09 to 1.35)]. While multimedia and hybrid CBT may be as effective as F2F CBT, there was substantial uncertainty in the estimates of treatment effectiveness. We found no evidence of specific effects of any content components or combination of components. Delphi Study on effective components of cognitive–behavioural therapy We aimed to establish an expert consensus on the effective components of CBT for adults with depression. An international panel of CBT experts (n = 120) was invited to participate in an online survey. In round 1, experts rated the effectiveness of 35 items covering both content and process components of CBT. In a second round, experts rerated components to reach a consensus. Of those invited, 32 participated in round 1 and 21 also provided data in round 2. Consensus was achieved in relation to nine content components (that facilitate behaviour change) and three process components (procedures for therapy delivery). Generic therapeutic competences comprised five of the nine content components. There was less agreement about the effectiveness of cognitive components of CBT. Cost-effectiveness of different formats for delivery of cognitive–behavioural therapy for depression We developed a decision model to evaluate the cost-effectiveness of F2F CBT, multimedia CBT and hybrid CBT, given in addition to TAU, in comparison with TAU alone. F2F and hybrid CBTs were modelled by treatment intensity defined by combinations of number and length of CBT sessions. The model covered an average treatment period of 4 months with a 5-year follow-up period to extrapolate long-term cost-effectiveness. The model inputs were derived from our NMA and the literature. The primary outcome was quality-adjusted life-years (QALYs). All CBT modes given in addition to TAU were more cost-effective than TAU alone. Probabilistic sensitivity analyses found that F2F CBT with intensities of six 30-minute sessions and sixteen 60-minute sessions had the highest probability of being cost-effective. However, neither option reached 50% probability of being most cost-effective (32.5% and 31.1%, respectively). There was substantial uncertainty around estimates. Development of the online therapy platform This comprised: (1) an iterative design stage, (2) a pilot study and (3) final refinement. In phase 1, we held individual interviews, prototype testing sessions, platform walkthroughs and workshops with stakeholders aimed at understanding the design requirements for the platform. Feedback informed the intervention design. Three requirements were identified for integrated CBT therapy platforms: (1) features to overcome depression-related barriers; (2) features that support engagement; and (3) that reinforce learning and support the acquisition and learning of new skills. Therapists highlighted the importance of collaborative working, and the impact of technology on therapists’ workflow and workload, and its potential in supporting clients’ engagement. In phase 2, we conducted a pilot study to evaluate usability and user experience of the intervention. Patients with depression were recruited from primary care and offered a course of integrated therapy using the newly developed platform. Qualitative interviews were conducted with patients, including those who completed and withdrew from therapy. Eighteen patients with depression were recruited from primary care. Of these, 10 completed therapy. Initial interviews were conducted with 13 patients (after 3–6 therapy sessions), and 9 ‘end of therapy’ interviews were completed. Patients appreciated the first session being F2F so they could meet their therapist and build rapport. Typing limited the amount discussed during online sessions, but some patients noted it aided focus and promoted reflection. Patients said that the integrated approach made therapy more accessible, but not all patients engaged with between-session tasks. Some found the worksheets too complex. Usage data showed that patients reviewed transcripts of the instant messaging therapy sessions and commented that these were a useful learning aid. Some technical issues were experienced that sometimes led to lower engagement. Phase 3 involved final refinement of the platform following feedback from patients, therapists and wider stakeholders, and input from the multidisciplinary study team. Platform security was assessed. Development of cognitive–behavioural therapy materials for the online platform We conducted a survey of 3665 accredited UK CBT therapists asking about their use and views of resources commonly described in CBT manuals. Overall, 994 individuals (27%) responded and a further 33 completed the questionnaire online. Over 85% of respondents used symptom measures, lists of problems/goals, activity schedules, behavioural activation diaries/plans, and case formulation worksheets ‘frequently’ or ‘very frequently’. Selection of platform materials was based on: therapist survey findings; the competency framework for CBT for depression; Delphi findings; and systematic review and NMA findings. Most worksheets were selected from resources already available and familiar to therapists. Development of a training package for therapists Training for the therapists employed for WP1.3 – phase 2 was developed and delivered with CBT experts in the research team and supported by the therapists’ supervisor. Online training for the RCT included lectures covering the trial background, therapy protocol, managing risk, and role play. Work package 2: randomised controlled trial We conducted a pragmatic RCT of 451 patients with depression from 67 general practices in 3 UK sites. Patients aged ≥ 18 years, scoring ≥ 14 on the Beck Depression Inventory, version 2 (BDI-II) and who met International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) criteria for depression were eligible. Patients were individually randomised to either integrated CBT [in addition to usual care (UC)] or to continue with usual general practitioner care (UC). The primary outcome was BDI-II score at 6 months post randomisation. Secondary outcomes included response and remission (based on BDI-II), other measures of depression and anxiety [Patient Health Questionnaire-9 items (PHQ-9)/Generalised Anxiety Disorder-7 (GAD-7)], function [Work and Social Adjustment Scale (WSAS)], EuroQol-5 Dimensions, five-level version, and costs of interventions and wider services at 6 and 12 months. The primary analyses were of 334 patients (171 integrated CBT; 163 UC). Those in the intervention group had a BDI-II score that was, on average, 4.4 points lower (less depressed) than those in the UC group at 6 months {difference in means: −4.4 [95% confidence interval (CI) −7.0 to −1.9]; p = 0.001}. In repeated-measures analyses using data from 6 and 12 months, the intervention group had a BDI-II score that was, on average, 3.8 points lower than the UC group (95% CI −6.1 to −1.5, p = 0.001). The intervention group had a twofold increased odds of response (≥ 50% improvement in symptoms) and remission (BDI-II score < 10), fewer symptoms of depression (PHQ-9) and anxiety (GAD-7), and less functional impairment (WSAS). Work package 3: economic evaluation We assessed cost-effectiveness from three perspectives: NHS and Personal Social Services (PSS); participants and their informal carers; and societal. Costs were collected accordingly. The primary analysis was an incremental cost–utility analysis of integrated CBT over and above UC, over 12 months from an NHS and PSS perspective. The incremental cost-effectiveness ratio (ICER) was calculated by dividing incremental costs by incremental QALYs. The mean costs of integrated CBT were £987 [standard error (SE) £14] per participant. The mean costs of UC were estimated at £382 (SE £57) in the UC group and £193 (SE £42) in the intervention group. In the primary analysis, costs from the NHS/PSS perspective were £753 (SE £77) per participant in the UC group and £1754 (SE £127) in the intervention group, with adjusted incremental costs of £1009 (95% CI £737 to £1286). The mean QALYs were 0.554 (SE 0.017) in the UC group and 0.597 (SE 0.017) in the intervention group, with adjusted incremental QALYs at 0.033 (95% CI −0.002 to 0.059). The ICER was £30,576 per QALY gain (probability of cost-effectiveness from £20,000 to £30,000: 13%–39%). Complete-case analysis (CCA) from NHS/PSS perspective showed a slightly more favourable picture with ICER at £22, 421 (probability of cost-effectiveness: 37%–66%). Work package 4: qualitative evaluation A qualitative study was nested within the main trial. We interviewed 30 patients (20 who received the intervention and 10 from UC), 9 therapists and 3 therapist supervisors. Data were analysed using thematic analysis. The combination of one-to-one sessions with a therapist and having access to integrated online CBT resources enabled patients to better manage their depression. Benefits included the opportunity to review transcripts to clarify homework tasks and track progress in managing their depression. Those who completed therapy valued talking to a therapist and accessing CBT resources within a single platform. Those who did not complete therapy found it difficult to express themselves through typing and found the CBT approach too demanding. As typing limited what could be covered in a single session, therapists restricted the session agenda and asked more focused questions. Conclusions Developing an integrated approach to delivering cognitive–behavioural therapy Based on the systematic review and NMA of trials of CBT interventions, we found no evidence of specific effects of any content components or combinations of components of CBT interventions to inform the platform development. Our decision model showed that there was uncertainty around estimates of cost-effectiveness for different treatment modalities and intensities limiting the extent to which this work could inform the intervention design. Through our survey of CBT therapists, we identified the most frequently used resources. The Delphi study provided insight into the generic therapeutic competences that CBT experts agreed on. In the absence of evidence on the effective or cost-effective components of CBT from the earlier systematic review, NMA and decision model, the findings from the survey and Delphi study were used to map and identify the CBT resources required within our online platform that aligned with the CBT competences framework. Iterative development of the online therapy platform resulted in a prototype platform that was acceptable to patients and therapists. Integrated CBT was provided to a small sample of depressed patients recruited from primary care. Effectiveness, cost-effectiveness and acceptability of an integrated approach to delivering cognitive–behavioural therapy Integrated CBT was effective in reducing depressive symptoms in primary care patients with depression. Benefits were also seen for the outcomes of response and remission in depressive symptoms, and symptoms of anxiety. Improvements were maintained over 12 months. The intervention was not cost-effective based on the current thresholds used by NICE. However, the intervention cost included training costs that would be unlikely to be maintained at the same level if the intervention was rolled out in NHS services. Further, once trained, therapists could treat more patients (beyond the study sample) which would reduce the per-patient training cost and may make the longer-term cost–benefit more favourable. The intervention was acceptable to patients and therapists. The qualitative study highlighted the importance of establishing patient and therapist goals and expectations about what can be achieved in CBT mediated by typing. Some patients are comfortable communicating via typing and are motivated to utilise online resources between sessions. Exploring the benefits and challenges of integrated CBT with patients will enable them to make an informed choice about referral for this novel therapy. Implications for health care The COVID-19 pandemic resulted in many services moving to remote delivery of therapy. There has also been a proliferation of mental health apps in recent years, but with little empirical evidence to support their use. The integrated approach that we have developed is of proven effectiveness. While the intervention was just above the upper limit in terms of accepted thresholds of value for money, the costs of training may be over-estimated compared to those incurred if this intervention was implemented within the NHS. Importantly, the approach was acceptable to patients and therapists and as such offers the potential to increase access for those who find it difficult to attend therapy appointments in person. Recommendations for research Future research needs to examine The longer-term clinical and cost-effectiveness of integrated cognitive–behavioural therapy Studies of in-person CBT have found that CBT is a clinically and cost-effective intervention over four years. Evidence is thus needed to quantify and substantiate the long-term gain for this integrated CBT approach. Which aspects of the platform led to improvements in depression This may enable refinement of the intervention to increase benefits. Study registration This study is registered as ISRCTN14850613 (phase 2 pilot study) and ISRCTN13112900 (RCT). Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0514-20012) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 13. See the NIHR Funding and Awards website for further award information.

Background Patients who present acutely to hospital are often suspected to have bacterial infections and are prescribed antibiotics rapidly. Emerging diagnostic information then shows many of these prescriptions were, or no longer are, necessary, requiring prescribers to stop antibiotics: but there is little evidence for how to support this. United Kingdom ‘Start Smart then Focus’ guidance promotes early effective antibiotics followed by active ‘review and revision’ 24–72 hours later. However, in 2017 < 10% of antibiotic prescriptions were discontinued at review, despite studies suggesting that 20–30% could be stopped safely but are not, largely because of behavioural and organisational factors. Objectives To develop and comprehensively evaluate a complex digital, organisational and behavioural change intervention (i.e. the Antibiotic Review Kit) designed to reduce antibiotic use among adult acute/general medical inpatients by increasing appropriate decisions to stop antibiotics at clinical review. Design and methods To provide underpinning data to design the intervention, we conducted: (1) an overview of the systematic review evidence for the effectiveness and safety of shorter versus longer antibiotic courses in hospitalised patients; (2) hospital-level and individual-level observational analyses of electronic health records to investigate associations between antibiotic usage and outcomes in inpatients; and (3) qualitative studies with healthcare professionals and patients/carers to deepen understanding of perceptions and experiences of inpatient antibiotic prescribing and to identify amenable barriers to change. We used findings from these studies to develop a multi-component healthcare professional and inpatient/carer intervention to support antibiotic ‘review and revise’, using a theory-, evidence- and person-based approach with iterative stakeholder consultation, and targeting practicality, generalisability and sustainability. We then tested its feasibility in one hospital, evaluated using mixed methods, before evaluating its effectiveness and safety in acute/general medical admissions in a non-randomised internal pilot in three hospitals and then a large randomised stepped-wedge cluster trial in 36 hospitals (randomisation unit) across the United Kingdom. The intervention was implemented in two hospitals per month over approximately 18 months. A further minimum 14-month follow-up in all hospitals assessed sustainability of any intervention effect. Analyses used meta-analytic techniques to combine hospital-specific immediate intervention effects (step-change) and underlying trends pre and post implementation, derived using negative binomial regression models for antibiotic outcomes and logistic regression models for binary outcomes. Process evaluation assessed which intervention components were used/useful. Setting and participants Acute/general medical specialties in acute hospital groups; participants were prescribers and non-prescribers involved in decision-making around stopping antibiotics and the inpatients themselves. Interventions The Antibiotic Review Kit comprised short internet-based training, standardised systems to review patients (the Antibiotic Review Kit decision aid), regular support from pharmacists/infection specialists through audit and feedback and an information leaflet for patients. Main outcome measures Coprimary outcomes were 30-day mortality post-admission (non-inferiority) and defined-daily-doses of antibiotics per acute/general medical admission (superiority). Data sources All outcomes were collected from routine electronic National Health Service databases. Results A key barrier to stopping antibiotics was lack of information about the original prescriber’s rationale for, and degree of certainty about their initial antibiotic prescribing decision. An integral component of the Antibiotic Review Kit was developing and optimising a decision aid to increase transparency around initial prescribing decisions. In the 3 pilot and 36 main trial sites (total = 39) who implemented the intervention between 25 September 2017 and 1 July 2019, adjusted estimates showed an immediate change of –1.0% [(95% confidence interval −4.0 to +2.1); p = 0.54] in total antibiotic defined daily doses per acute/general medical admission following the intervention and then sustained reductions of −4.8% per year (−9.1 to −0.2) (p = 0.042) greater post versus pre implementation. Among 7,160,421 acute/general medical admissions, the Antibiotic Review Kit intervention was associated with weak evidence of an immediate change of −2.7% [(−5.7 to +0.3); p = 0.079] in adjusted 30-day mortality and a year-on-year post versus pre implementation change of +3.0% per year [(−0.1 to +6.2); p = 0.060]; the latter was not found after excluding admissions after the COVID-19 pandemic started in March 2020 [year-on-year change −0.9% (−4.0 to + 2.4) (p = 0.60)]. Limitations COVID-19 had an important impact during the trial, and necessitated changes to the statistical analysis plan that could not have been foreseen. It also affected our ability to assess cost-effectiveness, since estimates were strongly affected by non-significant mortality excesses which were plausibly due to COVID-19. Conclusions The Antibiotic Review Kit intervention resulted in sustained reductions in antibiotic use among adult acute/general medical inpatients. The weak, inconsistent intervention effects suggesting a possible initial decrease in mortality associated with the intervention, followed by a longer-term increase, are likely due to the COVID-19 pandemic. Hospitals should consider using the Antibiotic Review Kit to reduce antibiotic overuse. Future work Most sites were using paper-based prescribing systems during the trial; implementing the Antibiotic Review Kit in electronic-based prescribing, now used by almost all National Health Service Trusts, can be more complex. Further research is needed to support recommendations for the Antibiotic Review Kit, and antimicrobial stewardship functionality more widely, in electronic prescribing systems. Study registration The overview of systematic reviews was registered with PROSPERO (CRD42016046907). The feasibility, pilot and main cluster-randomised trial was registered with ISRCTN (ISRCTN12674243). Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0514-20015) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 10. See the NIHR Funding and Awards website for further award information. Plain language summary Antibiotics kill bacteria, but do not cure viral infections. When sick patients arrive at hospital, doctors have to work out whether they have a bacterial infection, viral infection or another illness. This is difficult, because initial symptoms can be very similar (like cough, temperature). Finding out exactly why someone is sick can take time, so it often makes sense to give antibiotics straightaway, even before some test results are back. However, government guidance recommends stopping antibiotics when it becomes clear that they are not needed. This could be because doctors work out that the patient never had a bacterial infection, or because the patient has got better. Stopping antibiotics as soon as it becomes clear they are not needed cuts the risk of bacteria becoming resistant to antibiotics. But often antibiotics are not stopped when they should be. Sometimes this is ‘just in case’ or thinking ‘better safe than sorry’, or because of the belief that antibiotic courses must always be completed to avoid resistance. We developed a package of strategies (internet-based training, standardised systems to review patients, regular support from pharmacists/infection specialists and patient materials) to help doctors, nurses, pharmacists and patients stop antibiotics in hospitals when they are no longer needed. It was based on summarising all the evidence that stopping antibiotics is safe comparing information about antibiotic use and deaths in people in hospitals across England getting healthcare workers’ views about how patients on antibiotics were treated in hospitals and how antibiotic reviews could be improved talking to members of the public about what they needed to know about stopping antibiotics. We tested the package in 40 hospitals across the United Kingdom. It led to long-term reductions in antibiotic use among adult acute medical inpatients. We did not find evidence which reached statistical significance of an increase or decrease in mortality. This Antibiotic Review Kit could be considered to help hospitals to reduce unnecessary antibiotic use. Scientific summary Background Patients who present acutely to hospital are often suspected to have bacterial infections and are prescribed antibiotics rapidly. Emerging diagnostic information then shows many of these prescriptions were, or no longer are, necessary. To reduce antibiotic overuse, hospitals therefore target the continued ongoing indication for antibiotics, but there is little evidence for how to support this. UK ‘Start Smart then Focus’ guidance promotes early effective antibiotics followed by active ‘review and revision’ 24–72 hours later. However, in 2017, < 10% of antibiotic prescriptions were discontinued at review, despite studies suggesting that 20–30% could be stopped safely. Research suggests this is due to behavioural and organisational factors. Objectives To develop and evaluate a complex digital, organisational and behavioural change intervention [i.e. the Antibiotic Review Kit (ARK)] designed to reduce antibiotic use among adult acute/general medical inpatients by increasing appropriate decisions to stop antibiotics at clinical review. Methods To provide underpinning data and address concerns that stopping antibiotics early in those with genuine bacterial infections could compromise patient outcomes, we conducted an overview summarising the systematic review evidence for the effectiveness and safety of shorter versus longer antibiotic courses in hospitalised patients. We conducted hospital-level and individual-level observational analyses of electronic health records to investigate associations between antibiotic usage and outcomes in inpatients. Completely adjusting for confounding in such analyses is impossible; we aimed to assess whether there was strong evidence of harm (greater rates of treatment failure/mortality associated with lower antibiotic use), when confounding would have lesser impact. A theory-, evidence- and person-based approach was then used to develop and optimise the ARK intervention. This involved iterative stakeholder consultation and in-depth qualitative research with doctors, nurses and pharmacists in UK hospitals, particularly focusing on stopping antibiotics. Barriers to and facilitators of the intervention and its implementation, and ways to address them, were identified and used to inform the intervention’s development, which targeted feasibility, generalisability and sustainability. In parallel, we co-designed patient materials with members of the public. We then tested the intervention’s feasibility in one hospital, evaluated using mixed methods, including a qualitative interview study with patients to refine the patient materials. The effectiveness and safety of the finalised ARK intervention were then evaluated in acute/general medical admissions. Hospital organisations were the unit of randomisation, since the intervention could only be implemented at this level without contamination. These were a single hospital or group of hospitals organised with one executive board and governance framework (National Health Service trusts in England; health boards in Northern Ireland, Wales and Scotland), denoted ‘hospital’ subsequently. The main study comprised an internal non-randomised internal pilot in three hospitals (the first three to volunteer) and a cluster-randomised stepped-wedge controlled trial in 36 hospitals. Hospitals were eligible if they admitted adult non-elective acute/general medical patients, had a local representative to champion the intervention, and could provide the required study data. Hospitals were randomised to an implementation date at 1- to 2-week intervals over approximately 18 months using computer-generated sequences in seven calendar-time blocks. The date was concealed until 12 weeks before implementation, when local preparations for implementation were designed to start. The behavioural intervention was delivered to healthcare professionals involved in antibiotic prescribing or administration in adult inpatients admitted to acute/general medicine. Sustainability of intervention effects was assessed over a minimum 14 months at all hospitals (through October 2020). Outcomes were assessed using de-identified electronic health records in individual adult acute/general medicine admissions (including admission characteristics, mortality and Clostridioides difficile tests), ward-level antibiotic dispensing, prescription audits, and an implementation process evaluation. Coprimary outcomes were monthly antibiotic defined daily doses (DDD) per adult acute/general medical admission (hospital level, superiority) and all-cause mortality within 30 days of admission [patient level, non-inferiority margin (absolute) 5%]. Outcomes were assessed in the modified intention-to-treat population (i.e. excluding seven sites that withdrew after being placed into the randomisation list but before being informed of their implementation date and were replaced). Intervention effects were estimated using interrupted time series analyses within each hospital from the randomised implementation date, with overall effects estimated using random-effects meta-analytic techniques, and heterogeneity across prespecified potential modifiers assessed by meta-regression. Process evaluation assessed which intervention components were used/useful. Health-economic evaluations took a NHS perspective and compared the counterfactual situations of implementing the intervention on the same date in all hospitals versus never implementing the intervention. Antibiotic resistance among gut microflora was assayed in discarded faecal samples from three hospitals exemplifying different antibiotic prescribing patterns to objectively measure impact on antibiotic selection pressure. Results The overview of systematic reviews found no overall evidence of differences between shorter versus longer durations, but the overall quality of the underlying evidence was rated from very low to moderate. In the hospital-level and individual-level observational analyses we found no evidence of higher mortality risk associated with either lower antibiotic usage (hospital level) or earlier discontinuation of inpatient antibiotics (individual level). A key barrier to stopping antibiotics was lack of information about the original prescriber’s rationale for, and degree of certainty about, the initial need for antibiotics. The ARK decision aid was developed and optimised to capture and communicate the certainty surrounding initial antibiotic prescriptions. It categorised initial prescriptions as either ‘possible’ risk of infection (infection less likely, antibiotics are prescribed as a precaution) or ‘probable’ risk of infection (infection likely but more information is required before finalising the diagnosis/treatment). At review, prescriptions are either stopped (no continued need for antibiotics) or finalised (choice of final agent, route and duration). Ultimately ARK comprised: (1) short internet-based training (7–10 minutes) for staff involved in antibiotic prescribing, focusing on the evidence supporting shorter durations being sufficient, (2) a decision aid to support review and revise, (3) audit and feedback tools for antibiotic stewardship specialists to use in discussions with front-line staff during clinical team meetings and (4) a patient leaflet. A selection of resources was also developed to help ARK Champions (site leads) and their teams implement the ARK flexibly within their local context. The intervention was first tested in a feasibility study in one hospital. All the intervention components were introduced successfully. One hundred and thirty-two staff completed the online tool (19.4 healthcare professionals per 100 acute beds), including 97% (32/33) of the clinical staff who had been pre-specified by the local site team as essential for its uptake (target 70%). Of 588 prescription charts included in 7 point-prevalence surveys, 82% (76–90% at each survey) used the decision aid (target 50%). More prescriptions were categorised as ‘probable’ (335/588, 57%) rather than ‘possible’ (135/588, 23%) using the decision aid, but the proportion of ‘possibles’ increased over the 12-week implementation period {odds ratio [OR] possible vs. probable = 1.11 per week [95% confidence interval (CI) 1.04 to 1.17]; p = 0.001}. Overall prescription review rates increased from 91% (69/76) at baseline to 99% (450/457) post implementation [OR post vs. pre implementation = 6.52 (95% CI 2.22 to 19.16); p = 0.001]. Antibiotic stop rates at 72 hours increased significantly post implementation, to 36% (156/450) (range: 29–40% at each survey) from 9% (6/69) pre implementation [OR = 5.57 (95% CI 2.36 to 13.16); p < 0.001], and were higher for prescriptions categorised as ‘possible’ (48%, 51/106) than those categorised as ‘probable’ (29%, 75/260) [OR possible vs. probable = 2.29 (95% CI 1.44 to 3.64), p = 0.001], with a similar trend for those where the decision aid was not used (36%, 30/84) [OR possible vs. not used = 1.67 (95% CI 0.93 to 3.00); p = 0.09]. The 15 patients taking part in the qualitative study rated the leaflet highly on providing the information they needed, being helpful, trustworthy, and something they would recommend to others, with 13/15 (87%) stating that they would be likely or extremely likely to recommend it to friends and family taking antibiotics in hospital. Three pilot sites implemented the finalised ARK intervention between 25 September 2017 and 20 November 2017; 43 further sites were randomised to implement between 12 February 2018 and 1 July 2019, of which 7 withdrew before implementation (excluded from analyses as no data were collected), leaving 36 randomised sites. The 39 included sites (pilot + randomised) were distributed across the UK (32 England, 4 Northern Ireland, 2 Wales and 1 Scotland). In the 12 months before implementation, antibiotic use varied widely across sites, with WHO’s Access, Watch, and Reserve antibiotics accounting for 0.3–5.7% of total DDDs, Watch for 4.8–44.1% and Access for 30.7–85.2%. Antibiotic prescribing was paper-based in 25 (64%) sites at implementation. Twenty-one (54%) sites implemented the decision aid with a hard stop, meaning prescribers had to stop initial prescriptions within 72 hours unless already finalised, while 9 (23%) sites implemented as a soft stop, meaning prescribers were reminded to stop or finalise prescriptions within 72 hours, and the remaining 9 (23%) sites did neither. Implementation fidelity varied, with sites achieving a median 6 of the 8 pre-specified fidelity criteria [interquartile range (IQR): 5–7] and 9 (23%) sites achieving ≤ 4 criteria. Over the first 12 weeks, sites aimed to have ≥ 70% of essential staff complete the online training, but 16 (41%) sites did not meet this target. A median 52% (IQR: 31–76) of audited antibiotic prescriptions were categorised using the decision aid, and prescription review rates were high at implementation (median 91.0%) and 12 weeks later (median 89.9%). Prescriptions were more often stopped at review in the 12 weeks after implementation than at baseline (median 16.2% vs. 12.7%, Wilcoxon-matched pairs p = 0.006), and increases were greatest in sites with lower baseline rates. Implementation was associated with an overall adjusted −1.0% (95% CI −4.0 to +2.1) (p = 0.54) immediate change in the coprimary end point total DDDs/admission and a sustained −4.8% per year (95% CI −9.1 to −0.2) (p = 0.042) greater change in trend post versus pre implementation (incidence rate ratios from negative binomial regression). Among 7,160,421 admissions, implementation was associated with an overall immediate change in adjusted 30-day mortality odds of −2.7% (95% CI −5.7 to +0.3) (p = 0.079), which was consistent with non-inferiority (ORs from logistic regression). There was weak evidence of an overall year-on-year change in adjusted 30-day mortality of +3.0% (95% CI −0.1 to +6.2) (p = 0.060) post versus pre implementation, which was not found after excluding admissions after the onset of the COVID-19 pandemic in March 2020 [year-on-year change post vs. pre implementation −0.9% (−4.0 to +2.4) (p = 0.60)]. Implementation was associated with immediate reductions in DDDs/admission for broad-spectrum agents, Watch antibiotics and quinolones, and immediate increases in Access antibiotics. There were also sustained year-on-year reductions in Access, Watch, narrow-spectrum, and oral antibiotics. There was no evidence of sustained change in broad-spectrum or parenteral antibiotics, but carbapenems increased year-on-year with similar trends for Reserve agents, albeit from low levels and absolute increases were small. Overall implementation fidelity was not associated with effects on total DDDs/admission but select implementation fidelity criteria were associated with intervention effects on the coprimary end points, suggesting overall fidelity may be averaging across elements, only some of which are drivers of change. For example, sites with processes for ongoing audit and feedback in place by implementation had greater immediate reductions in DDDs/admission (interaction p = 0.02) and there was also weak evidence of greater sustained year-on-year reductions in total DDDs/admission among sites that introduced ARK categories into the prescribing process by implementation, and with higher uptake of the online learning (both interaction p = 0.08). There was no evidence of an impact of a hard stop (interaction p > 0.5). Overall, qualitative process evaluation showed that Champions and core team members viewed the ARK intervention as important and worthwhile. Many felt it fitted well with existing antibiotic guidelines and stewardship activities. Changing the drug chart to include the ARK decision aid categories was straightforward for some, but more challenging for others, particularly when involving electronic prescribing systems. A 48-hour review was considered particularly problematic when it fell at a weekend, when there would be less availability of senior staff to review the prescription. The average cost per hospital of implementing ARK was £27,440 [standard deviation £3792; range £16,230–£27,353]. These costs included both initial (such as training, kick-off meetings) and ongoing activity costs throughout the 33-month period (training and audit). The total intervention implementation cost across all 39 hospitals was estimated at £1,070,147, with additional consults adding a further £580,127. Total cost savings from antibiotics were estimated at £3,635,492 (95% CI −£1,427,888 to £8,887,271). Overall, implementing the intervention provided estimated savings of £124 (95% CI £101 to £145; equivalent to a 4% reduction in mean costs) per acute/general medical admission, mainly due to lower length of stay and hence excess bed-day costs, resulting in estimated total savings of £593,118,060 (95% CI £488,509,723 to £688,750,179) in admission costs across all 39 hospitals. There was no strong evidence that implementation of the intervention led to an increase in mortality, with a non-significant estimated increase in the number of deaths of 5997 (95% CI −2369 to 15,012) during the 33 months following implementation of the intervention in all 39 hospitals (likely affected by incomplete adjustment for the impact of COVID-19, see above). This equates to a non-significant increase of 1.1 deaths per hospital per week (95% CI −0.5 to 2.8), and a resulting non-significant estimated quality-adjusted life-year (QALY) loss of 43,125 (95% CI −15,867 to 104,796). Combining these estimates, the ARK intervention was estimated to reduce healthcare costs without outweighing the (statistically non-significant) reduction in QALYs [mean incremental net monetary benefit: −£267M]. The ARK intervention would be the optimal choice if the reduction in antibiotic consumption led to future cost savings of £267M, due to avoidance of antibiotic harms, such as antibiotic resistance. Under the assumption that the weak association between ARK implementation and 90-day mortality in the trial was due to a combination of chance (95% CI includes zero) and bias (e.g. the effect of the COVID-19 pandemic), ARK implementation would be clearly cost-effective, given the relatively low implementation costs (~£1M at 39 hospitals) compared to estimated reductions in antibiotic (~£3.6M) and admission (~£593M) costs. We found no evidence of differences in taxonomic diversity or antimicrobial resistance (AMR) gene carriage across three acute hospitals with different prescribing patterns, or over time within each hospital, suggesting that this method was not sufficiently sensitive to evaluate the impact of different antimicrobial stewardship (AMS) policies on AMR burden. Conclusions The ARK intervention resulted in sustained reductions in antibiotic use among adult acute/general medical inpatients. The weak, inconsistent intervention effects on mortality and cost-effectiveness are probably explained by the COVID-19 pandemic. Hospitals should consider using the ARK to reduce antibiotic overuse. Recommendations for future research are as follows: Most sites used paper-based prescribing during the trial. Implementing ARK in the different electronic prescribing systems now deployed in the NHS is more complex. Research is needed to support recommendations for AMS functionality more widely in electronic prescribing systems. Some hospitals chose to manage anxieties about prescribing shorter courses by employing an alert recommending stopping antibiotics rather than a ‘hard stop’; a further topic for future investigation is the extent to which this adaptation may negatively impact antibiotic stop rates. Given our incomplete understanding about the impact of shorter versus longer antibiotics on AMR, further studies should investigate whether more detailed/pooled sampling would provide measures of the impact of differential prescribing strategies on AMR. How to incorporate potential benefits due to reducing antibiotic use into cost-effectiveness analysis remains unclear, given the low cost of most antibiotics and the fact that the major detrimental effects of resistance are likely to be in the future, and in patients other than the one being treated for an infection now. Study registration The overview of systematic reviews was registered with PROSPERO (CRD42016046907). The feasibility, pilot and main cluster-randomised trial was registered with ISRCTN (ISRCTN12674243). Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0514-20015) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 10. See the NIHR Funding and Awards website for further award information.

Background People with penicillin allergy labels receive more antibiotics, different antibiotics and have worse health outcomes when compared to people without a penicillin allergy label. Objective To determine if a pre-emptive penicillin allergy assessment pathway can remove incorrect penicillin allergy labels, improve antibiotic prescribing and health outcomes, and is cost-effective. Design A multi-work package programme to develop and evaluate a complex intervention (penicillin allergy assessment pathway), including rapid literature review (work package 1.1), behaviour change material development (work packages 1.2 and 1.3), medical record updating processes (work package 1.4), a pragmatic randomised controlled trial (work package 2.1) with nested pilot (work package 1.5), health economic evaluation (work package 2.2) and process evaluation (work package 2.3). Setting English general practices and hospital sites in West Yorkshire, South Yorkshire and Cornwall. Participants Trial participants were adult general practice patients with a record of penicillin allergy. Interview subjects were general practitioners and healthcare workers involved in penicillin allergy assessment, trial and non-trial participants with a penicillin allergy. Interventions Complex intervention involving: identification of appropriate participants, risk stratifying and testing for penicillin allergy, promoting behaviour change and updating medical records. Main outcome measures Penicillin prescribing, total antibiotic prescribing, treatment response failure, duration of symptoms, hospital admission/length of stay, mortality, meticillin-resistant Staphylococcus aureus status and Clostridioides difficile infection, penicillin allergy de-labelling’ rates. People’s views about penicillin allergy and testing. Results Work package 1.1: Our review showed that patients had some worries about penicillin allergy testing but could see advantages. Clinicians were uncertain about referral criteria. Clinicians were wary about prescribing and patients were wary about consuming penicillins after a negative test. Fear of subsequent reactions were drivers. This informed WP1.3 development of resources to aid penicillin allergy assessment. Work package 1.2: Our qualitative research showed that patients were unaware of the benefits of penicillin allergy testing. Experience of negative effects of penicillin allergy labels motivated patients to be tested. Clinicians were reluctant to de-label patients using clinical judgement alone. Clinicians and patients were generally supportive of penicillin allergy testing. This informed work package 1.3. Work package 1.3: Materials were developed to increase knowledge of penicillin allergy testing, support beliefs about the importance of testing, and to increase confidence in performing the behaviours safely by providing a supportive environment. Materials comprised information booklets for general practitioners and participants, electronic health record (SystmOne) pop-ups, post testing result cards, trial work instructions/training material, participant and clinician letters. Work package 1.4: General practitioners were happy to update medical records after a negative test result. ‘Marking in error’ was used to update SystmOne if a participant had a negative penicillin allergy test, which removed allergy alerts but maintained an audit trail. General practitioner trial training/procedures informed general practitioners how to update records. SystmOne functions developed to facilitate de-labelling. Work package 1.5: Target recruitment for the pilot target was achieved. Identification of suitable patients from SystmOne was successful. Willingness of general practitioners to refer patients and participants to undergo testing was demonstrated. A data-secure ‘ALABAMA unit’ was set up within SystmOne. Processes for referring and following-up participants via the ALABAMA unit were established. Low symptom diary completion rate indicated case note data would also be needed. Work package 1.6: During the nested pilot trial, multiple measures were developed, approved and implemented to maximise recruitment for the main trial; including additional testing sites were recruited in South Yorkshire and Cornwall. Work package 2.1: The randomised controlled trial found there was a statistically significant increase in the number of patients de-labelled and penicillin prescribing was statistically significantly increased (adjusted relative risk 5.26) and median total defined daily dose of antibiotics was reduced by 25% in the intervention group. There were no significant differences in other secondary outcomes. Work package 2.2: Within-trial assessment indicated the assessment pathway was likely to be cost-effective. The cost-effectiveness modelling suggested the net benefit of testing was high but uncertainties were high. The value of information studies indicated another trial including health outcomes was worthwhile. Work package 2.3: General practitioners were motivated to join the trial because of perceived benefit to patients. General practitioners were positive about the trial design/delivery and support from the trial team. Many participants sought clarity about penicillin allergy status. Participants reported feeling safe and well looked after during testing. Most participants accepted test results and were relieved to receive a negative test that enabled them to take penicillin. The process evaluation provided important information on how to improve future National Health Service implementation. Specifically, the need for a safe, accessible and reliable service where general practitioners need to be provided with step-by-step guidance how to identify patients with different risk of true penicillin allergy to facilitate onward referral/assessment. The resources developed within the programme could be used as part of wider implementation to encourage and support patients and healthcare workers to engage with and deliver penicillin allergy assessment. Limitations The coronavirus disease discovered in 2019 pandemic affected recruitment and reduced the sample size meaning the primary outcome had to be agreed and changed to an antibiotic prescribing outcome. The research was conducted using functionality developed in SystmOne; there needs to be consideration of how to maximise knowledge transfer this functionality to other electronic health record systems. Conclusion The pre-emptive penicillin allergy assessment pathway was safe, increased penicillin prescribing reduced overall antibiotic prescribing, produced sustained de-labelling and was cost-effective. The assessment pathway was well received by participants and general practitioners. We conclude that patients who are recent users of antibiotics and are lower risk for true penicillin allergy would likely benefit from penicillin allergy assessment. Future work A larger trial is needed to determine if penicillin allergy assessment can improve patient health outcomes. Implementation studies are needed to determine how best to safely embed penicillin allergy assessment into clinical practice and to widen access beyond the current referral guidelines. Trial registration This clinical trial is registered as Current Controlled Trials ISRCTN20579216 and ClinicalTrials.gov NCT04108637. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-1214-20007) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 11. See the NIHR Funding and Awards website for further award information. Plain language summary Background Antibiotic overuse has resulted in more infections caused by antibiotic-resistant bacteria which make people ill for longer and increases their risk of death. New ways to reduce antibiotic prescribing are needed. Many people have a penicillin allergy label which means doctors avoid first-line penicillin antibiotics and prescribe alternatives, which may not work as well, may cause more side-effects, can be expensive, and can lead to more antibiotic resistance in bacteria. Most people with a penicillin allergy label are not actually allergic when tested and can safely take penicillins. Aim To find out if a pathway for assessing people with a penicillin allergy label allows better use of penicillin and other antibiotics, improves health outcomes, and is good value for money. Methods We used a mixture of research methods including interviews with patients and doctors and focus groups with experts to come up with a trial design that would work and be acceptable to patients and doctors. We developed ways to help patients get tested and for them to receive penicillins if they were not allergic. We carried out a clinical trial of an allergy testing pathway in comparison to usual clinical care to really find out if it was beneficial. The trial had to be reduced in size because of coronavirus disease discovered in 2019. Results Many people were not aware of the harms linked to penicillin allergy labels. Patients and doctors generally thought penicillin allergy assessment was a good idea but they did not know much about it. Trial materials helped patients and doctors understand the processes. The trial found the penicillin allergy assessment pathway was safe, increased penicillin prescribing fivefold, reduced all antibiotic prescribing, and was probably cost-effective. The pathway was well received by patients and general practitioners. Conclusions Penicillin allergy assessment benefits patients by safely and efficiently allowing them to receive first-line penicillin treatment and by reducing the amount of antibiotics they need. Scientific summary Background The importance of antimicrobial resistance (AMR) and the need to reduce its impact is well recognised. One way to reduce AMR is to improve the quality of antibiotic use as well as reduce the quantity of antibiotics used. Penicillins are first-line therapy for many common infections, but patients with a penicillin allergy label (PAL) are much less likely to be prescribed penicillins (West RM, Smith CJ, Pavitt SH, Butler CC, Howard P, Bates C, et al. ‘Warning: allergic to penicillin’: association between penicillin allergy status in 2.3 million NHS general practice electronic health records, antibiotic prescribing and health outcomes. J Antimicrob Chemother 2019;74:2075–82. https://doi.org/10.1093/jac/dkz127, Blumenthal KG, Lu N, Zhang Y, Li Y, Walensky RP, Choi HK. Risk of meticillin resistant Staphylococcus aureus and Clostridium difficile in patients with a documented penicillin allergy: population based matched cohort study. BMJ 2018;361:k2400. https://doi.org/10.1136/bmj.k2400) and may therefore be disadvantaged. Penicillin allergy labels are common, affecting about 6% of the English population. (West et al. 2019) Importantly, fewer than 10% of these patients are truly allergic (Powell N, Stephens J, Kohl D, Owens R, Ahmed S, Musicha C, et al. The effectiveness of interventions that support penicillin allergy assessment and delabeling of adult and pediatric patients by nonallergy specialists: a systematic review and meta-analysis. Int J Infect Dis 2023;129:152–61. https://doi.org/10.1016/j.ijid.2022.11.026). Spurious PALs arise for many reasons including side effects and symptoms related to the infection requiring antibiotic treatment being mislabelled as allergies; health record systems not easily allowing distinction of allergic symptoms and side effects and allergy waning over time. Consequently, a significant proportion of the population are restricted access to highly effective penicillins. Several negative health outcomes have been associated with PALs. Penicillin allergy records are associated with AMR; for example, evidence from the UK and USA suggests that patients with a penicillin allergy record are more likely to acquire multi-drug-resistant bacteria, including meticillin-resistant Staphylococcus aureus (MRSA) (West et al. 2019, Blumenthal et al. 2018, Macy E, Contreras R. Health care use and serious infection prevalence associated with penicillin ‘allergy’ in hospitalized patients: a cohort study. J Allergy Clin Immunol 2014;133:790–6. https://doi.org/10.1016/j.jaci.2013.09.021). This appears to be because patients with a penicillin allergy are usually prescribed alternative antibiotics like fluoroquinolones, macrolides and clindamycin (Blumenthal et al. 2018). Our preliminary research found macrolide, tetracycline, cephalosporin, quinolone, and clindamycin prescribing were all more common in patients with a record of penicillin allergy compared to those without, and that antibiotic prescriptions were almost twice as frequent in patients with a penicillin-allergy record (West et al. 2019). Systematic review evidence indicates that antibiotic allergies affect health outcomes, increasing mortality, length of hospital stay, and costs (Krah NM, Jones TW, Lake J, Hersh AL. The impact of antibiotic allergy labels on antibiotic exposure, clinical outcomes, and healthcare costs: a systematic review. Infect Control Hosp Epidemiol 2021;42:530–48. https://doi.org/10.1017/ice.2020.1229). The large discrepancy between reported and true allergy rates suggests that introducing a ‘pre-emptive’ penicillin allergy assessment pathway (PAAP) for patients who are more likely to receive antibiotics could impact upon antibiotic prescribing, yield patient benefits, limit AMR/healthcare-associated infection, and deliver NHS cost savings. The focus of ALABAMA was spurious or ‘false-positive’ penicillin allergy records, and whether a trial of a complex intervention aimed at correcting such records is safe, clinically effective and cost-effective. Objectives Workstream 1: Intervention development and feasibility Objective 1a: Development of the complex intervention (work packages 1.1, 1.2, 1.3, 1.4) To design a complex intervention to support and provide a PAAP for primary care patients with a record of a penicillin allergy, following Medical Research Council guidelines. Objective 1b: Feasibility (work packages 1.5 and 1.6) To undertake a nested pilot (NP) trial to assess safety, acceptability and practicality of delivering a trial of the PAAP intervention. Workstream 2: Evaluation and implementation of the penicillin allergy assessment pathway intervention Objective 2a: Full trial (work package 2.1) To conduct a randomised clinical trial of the PAAP intervention versus usual clinical care to assess impact on antimicrobial prescribing and health outcomes. Objective 2b: Cost-effectiveness: to determine whether the penicillin allergy assessment pathway intervention is cost-effective (work package 2.2) Objective 2c: Process evaluation: to conduct a process evaluation of the main trial to assess implementation of the penicillin allergy assessment pathway intervention and to produce recommendations for wider National Health Service uptake (work package 2.3) Work packages, methods and results Workstream 1: Complex intervention development and feasibility evaluation Work package 1.1: Rapid review of evidence regarding patient/prescriber views of penicillin allergy and behaviour change in relation to allergy testing and antibiotic prescribing Aim To synthesise published evidence of patients’ and/or clinicians’ views and experiences of penicillin allergy testing (PAT) services and influences on antibiotic prescribing behaviour in the context of penicillin allergy. Findings Patients had some worries about PAT but could see advantages. Clinicians were uncertain about referral criteria. Clinicians were wary about prescribing and patients were wary about consuming penicillins after a negative test. Fear of subsequent reactions were drivers. Limitations Limited numbers of cross-sectional survey studies were identified, assessing clinical aspects of penicillin allergy tests and accuracy. A full systematic review might have found more studies. Lack of quality assessment and double-checking could have introduced a bias. What this meant For key behaviours to change regarding engagement with penicillin allergy assessment, a behavioural intervention would need to address these issues and target both patients and general practitioners (GPs). Work package 1.2: Qualitative study of patient/prescriber attitudes to penicillin allergy and penicillin allergy testing Aims To identify primary care clinician and patient views on barriers and enablers to PAT and subsequent antibiotic use. Findings Patients were generally unaware of the benefits of PAT. Experience of the negative effects of PALs motivated patients to be tested. Clinicians were reluctant to de-label patients using clinical judgement alone. Clinicians had limited experience of referring patients with PALs for assessment and testing processes. Clinicians described numerous potential benefits of testing. Patients who had been tested reported benefits of the process. There was uncertainty about who should ensure the patient understood allergy test results among GPs. Limitations This was a qualitative study with a purposeful sample that recruited from one region in England; the results should be interpreted cautiously in terms of their transferability to other settings. What this meant The behaviour change component of the intervention would need to provide information for both patients and GPs about PAT and the benefits of penicillin allergy assessment. The GP facing component would need to cover referral processes and interpretation of test results. Work package 1.3: Development of intervention materials to support patient and clinician engagement with penicillin allergy assessment pathway Aim To develop intervention materials to support the implementation and uptake of PAAP by clinicians and patients. Results To support clinicians and patients in undertaking the desired behaviours involved in the PAAP, materials were developed that were intended to increase knowledge of PAT, support beliefs about the importance of testing and the importance of tackling AMR and to increase confidence in performing the behaviours safely by providing a supportive environment. Materials comprised separate information booklets for GPs and patients, electronic health record pop-ups, post testing result cards for patients, trial work instructions and training material, patient and clinician letters. Feedback from think-aloud interviews identified changes required to draft intervention materials in order to increase acceptability and feasibility of the PAAP with patients and clinicians. Limitations A limitation of the ALABAMA intervention package is uncertainty about its wider application outside the trial. The intervention was centred around functionality within SystmOne (an electronic medical record system widely used in general practice). The intervention package developed within ALABAMA will need to be modified to allow full functionality in other GP electronic health records systems. What this meant Behaviour change materials were prepared and ready for embedding into the trial intervention and testing in the pilot trial. Work package 1.4: Design of systems and processes for recording/updating penicillin allergy records Aim To determine how best to update penicillin allergy records to maximise the impact of PAT while maintaining patient safety. Results Qualitative interview research found GPs were happy to update medical records after a negative test result provided there were clear directions. GPs also expressed concerns about patients becoming relabelled as penicillin allergic. GP trial training and standard operating procedures were developed to inform GPs how to update records. To improve de-labelling, a new function in SystmOne was utilised to send a ‘task’ to the GP. Processes were developed to update secondary care records in line with primary care. Limitations The processes that were developed were specific to the SystmOne electronic health record system and equivalent functionality would need to be established for other electronic health record systems. What this meant Processes for updating primary and secondary health records were ready for testing in the pilot trial. Work package 1.5: Assessment of the feasibility of delivering a randomised controlled trial of the penicillin allergy assessment pathway intervention versus usual clinical care Aim To undertake a NP randomised controlled trial (RCT) with feasibility outcomes to assess safety, acceptability and practicality of delivering a trial of the PAAP intervention. Results Before having to pause trial activities in April 2020, due to coronavirus disease discovered in 2019 (COVID-19), 81 participants were consented satisfying stop/go criteria. The Data Management and Ethics Committee (DMEC) reviewed safety data on completion of the NP trial and concluded it was safe to continue to main trial. Feasibility outcomes: Identification of suitable patients from GP electronic health records was successful. Willingness of GPs to refer patients to PAAP and participants to undergo testing was demonstrated. The processes of setting up a data-secure ‘ALABAMA unit’ within the SystmOne electronic health record system, referring participants into the unit, allocating to treatment group, and follow-up via the ALABAMA unit were built, tested and piloted. Research nurse follow-up of primary events and post testing safety checks were demonstrated to be feasible. Symptom diary completion rate was low. Limitations Onset of the COVID-19 pandemic delayed randomisation and testing for some consented participants. COVID-19 interrupted follow-up for only a minority, meaning the planned monthly telephone calls did not occur for one participant. Many patients actively avoided contacting GPs during the pandemic, alongside socialisation restrictions, and use of face mask that collectively resulted in antibiotic prescribing falling during this period, potentially reducing the number of primary events. (Primary events were antibiotic prescriptions for a pre-defined list of infections which prompted the participant to complete a symptom diary for trial outcome data analysis.) What this meant As the pilot recruitment target was achieved, the independent DMEC and Trial Steering Committee (TSC) recommended automatic progression to main trial. The ALABAMA trial restarted 1 November 2020. Given low diary completion rates, it was clear that the 28–30 follow-up call would be needed to ensure ‘treatment response failure’ data could be captured. This confirmed SystmOne could be used to deliver key elements of the ALABAMA trial. Work package 1.6: Understanding recruitment to ALABAMA penicillin allergy assessment pathway trial Aims To optimise recruitment to the trial and understand why patients wanted to participate in the trial. Results Initially there were low rates of conversion from expression of interest to consent. During the NP trial, the following measures were approved and implemented to maximise recruitment: NHS research staff were seconded to the local GP confederation enabling them to work as part of the GP practice team and follow up non-responders/pre-consented participants. Verbal telephone consent, which saved GP/participants’ time. Non-responder letters were sent to patients who did not respond to the first mail-out. Up to two follow-up calls could be made to non-responders by research nurses. Posters for display in GP practices and local pharmacies, to improve awareness. Non-responder follow-up (via mail or telephone) had a noticeable positive impact on response rates. As part of our recruitment strategy for the main trial we recruited additional sites: South Yorkshire and Cornwall. Limitations Formal analysis of the different approaches used to improve recruitment was not undertaken, rather an iterative consensus process was used. What this meant We moved to recruit to the main trial with an improved recruitment strategy and trial processes that had been refined and tested. Measures were put in place to allow recruitment to continue in spite of COVID-19. Workstream 2: Evaluation and implementation of the penicillin allergy assessment pathway intervention Work package 2.1: Randomised controlled trial of the safety, clinical and cost-effectiveness of the penicillin allergy assessment pathway intervention versus usual clinical care, on health outcomes, antimicrobial prescribing and antimicrobial resistance Aim To determine whether a pre-emptive (carried out in advance of need) PAAP intervention is safe and clinically effective in improving antibiotic prescribing and patient health outcomes in patients with a record of penicillin allergy in their primary care electronic health records. Results Penicillin allergy assessment pathway resulted in a statistically significant increase in penicillin allergy de-labelling [risk difference 87.4% (84.1 to 90.9), p-value < 0.0001] and also in penicillin prescribing [adjusted relative risk 5.26, 95% confidence interval (CI) 3.03 to 9.18]. Total prescriptions and total defined daily doses were reduced (adjusted median difference −3.17, 95% CI −6.11 to −0.23). Prescriptions of non-penicillin antibiotics reduced in the PAAP group compared to the usual care group. There were no important differences in safety outcomes. The PAAP intervention did not significantly affect mortality, hospital admissions or hospital length of stay, duration of symptoms, MRSA or Clostridioides difficile infection (CDI) rates or treatment response failure rates. The PAAP intervention significantly increased de-labelling at 3 and 12 months. Limitations The trial took place during the COVID-19 pandemic which impacted on recruitment. The primary outcome was changed, with full consideration and agreement of the DMEC and funder, from ‘treatment response failure’ to one measuring penicillin prescribing. In addition, the trial may underestimate the effect of the intervention as visits to GPs and antibiotic prescribing were both reduced during the COVID-19 pandemic. The trial was open label so there was a risk of ‘contamination’ of the control group through behaviour change materials delivered to GPs. What this meant The PAAP intervention was safe and effective at increasing penicillin use and reducing total antibiotic use. Work package 2.2: Health economic evaluation: assessing the consequences of introducing the penicillin allergy assessment pathway intervention into the National Health Service Aims The aim was to conduct an economic evaluation of the cost-effectiveness of the PAAP versus usual clinical care (UCC) from the perspective of the NHS and Personal Social Services in England. This was done in two parts: first, on a within-trial basis using up to 12-month follow-up and, second, using a decision analytic model with a 5-year time horizon. We also assessed whether conducting a new follow-on RCT of PAAP was worth the extra cost. Results The mean per patient costs of PAAP were £170 at 12 months. Mean (95% CI) cost differences in the intervention group relative to UCC were: −£11 (−24, 3) for primary care, −£114 (−432, 204) for hospital admissions, £0 (−100, 100) for outpatient attendances and −£13 (−42, 17) for emergency care. Mean total costs were £1301 and £1269 in intervention and UCC group, respectively (difference: unadjusted 32.31, imputed + adjusted 55.61); mean quality-adjusted life-years (QALYs) were 0.866 and 0.830, respectively (difference: unadjusted 0.036; imputed + adjusted 0.0051). The within-trial analysis showed that PAAP was associated with an incremental cost-effectiveness ratio of £10,938 per QALY gained relative to UCC. It generated an incremental net health benefit of 0.0023 QALYs and incremental monetary benefit of £41 (58% probability of cost-effectiveness) per patient, at a threshold of £20,000/QALY and 0.0032 and £92 (64%), respectively, at a threshold of £30,000/QALY. The probability of PAAP being cost-effective varied the most by gender and number of antibiotic prescriptions in the year before trial enrolment. The decision tree model yielded predictions of both cost savings and QALY benefits with PAAP versus UCC in the long term. The incremental net monetary benefit was 806 (48% probability of cost-effectiveness). The cost of a new follow-on trial modelled on ALABAMA and involving 1267 participants was estimated to be £2.37 M. The expected value of sample information produced by the new trial from reducing the costs of making the wrong decision based on a 5-year analysis of PAAP costs and benefits is £19.60 M over the adult population in England expected to benefit from PAAP in the next 10 years. Limitations Health-related quality-of-life outcomes of trial participants were collected at baseline and at the trial end, which for some participants was < 12 months due to early trial termination due to COVID-19. The economic modelling relied on simplifying assumptions regarding the long-term cost and QALY outcomes due to the limited follow-up period of the trial. What this meant Penicillin allergy assessment pathway was likely to be cost-effective, modelling suggested PAAP would be cost-effective, but uncertainty was high. A follow-on trial would be expected to be worthwhile. There is, however, additional considerable uncertainty about long-term costs of AMR which we have not factored in as they are out of the scope of this analysis or perhaps would be related to a future trial. Work package 2.3: Process evaluation: assessing implementation of the penicillin allergy assessment pathway intervention Aims We aimed to explore participants and healthcare professionals’ experiences of trial procedures. Specifically, we explored participants and healthcare professionals’ views and experiences of PAT, test results and future antibiotic use. Results Clinicians and patients’ interviews: General practitioners were motivated to join the trial because they bel

Aims To develop, assess the feasibility and test the effectiveness and cost-effectiveness of a multifaceted ePrescribing-based Antimicrobial Stewardship intervention to safely reduce inappropriate antibiotic use in adult medical inpatient settings. Methods We undertook a mixed-methods programme of work organised into four interlinked work packages: Work package 1: (a) planning, developing and optimising all elements of ePrescribing-based Antimicrobial Stewardship; and (b) carrying out qualitative, feasibility and process studies of ePrescribing-based Antimicrobial Stewardship implementation. Work package 2: agreeing secondary outcome measures through an expert consensus building workshop, developing the methods to collect outcome data and establishing the feasibility of evaluating ePrescribing-based Antimicrobial Stewardship through a hybrid cluster-randomised stepped-wedge trial. Work package 3: evaluating the effectiveness of ePrescribing-based Antimicrobial Stewardship in achieving our coprimary outcomes of reducing antibiotic consumption without any adverse impact on 30-day mortality through piloting and undertaking a formal evaluation. Work package 4: estimating the cost-effectiveness of ePrescribing-based Antimicrobial Stewardship, including cost-minimisation and sensitivity analyses through evidence synthesis and expert engagement. Our research (which started on 1 January 2019) was majorly disrupted by the COVID-19 pandemic, with members of the team being redeployed to clinical and government advisory roles, and limited opportunities to undertake hospital fieldwork compounded by strikes. These delays resulted in our trial plans being pushed back by ≈ 3 years, by which time it was no longer possible to recruit sufficient numbers of hospitals into the planned definitive trial. Following discussion with the funders, we aborted plans to run the pilot and definitive trials, replacing this with qualitative scoping work to understand the ways in which trusts had innovated in the intervening period to promote antimicrobial stewardship (work package 5). Results We developed the ePrescribing-based Antimicrobial Stewardship intervention, comprising technical, educational and behavioural components and were able to agree secondary outcome measures and the cost-effectiveness model. The feasibility trial was undertaken in two hospitals in one trust, studying 24,884 antibiotic orders on 1958 admissions. This demonstrated that several aspects of the intervention were deliverable and that we were able to extract data on 20/43 outcome measures. Modelling of the coprimary outcomes, total antibiotic defined daily dose per admission, was feasible and enabled its variability to be estimated to support further research planning. There was, however, limited engagement of clinical staff with the training element of the intervention, which was initially only voluntary. Given the severe delays, many trusts had begun implementing their own approaches to deal with growing challenge of antimicrobial stewardship. As such, we were unable to find sufficient naive sites to progress the trial. Our additional qualitative scoping work identified the myriad ways in which trusts were attempting to respond to antimicrobial stewardship. These included order sets and order protocols, default durations, alerts for review, incorporation of guidelines, and audit and feedback functionality. Limitations We identify five key limitations: (1) the COVID-19 pandemic caused major delays, and non-COVID research activities had a lower priority than front-line clinical work and COVID-related research. The timescale for the development work is, therefore, unlikely to be representative of the time it should take to develop an intervention of this scope outside of pandemic contexts. (2) Although there was considerable stakeholder input into the development of ePrescribing-based Antimicrobial Stewardship, this did not include the entire potential user population. Nurses were under-represented. Although not directly involved in prescribing, more engagement with nurses would have been useful because they were involved in medication management workflows. (3) Undertaking the work in one trust limits generalisability, although inclusion of several ward types should help ensure that our findings related to a breadth of clinical contexts. (4) Implementation of ePrescribing-based Antimicrobial Stewardship within Cerner Millennium (North Kansas City, Missouri, USA) means feasibility in other systems still needs to be established. (5) The add-on study of antimicrobial stewardship practices included views of antimicrobial stewardship pharmacists and consultant microbiologists but did not include all hospitals nor ward prescribers. Nonetheless, this study provided many valuable insights into approaches, including the need for development of evidence for different practices. Conclusions We developed ePrescribing-based Antimicrobial Stewardship and conducted a feasibility trial of this in inpatient settings. This development and feasibility work was, however, undertaken during challenging circumstances resulting from a combination of the COVID-19 pandemic and National Health Service pressures, making it difficult for clinicians to fully engage. Delays with our work resulted in provisionally recruited trusts innovating in relation to antimicrobial stewardship such that it was no longer possible to deliver the planned follow-on pilot and definitive trials. In summary, ePrescribing-based Antimicrobial Stewardship is, in its present format, not suitable for use in National Health Service hospitals. Future work Hospitals are currently pursuing a range of approaches in an attempt to promote antimicrobial stewardship. There is a need to develop a typology of these approaches and establish the strength of the underpinning evidence using naturalistic designs. Trial registration This trial is registered as ISRCTN13429325. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0617-20009) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 12. See the NIHR Funding and Awards website for further award information. Plain language summary Increasingly, bacteria (germs) no longer respond to antibiotics. This can result in patients experiencing more severe or longer infections and may lead to an increased risk of dying from infections. This is due to too much – inappropriate – use of antibiotics. The reasons for overuse of antibiotics include doctors not having relevant information when prescribing, concerns about missing possibly serious infections, and limited opportunities to review diagnoses and stop antibiotics if no longer needed. From our previous research, we found that hospital electronic prescribing (ePrescribing) systems can support new ways of working to safely reduce unnecessary antibiotic prescribing. We built on our previous work to see if ePrescribing systems can safely reduce antibiotic prescribing. We conducted interviews with clinicians and undertook observations in hospitals, and used these insights to develop new tools to support prescribers, training of prescribers and organisational changes. The development and go-live date of this complex intervention, which we called ePrescribing-based Antimicrobial Stewardship, were delayed due to COVID-19. Once pandemic restrictions were lifted, we were able to progress a trial of ePrescribing-based Antimicrobial Stewardship across two hospitals, showing it was possible to implement ePrescribing-based Antimicrobial Stewardship and to extract the information we needed to see if it was reducing inappropriate antibiotic usage. We found little appetite from clinicians to engage with the training and use ePrescribing-based Antimicrobial Stewardship as hospitals were under pressure from pandemic backlogs and National Health Service strikes. The delays to our main trial resulted in other already recruited hospitals having to innovate locally to address antibiotic over-usage. These innovations overlapped with ePrescribing-based Antimicrobial Stewardship such that it was no longer possible to progress the main trial as planned. We decided not to progress the trial, as we would have struggled to recruit enough hospitals to generate a clear answer. Finally, we interviewed National Health Service staff to understand new ways hospitals were attempting to reduce inappropriate antibiotic use. Scientific summary Background Antibiotic-resistant infections are responsible for considerable morbidity and mortality globally. In response to the growing problem of antimicrobial resistance (AMR), there are global efforts afoot to minimise the inappropriate, excessive use of antibiotics, particularly in healthcare settings. Aims We sought to develop, assess the feasibility of and test the effectiveness and cost-effectiveness of a multifaceted ePrescribing-based Antimicrobial Stewardship (ePAMS+) intervention to safely reduce inappropriate antibiotic use in adult medical inpatient settings. Research questions We sought to answer the following research questions that were pursued through five interlinked work packages (WPs): What were the main challenges to and facilitators for reducing inappropriate antibiotic use? (WP1 phase 1 and phase 2a) In what ways could ePAMS+ promote antimicrobial stewardship (AMS)? (WP1 phase 1 and phase 2b) How should ePAMS+ best conceptualised/delivered? [WP1 (phase 1 and phases 2a and b) and WP2] What are suitable secondary outcomes and process measures? [WPs1 (phase 1 and phase 2a and b) and WP2] Was ePAMS+ acceptable/feasible to implement? (WP2) Were our trial procedures acceptable/feasible? (WP2) How much patient-level variability was there in the coprimary outcomes? (WP2) Was ePAMS+ effective in safely reducing inappropriate antibiotic use? (WP3) What were the mechanisms of action/unintended consequences of ePAMS+? [WP1 (phase 2b), WP2 and WP3] Was ePAMS+ cost-effective? (WP4) What approaches were being pursued by hospitals in the pandemic era to promote AMS? (WP5) Methods We undertook a mixed-methods programme of work that comprised of in-depth qualitative work, consensus building approaches, a feasibility trial and evidence synthesis. This work was originally organised into four interlinked WPs, but not all WPs were able to proceed as planned because of the emergence of the COVID-19 pandemic, with associated major impacts on the NHS. A fifth WP was added towards the end of the work to help contextualise our work within the changed AMS landscape in the aftermath of the pandemic. The methods employed in our WPs are summarised below: Work packages 1–5 Work package 1 phase 1 Some text in this section has been reproduced with permission from Mureyi D, Cresswell K, Sivyer K, Heed A, Weir CJ, Adamestam I, et al. The development of a complex digital and behavioural Antimicrobial Stewardship intervention for hospitals in England. BMC Med Inform Decis Mak 2024. https://doi.org/10.21203/rs.3.rs-3715230/v1 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text. ePrescribing-based Antimicrobial Stewardship’s development was informed by a qualitative person-based approach. This involved literature reviews, structured stakeholder workshops and interviews with policy-makers, practitioner and patient representatives, vendors and international experts. Engagements and interviews focused on identifying key barriers to and facilitators of appropriate prescribing and post-prescription reviewing of antimicrobial therapy by users of ePrescribing systems in hospitals. Qualitative data were thematically analysed and, where feasible, used to inform the design ePAMS+ features. Work package 1 phase 2a We conducted 18 semistructured interviews with medical prescribers and pharmacists with varying levels of seniority exploring current AMS practices and investigating potential areas for improvement before ePAMS+ was implemented. Participants were recruited with the help of local gatekeepers. Topic guides sought to explore both formal and informal practices surrounding AMS, and challenges and opportunities for ePrescribing-based intervention. Work package1 phase 2b Some text in this section has been reproduced from Cresswell K, Hinder S, Sheikh A, Pontefract S, Watson NW, Price D, et al. ePrescribing-based antimicrobial stewardship practices in an English National Health Service Hospital: qualitative interview study among medical prescribers and pharmacists. JMIR Form Res 2023;7:e37863. https://doi.org/10.2196/37863 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text. Two and 7 weeks after intervention initiation, we interviewed 25 users of the intervention, including senior and junior prescribers, a senior nurse, a pharmacist and a microbiologist. Topics discussed included perceived impacts of different elements of the intervention, and facilitators and barriers to effective use. Interviews were supplemented by two observations of ward rounds to gain insights into AMS practices. Work package 2a A scoping review of the literature was conducted to identify AMS interventions and outcomes measures for the hospital setting. The review was published in the Journal of Antimicrobial Chemotherapy – AMR (AMS using electronic prescribing systems in hospital settings: scoping review of interventions and outcome measures). This identified 33 indicators that have been used to measure AMS in hospitals, which, together with an additional 3 indicators from other scoping work, was used to inform an eDelphi to gain consensus on 36 potential study outcome measures: 20 quality indicators (relating to the appropriateness of antibiotic use) 11 quantity indicators (relating to the cost or volume of antibiotic use) 5 indicators relating to clinical decision support. Additional measures from expert review at eDelphi identified an additional 43 outcome measures to a total of 79 outcome measures that were taken forward to assess for appropriateness, reliability and feasibility. Work package 2b We undertook a before-and-after feasibility trial (non-randomised), introducing ePAMS+ in two English hospitals which were established users of the Cerner Millennium ePrescribing and Medicines Administration (EPMA) system. Several specialties were studied among the wards participating in the trial. Patient participants, but not prescribers, were blinded to whether ePAMS+ was in use. By applying a mixed-methods evaluation, we aimed to establish: the usability and acceptability of ePAMS+ and trial procedures; the feasibility of implementing ePAMS+ and recording quantitative outcomes; and, to measure the degree to which ePAMS+ was delivered as planned, a Fidelity Index. We performed a series of qualitative semistructured interviews of doctors, nurses and pharmacists, as well as non-participant observations; we derived quantitative summaries of prescribing data from the EPMA system. Through normal linear modelling of the defined daily dose (DDD) of antibiotic per admission, we informed the sample size calculation for a future large-scale evaluation of ePAMS+. Work package 3 We intended to evaluate ePAMS+ using a hybrid stepped-wedge cluster-randomised trial (with internal pilot study) to determine its impact on the coprimary outcomes of antibiotic usage (total DDD per admission) and mortality 30 days following admission, alongside a series of secondary outcomes and prescribing process measures. Work package 4 A scoping review of the literature on cost-effectiveness studies of Antimicrobial Stewardship Programmes (ASPs) was undertaken. In line with the original intention of the programme, we considered ASPs that affect the duration, route and type of antibiotic administered during follow-up after the initial antibiotic prescription. We selected studies that measured both cost (intervention cost, antibiotic cost or hospital costs) and clinical effectiveness. Only studies that included a valued (i.e. clinical) outcome [death, length of stay (LOS), Clostridioides difficile infection (CDI)] were included. We did not include studies that only mentioned process outcomes such as DDD and total antibiotic use. We also excluded antifungal stewardship programmes and studies using diagnostic tests to improve antibiotic utilisation (e.g. procalcitonin tests and rapid diagnosis tests). Following extensive consultation with experts, we identified three causal routes through which an ASP might affect patient health and health service costs: Reduction in opportunistic infection by a micro-organism agent that is normally held in control by other micro-organisms that are inactivated by antibiotic agents. We focused on Clostridioides difficile as it is, by a considerable margin, the most important of these opportunistic infections. Changing from intravenous (i.v.) to oral therapy, which has numerous advantages in terms of cost (including staff time) and patient comfort/mobility. Switching to a more appropriate antibiotic in the case of bug-antibiotic mismatch (BAM). In populating the model, we utilised 22 parameter estimates. Some of these estimates were taken from the literature. For many parameters, there was no estimate available from the literature (e.g. how often is a BAM mis-diagnosed), and for others, the link to the literature was tenuous (e.g. effect of DDD of antibiotic on risk of CDI). Some data (e.g. the proportion of patients on antibiotics where a sample was submitted for microbiological analysis) were estimated from hospital summary records. The model was subjected to deterministic and probabilistic sensitivity analysis. Work package 5 We carried out semistructured qualitative interviews with 38 AMS leads, pharmacists and microbiologists in 23 hospitals. We explored the participant experiences with AMS electronic prescribing systems and other activities relating to AMS. We sought to provide insights into the AMS functionality of EPMA systems and other related AMS activities, and make recommendations with regard to the interventions most valued by participants. We also sought to understand future strategic priorities and challenges and possible ways to tackle these. We analysed the data thematically using technology, people, organisations and macroenvironmental as a coding framework. Qualitative analysis was iterative, allowing emerging themes to be explored further and disconfirming evidence to be sought. Thematic analysis allowed exploration of various perspectives and contexts. Results We summarise our findings below by WP. Work package 1 phase 2a Antimicrobial prescribing and review processes were characterised by competing priorities and uncertainty of prescribers and reviewers around prescribing decisions. For example, medical prescribers often had to face trade-offs between individual patient benefit and more diffuse population health benefits, and the rationale for prescribing decisions was not always clear. Prescribing involved a complex set of activities carried out by various healthcare practitioners who each only had a partial and temporary view of the whole process, and whose relationships were characterised by deeply engrained hierarchies that shaped interactions and varied across specialties. For example, newly qualified doctors and pharmacists were hesitant to change a consultant’s prescribing decision when reviewing prescriptions. Multidisciplinary communication, collaboration and co-ordination promoted good AMS practices by reducing uncertainty. Work package 1 phase 2b Tracing the adoption and impact of the various components of the intervention was difficult as it had been introduced into a setting with competing pressures. These particularly affected behavioural/educational components (e.g. training, awareness building activities) which were often delivered ad hoc. We found that the participatory intervention design had addressed typical use cases but had not catered for edge cases that only became visible when the intervention was delivered in real-world settings (e.g. variations in prescribing workflows across different specialties and conditions). Work package 2a An online learning tool was developed to develop knowledge relating to AMS and to communicate information about the EPAMS. Online AMS and ePAMS+ training (approximately 30 minutes of learning): Hosted online using an NHS-approved URL: https://epams.helmlms.com/login. The online module had the following learning outcomes: List the factors to consider when initiating antibiotics, switching the route of administration, and stopping treatment. Discuss the risks of staying on antibiotics for longer than is clinically needed. List the tools provided by the ePAMS+ intervention in this hospital. Explain how order plans work and how these have been set up to aid your decision-making. Explain the components of the antibiotic reduction and conservation decision aid and how these can encourage good stewardship within ePAMS+. A pre/post-test was developed so learners can assess baseline knowledge and knowledge acquisition. The module comprised videos of the ePAMS+ tool to explain to learners how the tool is accessed and utilised in practice. Work package 2b It was feasible to gather 20 of the 43 planned outcome and process measures using data extracted from the Cerner EPMA system. Overall, where data extraction was possible, data completeness was high. We successfully derived total antibiotic DDD per admission, a key outcome measure for future research on ePAMS+, and analysed its variability to inform the sample size calculation for future studies. The small number of antibiotic review records available confirmed the limited use of ePAMS+ within wards during the feasibility trial. Future data extraction efforts should focus on obtaining antibiotic indication data, to allow indication-based outcomes to be measured (e.g. the number of antibiotic courses for the same indication). In developing the Fidelity Index, we identified the critical decision-making points for prescribers that are linked to adherence to the ePAMS+ intervention, enabling proxy measures for these to be developed using the electronic patient record data. The limited use of ePAMS+ order sets during the feasibility trial meant that we could not, as originally planned, develop a scoring system to quantify each decision point and specify automation of the scoring within the Cerner EPMA system. Work package 3 Due to the challenges in recruiting sufficient study sites because of the changing landscape of AMS across NHS Cerner sites, as well as the further ePAMS+ development requirements identified during the feasibility trial, the planned pilot and hybrid stepped-wedge cluster-randomised trial was not undertaken. Work package 4 Our literature review on economic evaluations of ASPs found that the intervention types evaluated research methods, outcomes assessed and contexts all varied widely across studies. Given this heterogeneity, any attempt to estimate a ‘typical’ cost-effectiveness parameter would be a quixotic undertaking. It was this analysis that led us to a completely different approach – namely, the generation of a causal model to help derive the essential effectiveness and cost parameters that might drive the decision-analytic (cost–utility) model. The base-case results from our model suggests that, even at the low absolute health gains estimated and despite uncertainty as to intervention costs, the intervention is highly likely to dominate by yielding a net saving of money and a health benefit – albeit a small one. The sensitivity analyses found that the cost savings predominately turn on the impact of i.v. to oral switch on LOS, so this should be a parameter that is closely monitored. Work package 5 We found there was considerable variety across the 23 English hospitals studied in terms of their adoptions of functionality and other interventions to support AMS and a lack of evidence base to support these approaches. Secondary uses of data allowing AMS pharmacists and microbiologists to view prescribing data remotely (e.g. which patients are on antibiotics) were viewed as important in improving AMS. An indication of the source of infection was viewed as an important part of prescribing antibiotics to help other clinicians in the patient’s journey. The views of interviewees varied on the usefulness of order protocols (prescribing on the basis of infection, e.g. community acquired pneumonia). For those with ePrescribing systems without in-built protocols, designing them took some time. Conclusions We developed the ePAMS+ intervention and conducted a feasibility trial of this in hospital medical inpatient settings. This feasibility work was, however, undertaken during challenging circumstances resulting from a combination of the COVID-19 pandemic and NHS strikes, making it difficult for clinicians to find time to engage with the intervention. Challenges with the feasibility trial and delays to the research associated with the pandemic resulted in the wider AMS landscape having changed such that it would not have been prudent to proceed with the planned follow-on pilot and definitive trials. Our qualitative work mapping the NHS landscape has identified a number of promising areas for further development and evaluation to support AMS in NHS hospitals. In summary, the ePAMS+ intervention has been pre-empted by the adoption of similar measures, often resulting from local innovation which in consequence may lack an evidence base. The landscape into which complex interventions supporting AMS may be implemented is extremely heterogeneous in terms of both technological functionality and organisational practices. Interventions need to be overlaid onto this installed base. More systematic research is needed to better understand the context for further improvement. Trial registration This trial is registered as ISRCTN13429325. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0617-20009) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 12. See the NIHR Funding and Awards website for further award information.

Background Delirium is a common neurocognitive disorder in older adults and is associated with poor outcomes. Cognitive, mood and functional deficits can persist for months following an episode, impacting long-term well-being. While previous research has focused on delirium prevention and guidelines exist for short-term management, there remains a critical gap in understanding and supporting longer-term recovery. Objectives Objectives were to: (1) develop an intervention to improve recovery after delirium; (2) conduct a feasibility study of the intervention in people who have had delirium; and (3) conduct a definitive randomised controlled trial and cost-effectiveness analysis of the intervention in people who have had delirium compared with usual care, with parallel process evaluation and an embedded implementation study. Design The study consisted of four work packages. In work package 1, we designed a complex, theory-based rehabilitation intervention informed by a realist review, qualitative research and expert panel input. Work package 2, using a single-arm study, assessed the feasibility of the intervention and of undertaking a definitive randomised controlled trial and cost-effectiveness analysis with an embedded process evaluation, while the aim of work package 3 was to test its effectiveness in a definitive randomised controlled trial, and the aim of work package 4 was to focus on implementation in real-world settings. Setting Acute admissions wards and community follow-up after discharge at six National Health Service hospital sites. Participants Patients aged ≥ 65 years experiencing delirium during acute hospital admission. Intervention Home-based rehabilitation programme designed to support recovery after hospital discharge, addressing cognitive, physical, physiological and psychosocial needs. Delivered by a trained team of occupational therapists, physiotherapists and rehabilitation support workers, the intervention included a comprehensive home assessment, collaborative goal setting, up to 10 personalised therapy sessions over 12 weeks, the use of a recovery record to guide progress, education and psychosocial support. Main outcome measures The main objective of work package 1 was the development of a theory-based and person-centred rehabilitation intervention for individuals recovering from delirium. It included structured components targeting physical, cognitive and emotional recovery, along with an intervention manual and training materials for healthcare professionals. In work package 2, we assessed feasibility, including eligibility, recruitment, data collection, attrition, acceptability of the rehabilitation intervention and the cost-effectiveness framework for a subsequent definitive trial. Patient and carer participants and professional delivering the intervention were interviewed in a process evaluation. Results The intervention was developed using insights from a realist review, stakeholder interviews and expert panel discussions, identifying key recovery domains: physical, cognitive and emotional. Key facilitators included carer involvement, tailoring to individual needs, fostering interpersonal connections and promoting positive self-expression. The feasibility study identified 435 patients with delirium across 6 hospitals, of whom 36 (8%) met eligibility criteria, with 19 (53%) of these consenting to participate. Among those enrolled, 13 (68%) started the intervention, and 10 (53%) completed the final follow-up. Participants had a mean (standard deviation) baseline disability assessment for dementia score of 43.7 (27.1), reflecting the functional status of the recruited population. The mean cost of delivering the intervention was £1249 per participant. The process evaluation highlighted the intervention’s flexibility, acceptability and strong continuity of care, although challenges in recruitment and retention were noted. Limitations In work package 2, the study was limited by a lack of ethnic diversity, reliance on clinical teams for delirium diagnosis and potential selection bias due to staff capacity constraints, which may have impacted recruitment and generalisability. Recruitment was insufficient to progress to the definitive randomised controlled trial, preventing further evaluation and the final implementation study. Following a funding review, the programme was discontinued after work package 2. Patient and public involvement and engagement A patient and public involvement and engagement group of people who had experienced delirium and carers was involved throughout the programme. They assisted with participant materials, design of the intervention, interpretation of the results and recruitment strategies. Conclusions A multidisciplinary rehabilitation intervention for delirium recovery was successfully developed and implemented, with high engagement from participants who remained in the study. However, challenges in recruitment and retention to an evaluation study must be addressed. Future work Future research should focus on understanding hospital processes, the natural history of delirium recovery, community pathways and available services. Refining inclusion criteria will be essential to ensure the feasibility of conducting a randomised controlled trial or other evaluation of the intervention. Study registration This study is registered as ISRCTN15676570, registered 13 December 2022. https://doi.org/10.1186/ISRCTN15676570 Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: NIHR202338) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 7. See the NIHR Funding and Awards website for further award information. Plain language summary Delirium is a serious condition that often affects older people, especially those with dementia. It causes confusion, difficulty focusing and problems with memory and thinking. People with delirium in hospital often need more help on returning home due to problems with mobility and thinking. While most research has focused on preventing delirium, not enough attention has been given to helping people recover from it once back at home. The RecoverED programme was designed to help older adults recover from delirium by supporting their physical, mental and emotional well-being. The programme included exercises to help with physical recovery, activities to improve memory and thinking, and support to help people feel emotionally better. The first part of the study focused on creating the best possible recovery plan by talking to experts, patients and carers. In the second part, the programme was tested with older adults who had delirium in hospital, once they were back at home. We found 435 patients with delirium across 6 hospitals. Thirty-six (8%) were suitable for the trial and 19 took part. Ten people (53%) completed the final follow-up. The results showed that people found the programme to be helpful and felt better after taking part. However, if we had been able to recruit more people, we may have found more changes that needed to be made. The study also faced difficulty in finding enough participants, community rehabilitation staff to deliver it and making sure the programme reached people from all backgrounds. We planned to go on to a full randomised controlled trial of the programme and a study of how to make it work in the National Health Service. However, we were unable to do this because of the difficulties in finding participants in the second part of the study. The study showed that we need to improve how we recruit people, make it clearer who the programme is best suited to and find better ways to support those with complex health needs. Future studies will need to address the challenges we found in recruitment. They should focus on involving people from different backgrounds and finding ways to make the programme available to more people who need it. Scientific summary Background Delirium is a common and serious neurocognitive disorder affecting older adults, particularly those with pre-existing dementia. It is characterised by acute disturbances in attention, awareness and cognition, and it is associated with prolonged hospital stays, functional decline and increased mortality. Beyond the immediate episode, delirium has lasting consequences, with many individuals experiencing persistent cognitive and functional impairments and an increased likelihood of institutionalisation. The condition also places significant emotional and practical burdens on carers while contributing to healthcare costs. Despite its widespread impact, research on delirium has largely focused on prevention and short-term management, leaving a critical gap in understanding and supporting long-term recovery. Emerging evidence suggests that rehabilitation could aid recovery following delirium, yet older adults with cognitive impairment are often incorrectly perceived as having limited rehabilitation potential. Current guidelines highlight the need for effective, non-pharmacological interventions to support long-term recovery after an episode of delirium, particularly for individuals with delirium superimposed on dementia. However, there remains a lack of structured rehabilitation strategies tailored to this population. Addressing this gap requires a deeper understanding of the factors influencing recovery and the development of targeted interventions that promote cognitive and functional improvements, enhance patient and caregiver well-being and reduce healthcare burdens. Aims, objectives and summary of approach The aim of the RecoverED research programme was to develop an intervention to improve recovery following delirium and to evaluate its effectiveness, cost-effectiveness and feasibility for implementation in clinical practice. The intervention targeted the cognitive, functional and emotional aspects of recovery, providing a comprehensive approach to supporting individuals who have experienced delirium. Work package 1: intervention design Work package (WP) 1 aimed to develop a comprehensive programme theory to understand what improves recovery after delirium, for whom and in what context. This programme theory provided the foundation for co-designing a complex rehabilitation intervention for older adults (aged ≥ 65 years) recovering from delirium following acute hospital admission. Work package 2: feasibility and acceptability Work package 2 was a single-arm feasibility single-arm study aimed at assessing the feasibility and acceptability of a home-based rehabilitation intervention designed for older adults who have experienced delirium during acute hospitalisation. The primary objective was to evaluate trial feasibility, whether the intervention was acceptable to both participants and their carers. Secondary objectives included examining the intervention’s acceptability among diverse populations, testing the feasibility of collecting outcome data for a definitive randomised controlled trial (RCT), testing the framework for a future cost-effectiveness analysis (CEA) alongside a definitive trial and engaging in iterative refinements of the intervention. Stop-go criteria determined whether to go on to a full trial. The programme was coproduced with our patient and public involvement and engagement (PPIE) group. Work package 3 was planned to be a definitive RCT of the intervention in 12 sites across the UK. WP4 was planned to be an implementation study of the intervention in two sites, with clinical teams identifying participants rather than research teams. However, following a funding review, the programme was discontinued after WP2 due to significant challenges in identifying and recruiting participants. Despite efforts to refine recruitment strategies, the feasibility study faced persistent difficulties, which limited the number of participants who could take part. These challenges raised concerns about the viability of progressing to a full-scale RCT. As a result, plans for a definitive RCT, CEA and implementation study were halted. Methods and results Work package 1: methods Work package 1 of the RecoverED programme followed Medical Research Council guidelines for complex interventions, using a realist approach to develop a programme theory on improving delirium recovery. This involved five key stages: (1) a rapid realist review to synthesise mechanisms for improving recovery from delirium; (2) qualitative realist interviews with 8 older people who had experienced delirium, 14 carers and 24 healthcare professionals (HCPs) to understand rehabilitation needs; (3) an expert panel workshop to refine the theory and review findings; (4) a series of programme theory development meetings with the core team to revise and produce a logic model and (5) an evaluation of the intervention’s implementation, assessing fidelity and acceptability. A multidisciplinary approach was used in the intervention design, integrating expertise from physiotherapy, occupational therapy, psychology, geriatric medicine and other specialists, informed by the International Classification of Functioning, Disability and Health framework. Continued engagement with our PPIE group ensured that the intervention remained practical, patient-centred and culturally relevant. Work package 1: results The realist review identified three recovery domains: physical recovery through exercise, cognitive recovery with reality orientation and stimulation and emotional recovery through conversations with skilled professionals. Four key facilitators were identified: carer involvement, tailoring the intervention to individual needs, fostering interpersonal connections and promoting positive self-expression. Stakeholder interviews further emphasised the importance of rehabilitation, emotional support, delirium education and addressing health conditions. Expert panel discussions refined the programme theory, emphasising a person-centred approach, integration of rehabilitation into daily life, carer engagement and continuity of relationships with professional carers. This theory guided the development of the RecoverED intervention, ensuring that it addressed the complex needs of delirium recovery. Work package 2: methods Work package 2 employed a multicentre, single-arm feasibility study design with an embedded process evaluation. Participants were recruited from six NHS hospitals in the UK, focusing on individuals aged ≥ 65 years who had been diagnosed with delirium and were expected to return home. The RecoverED intervention, aimed at supporting recovery, began within 2 weeks of discharge and consisted of a structured rehabilitation programme, including assessments by physiotherapists (PTs) and occupational therapists (OTs) and up to 10 sessions delivered by a rehabilitation support worker (RSW) over 12 weeks. Feasibility outcomes were assessed through recruitment and retention rates, with mixed methods used to evaluate implementation fidelity and acceptability. The resources required to deliver the RecoverED intervention and their associated costs, health-related quality of life and well-being outcomes measured using instruments suitable for generating quality-adjusted life-years (QALYs) and well-being-adjusted life-years (WALYs), and participants’ use of health, social care and wider societal resources and their associated costs were also assessed to inform a potential future CEA. Work package 2: results Out of 435 patients identified with delirium across 6 hospitals, 36 (8.2%) met the eligibility criteria, and 19 (53%) of patient–carer pairs consented to participate. Of these, 13 participants (68%) started the intervention, and 10 (53%) completed the final follow-up. The mean participant age was 83.8 years, with 60% being men and 26% having dementia as a comorbidity. Participants had a mean [standard deviation (SD)] baseline disability assessment for dementia scores of 43.7 (SD: 27.1), reflecting the functional status of the recruited population. Intervention adherence was 53%, with 10 participants completed at least 6 or more intervention sessions. At baseline, 79% of participants showed delirium markers, but none did at follow-up. The mean cost per participant for delivering the RecoverED intervention was £1249, although challenges arose in gathering data for non-contact activities. Mean (SD) QALY gains for patients and carers, based on self-reported EuroQol-5 Dimensions, five-level version responses over the 6-month follow-up period, were 0.273 (0.168) and 0.403 (0.092), respectively. Mean (SD) WALY gains were 0.368 (0.075) for patients (based on ICEpop CAPability measure for older people) and 0.427 (0.056) for carers (ICEpop CAPability measure for adults), indicating positive effects on health-related quality of life and well-being in both groups. Analysis of total participant resource use and costs was limited due to missing data. Challenges in recruitment and retention were identified, with health-related withdrawals and inconsistent delirium screening contributing to lower recruitment than anticipated. The process evaluation showed that the intervention was well received and delivered with good fidelity. It was flexible and tailored to individual needs, with strong continuity of care provided by the same HCP team. While some mobility tasks were underdelivered due to health issues or poor weather, the intervention was largely seen as acceptable and beneficial. Discussion The RecoverED programme has several notable strengths, including its comprehensive, theory-driven design developed through an iterative co-design process. This process integrated expertise from geriatricians, PTs, OTs and clinical psychologists, ensuring the intervention’s clinical relevance and theoretical foundation. The involvement of a PPIE group ensured that the intervention was practical and patient-centred. Moreover, the development of a structured manual and training programme for RSWs was a significant achievement, enhancing their preparedness and ability to implement the intervention effectively. The intervention was well received by participants, with positive feedback regarding its perceived value and benefits. High consent rates among eligible participants and strong engagement from those who remained in the study further underscore the potential of the intervention. Despite these strengths, the study encountered several limitations. Recruitment was a significant challenge, with only a small percentage of screened patients meeting the eligibility criteria. This was partly due to staff capacity constraints, inconsistent routine delirium screening practices and the predominance of clinical teams diagnosing delirium rather than research staff. Since clinical teams were responsible for diagnosing delirium as part of routine care, there was variability in how and when delirium was identified, leading to potential inconsistencies in patient eligibility for the study. By contrast, if research staff had been directly involved in the screening and diagnosis process using standardised criteria, it may have improved consistency and ensured a more reliable identification of eligible participants. Additionally, missing data on intervention costs and participant resource use limited the ability to conduct a full economic evaluation. Future studies should implement more robust data collection strategies to ensure a more comprehensive assessment of cost-effectiveness. The process evaluation was limited due to the small number of participants and the withdrawal of more impaired participants from the trial. A more extensive evaluation may have revealed more weaknesses in the intervention and a need for further refinement. Another limitation was the homogeneity of the study sample, as all participants were White, which limits the understanding of how the intervention may be received by different ethnic and cultural groups. Furthermore, the reliance on clinical teams for diagnosis rather than standardised research staff-based approaches revealed weaknesses in the screening process, leading to missed recruitment opportunities. Finally, the absence of participants from minority ethnic, racial or linguistic groups and the challenges in retaining participants due to health deterioration or relocation suggest the need for improved inclusion strategies and a more inclusive approach to recruitment in future studies. Conclusion The RecoverED programme successfully developed a theory-driven, multidisciplinary intervention aimed at supporting recovery following delirium. Despite recruitment challenges and limitations in participant diversity, the qualitative data from the feasibility study demonstrated some evidence that the intervention could be delivered in clinical practice, was well received and participants reported perceived benefits. The process evaluation provided some insights into how the intervention was implemented and how it could be refined for future studies, but more participants were needed for a comprehensive evaluation. The findings from this study underline the importance of addressing key challenges in recruiting and retaining a diverse sample, including individuals from minority ethnic groups and those transitioning to care homes. Further research is needed to explore how the intervention can be adapted and optimised to support these groups. Given the pressing need for effective delirium recovery interventions, future studies should consider novel trial designs and alternative funding models to enable the continued development and evaluation of this important area of health care. Study registration This study is registered as ISRCTN15676570, registered 13 December 2022. https://doi.org/10.1186/ISRCTN15676570 Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: NIHR202338) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 7. See the NIHR Funding and Awards website for further award information.

Background Globally, the health of young people is in decline, with increasing rates of mental health issues, obesity and sedentary lifestyles. Existing health behaviour change interventions for this group face challenges in engagement and implementation. Effective, targeted interventions are needed to improve lifelong health outcomes for young people. Objective To develop, implement and evaluate a multicomponent health behaviour change intervention aimed at improving diet and increasing physical activity in young people in the United Kingdom. Design and methods A person-based approach drawing on systematic and literature reviews and qualitative data (interviews, group discussions and participatory ‘game jams’) was used to develop the Engaging Adolescents in Changing Behaviour intervention, which was tested using a cluster-randomised trial. A pilot trial informed modifications to recruitment, data collection and app usage processes. Schools selected two Year 8 classes (12–13 years old) to participate, with randomisation at the school level. Baseline and 12-month follow-up data on diet, physical activity and general well-being were collected from students, and demographic data were provided by schools with parental permission. An economic model was designed to estimate future effects on health outcomes of the intervention, and mixed-methods process evaluation examined participant perspectives (students, parents and teachers), LifeLab® attendance and app download statistics. Setting and participants Of our planned sample of 2300, we collected baseline data from 2065 students from 45 of 49 recruited and randomised secondary schools across Hampshire and the south coast of England. Four schools dropped out before baseline data collection, and we followed up 55% of students at 1 year. The process evaluation involved 88 of these students, 23 teachers and 11 parents. Intervention The Engaging Adolescents in Changing Behaviour intervention had three components: experiential science education for students at LifeLab, teacher training to support students’ health behaviours and a gamified smartphone app for students to encourage better diet and physical activity. Main outcome measures Diet quality (standard deviation scores), total physical activity (minutes/day). Secondary outcomes included further diet and physical activity outcomes, self-efficacy for health behaviours and general well-being, assessed using both objective and self-report methods. Data were analysed using linear mixed-effects models and binary regression, accounting for clustering. Results Development work indicated that effective interventions should support young people’s autonomy, competence and relatedness while making healthier choices accessible and appealing. Digital platforms and social media were identified as key for engagement. Pilot findings prompted adjustments to school recruitment, data collection and app presentation for the main trial. At baseline, the intervention group had a mean diet quality score that was 0.22 standard deviation (95% confidence interval −0.05 to 0.51) higher than in the control group. At follow-up, the difference between groups was −0.02 standard deviation (95% confidence interval −0.32 to 0.27), and after adjustment for the baseline, diet quality score was −0.10 standard deviation (95% confidence interval −0.29 to 0.09). For the total physical activity outcome, the baseline difference between groups was minimal; at follow-up, the difference was 1.19 minutes/day (95% confidence interval −15.5 to 17.9) and after adjustment for baseline was 8.65 minutes/day (95% confidence interval −7.79 to 25.1). Minor differences were observed in secondary outcomes. Index of Multiple Deprivation scores of schools in intervention and control groups showed no baseline differences. Overall, 88% of eligible students attended LifeLab, and 42.5% downloaded the app. Students and teachers reported positive experiences with LifeLab and teacher training, though COVID-19 impacted in-person delivery for some schools. Students valued learning about their health but engaged minimally with the app. Parents viewed the intervention positively but often lacked communication from their children or schools about it. Teachers were eager to support health initiatives but had limited opportunities during the school day. A Markov model was produced to estimate the cost-effectiveness of LifeLab Plus, but our findings did not warrant such an analysis. Limitations Most students did not engage with the app due to its lack of appeal and insufficient motivation to revisit it. Only 55% of students participated in the follow-up and provided data on the diet quality score, and only 28% had complete physical activity data at baseline and follow-up. Thus, the power of the trial was reduced, making it difficult to detect differences between groups. Trials in school settings tend to include more advantaged children, as opt-in consent favours parents who complete forms. Even opt-out consent does not fully address this issue, as disadvantaged students are more likely to miss school during data collection. Demographic data collection required parental opt-in, introducing further bias. Conclusions The trial detected no significant impact of the intervention on diet and physical activity among secondary school students. Substantial baseline differences in diet and physical activity between intervention and control groups introduced noise, complicating conclusions about intervention effects. While schools are suitable settings for health interventions, randomised trials may not be ideal for evaluating large-scale, real-world interventions. Intervention design, particularly for smartphone applications, requires greater attention to foster sustained engagement. Future work Although the Engaging Adolescents in Changing Behaviour intervention did not improve diet or physical activity, it provided valuable insights. Interventions must support young people’s autonomy, competence and relatedness and offer them meaningful opportunities to support their aspirations in life. Health topics have to be important to young people to be engaging, and research has to be coproduced. Our current and future research focuses on evaluating the impact of this participatory approach on health and well-being and aims to translate findings into policy through collaborations with local authorities and Integrated Care Systems in Wessex. Exploring alternative designs for evaluating large-scale public health interventions is another important research agenda. Trial registration This trial is registered as ISRCTN74109264. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0216-20004) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 8. See the NIHR Funding and Awards website for further award information. Plain language summary Young people in the United Kingdom have the worst diets of any age group and insufficient exercise, increasing risks of conditions such as heart disease and diabetes later in life. Since health habits formed in adolescence can last a lifetime, researchers are trying to find ways to help young people adopt healthier behaviours. Interactive programmes, called interventions, can improve health habits but are difficult to design and implement with young people. The Engaging Adolescents in Changing Behaviour programme developed and tested a school-based intervention for healthier choices among students. A method of assessing the money that could be saved on health care if this intervention was successful was developed. To inform the programme, researchers consulted young people, parents and teachers, finding that: young people valued health but felt it had to fit into their social lives and activities the environment around young people made it hard for parents and teachers to support young people’s healthy habits effective interventions must respect adolescents’ need for independence and a social life while making healthy choices enjoyable and accessible. The intervention involved 2065 students (aged 12–13 years at baseline and 13–14 years at follow-up) from 45 schools in Hampshire and Southern England. Forty-nine schools were recruited and randomly allocated to intervention or control, but four schools (two intervention and two control) subsequently dropped out. Half of the schools implemented the programme, while the others continued regular activities as a comparison group. The Engaging Adolescents in Changing Behaviour intervention included: teacher training for supportive health discussions LifeLab (in-school lessons and a trip to a health education facility for an experiential science day) the Engaging Adolescents in Changing Behaviour gamified smartphone app to promote eating better and exercising more. Students completed diet and activity questionnaires and wore activity trackers for a week. Data were collected again a year later, alongside feedback from participants, parents and teachers. Findings revealed no meaningful improvements in diet quality or physical activity between intervention and comparison groups, though small differences in well-being and food choices were noted. We may not have found any effect of the intervention because i) there were differences in initial diet scores between groups, ii) only being able to see 55% of the original group of students a year after the intervention and iii) feedback that the app was not engaging enough to compete with social media for young people’s attention. Because we did not see any impact of the intervention, we were unable to carry out a cost-effectiveness evaluation. Key recommendations: continue partnering with schools despite logistical challenges ensure digital interventions work alongside other digital activities like social media rather than compete. Scientific summary The health of young people across the world is in decline, and most existing health behaviour change interventions have failed to shift young people onto more positive health trajectories. The National Institute for Health and Care Research (NIHR) funded the Engaging Adolescents in Changing Behaviour (EACH-B) programme to examine whether a multicomponent intervention could improve health behaviour outcomes for young people on the south coast of England. We developed the intervention, evaluated its effectiveness using a cluster-randomised controlled trial and conducted a process evaluation of the delivery, implementation and mechanisms of impact of the EACH-B intervention. Programme research stream and work package aims and objectives Research stream (RS) 1 [work package (WP) 1.1, WP1.2, WP1.3]: to find the best methods for helping young people make healthier choices by looking at previous research and talking to experts. To build a health economic model to estimate the costs and benefits of existing interventions. Research stream 2 (WP2.1, WP2.2, WP2.3, WP2.4): to develop an intervention that young people, parents and teachers found helpful in improving diet and increasing exercise. Research stream 3 (WP3.1, WP3.2): to test the intervention to see how well it works, whether it offered good value for money and the best way of implementing it in schools. Methods Research stream 1 Work package 1.1: three systematic reviews were conducted to inform the economic modelling (RS1) and intervention development (RS2). The first synthesised evidence from quantitative studies on the effectiveness of digital interventions to improve diet quality (www.sciencedirect.com/topics/medicine-and-dentistry/diet-quality) and increase physical activity (PA) (www.sciencedirect.com/topics/psychology/physical-activity) in adolescents, to identify effective intervention components and to assess the cost-effectiveness of these interventions. The second focused on the literature investigating the effectiveness of health education interventions delivered in school settings to prevent overweight and obesity and/or reduce body mass index (BMI) in adolescents, and to explore the key features of effectiveness. The third systematic review examined the influence of community and consumer nutrition environments on young people’s food purchasing and dietary behaviours in high-income countries was conducted. The protocol paper described the design of the EACH-B trial to test the intervention. Work package 1.2: a Markov model, considering mental health, earnings, type 2 diabetes (T2D) and pregnancy outcomes over 20 years, was developed using published literature and sensitivity analyses to assess the cost-effectiveness of a school-based intervention for improving diet and PA in young people. The model was designed to be used with data produced in WP3.2 to test cost-effectiveness. Work package 1.3: stakeholders were identified and regular meetings set up throughout the duration of the programme. Opportunities to contribute to consultations, briefings and policy documents were sought. A stakeholder consortium was formed in the last 3 years of the programme. Research stream 2 Work package 2.1: a review examined the reasons for the ineffectiveness of existing behaviour change interventions. Findings informed questions asked in group interviews with young people aged 13–14 years. Semistructured interviews with parents of adolescents were also conducted, exploring parental involvement in adolescent health interventions. Another study analysed quantitative adolescent dietary data and qualitative interviews with parents, teachers and adolescents to examine energy drink consumption and associated factors. A further study conducted focus groups with adolescents aged 11–18 years to explore the interplay between social and food environments in adolescent food choices. Work package 2.2: the content of the intervention was designed using findings from RS1 and RS2. Components of the intervention were developed using data collected in WP2.1 and drawing on evidence review from RS1. Work package 2.3: the smartphone application, referred to as ‘the app’ or ‘LifeLab Plus’, was developed by game designers at Glasgow Caledonian University and tested by young people. Think-aloud interviews were conducted with young people testing prototypes of the app to capture immediate reactions to aspects of the intervention. Work package 2.4: focus groups were conducted with secondary school teachers to explore their perceptions of how they could use Healthy Conversation Skills (HCS) to support students to make lifestyle changes, what additional training needs they had and how the final HCS training programme could best be delivered in the school context. Following a pilot of the HCS training, further interviews were conducted with teachers who took part (see Report Supplementary Material 6). Research stream 3 Work package 3.1: the EACH-B trial was piloted in six secondary schools in the Southampton area, randomised to intervention or control. A process evaluation was conducted to assess the implementation of the intervention, the influence of context on the trial and the mechanisms of impact underlying the intervention. The research team analysed data from the UK National Diet and Nutrition Survey to develop a short, 20-item food frequency questionnaire (FFQ) to use in the EACH-B trial. Principal component analysis was used to identify key dietary patterns and create diet quality scores. Work package 3.2: in total, 49 schools were recruited to the main trial, and 45 took part in both baseline and follow-up measures. Schools were randomised to receive the EACH-B intervention or not. Two middle-ability classes were chosen to take part by teachers in each school. Outcomes were dietary change (measured by the 20-item FFQ) and PA change [measured using GENEActiv™ (ActivInsights, Cambridge, UK) accelerometers]. Secondary outcomes were self-reported PA, behavioural regulation and self-efficacy for diet and PA, quality of life and well-being. A mixed-methods process evaluation was also conducted, following the format of that carried out with the pilot trial (WP3.1). Interviews and focus groups were conducted with students, teachers and parents from 11 schools (5 intervention and 6 control) that took part in the trial. Quantitative data were collected describing numbers of students who attended LifeLab and downloaded and used the app. Results Research stream 1 Work package 1.1: the systematic review of digital interventions found that the most effective digital interventions incorporate education, goal-setting self-monitoring and parental involvement. We found too little data on the cost-effectiveness of digital health interventions to be able to draw conclusions. The systematic review and meta-analysis of school-based adolescent health interventions suggested that school-based health education interventions have the potential to lower BMI towards a healthier range in adolescents. Multicomponent interventions involving teachers and parents and digital components are promising. The systematic review of influences of the community and consumer nutrition environment on the food purchases and dietary behaviours of young people found no clear associations between exposure to healthy community nutrition environments and dietary outcomes. However, most of the studies reviewed (57%) reported that greater exposure to unhealthy food outlets was associated with less healthy food purchases and dietary intakes. Results for consumer environments were inconsistent. Work package 1.2: a Markov model was developed to extrapolate short-term observed effects on diet, activity and/or BMI to estimate future effects on the incidence of T2D and cardiovascular disease. While the intention was to apply the model to estimate the cost-effectiveness of LifeLab Plus, our findings did not warrant such an analysis. The published model has utility beyond the present study, however, and could be applied by others undertaking such complex intervention research. Work package 1.3: stakeholders were engaged systematically throughout the programme to ensure buy-in to the intervention and insights for subsequent implementation. Significant achievements were the Innovation-to-Commercialisation fellowship for Woods-Townsend, WP1.3 lead. This involved intense stakeholder engagement to understand the implementation aspects of the programme, including conversations with a wide variety of stakeholders to explore the potential for scale-up and roll-out of EACH-B. Insights from EACH-B fed into the ukactive ‘Generation Inactive 2’ report, the Southampton City Council’s scrutiny review of their Childhood Obesity Policy and the formation of the Pathways to Health (www.pathways-to-health.org/) Consortium, a grouping of more than 30 organisations in Southampton that is committed to improving young people’s health and well-being. Research stream 2 Work package 2.1: The article summarising the findings of a review of factors involved in reducing intervention effectiveness identified three issues: a focus on mean effect sizes, assumptions about the value young people place on health, and overemphasis on cognitive mechanisms in interventions. The article suggests that interventions should consider both rational and emotional influences on health behaviours. Subsequent qualitative work (n = 54, 13- to 14-year-olds) identified that young people valued being with friends, being active and being healthy without compromising other priorities. Health alone is not a motivating factor for them. They prioritised social experiences over food quality, valued familiar food outlets and chose food promotions that aided quick decision-making. Energy drink consumption among young people (n = 2587 adolescents aged 11–18) was associated with poorer dietary quality and higher deprivation. Qualitative data from 74 adolescents, 24 parents and 15 teachers revealed that young people drink energy drinks because of peer influence and low cost and believe voluntary bans on their sales are ineffective. In additional interviews, parents (n = 24) saw themselves as gatekeepers and role models for healthy behaviour but struggled with balancing their influence and their children’s increasing autonomy. Work package 2.2: guiding principles for intervention content were that it (1) persuades adolescents that healthy eating and PA can be a positive experience with positive outcomes, (2) enhances Self-Efficacy for Healthy Eating and Physical Activity and persuades users that it can be easy and (3) supports adolescents to seek support from their social network with healthy eating and PA. The logic model developed for the funding application which described the programme and outlined the hypothetical mechanisms of the intervention was modified based on the data collected in RS2 (Figure 1). Work package 2.3: a description of the app development process is described in a paper entitled, ‘Enhancing the relevance of nutrition and physical activity interventions for adolescents: design and development of the LifeLab Plus App’ (see Report Supplementary Material 5). Findings from the qualitative research and coproduction activities informed the integration of behaviour change components into each element of the app, for example, the ‘Gutsy’ game, to maximise engagement. The paper concludes that theoretical and evidence-informed approaches, while crucial for initial development, were not sufficient to deliver an app that was meaningful and relevant to adolescents. Genuine coproduction and testing with target users is an essential step in app development. Work package 2.4: the qualitative research explored teachers’ perceptions of their roles in supporting their students to make healthy choices. Interviews with 15 secondary school teachers and focus groups with another 14 revealed that they were enthusiastic about talking to their students about health in supportive and encouraging ways and found opportunities to have these sorts of conversations. Teachers had ideas about strategies that could help them to support their students’ health but felt powerless because they witnessed the way environmental, individual, financial and social factors made it more difficult for students to make healthy choices. They felt school cultures and systems did not facilitate health behaviour change but, in fact, actively made it harder for teachers to support their students. Research stream 3 Work package 3.1: the results of the study developing the shortened version of the FFQ showed a strong correlation (0.87) between the 20-item and 139-item scores. Both scores correlated with nutritional biomarkers. The study concluded that the 20-item questionnaire could effectively assess diet quality in large-scale studies with young people. The pilot demonstrated that the trial could be conducted in schools, though some minor modifications were made before the main trial started. Just as the main trial started, schools were shut as part of the first COVID-19 lockdown period of the COVID-19 pandemic, and further modifications had to be made to accommodate that. Work package 3.2: despite using a minimisation algorithm, baseline imbalances existed in this cluster-randomised school trial, notably a 0.22 standard deviation (SD) difference in diet quality score – a primary outcome –almost as large as the 0.25 SD the study aimed to detect. This complicated analysis, as cluster-randomisation at the school level, with only 45 schools and over 2000 students, is prone to imbalance. Income Deprivation Affecting Children Index score (a measure of area household deprivation for children) also differed at baseline in control and intervention groups. The intention-to-treat analysis, including adjustment for clustering, showed minimal difference between groups in the primary outcomes at follow-up. However, follow-up rates were only 55% for the diet quality score and 28% for the PA measure. Further analyses incorporating adjustment for the baseline values of the outcome being assessed also revealed minimal group differences. The intervention group had a mean diet quality score that was 0.10 SD [95% confidence interval (CI) −0.29 to 0.09] lower than in the control group, and 8.65 minutes/day (95% CI −7.79 to 25.1) more total PA than in the control group. Both these differences are small. Secondary outcomes highlighted some effects, notably reduced risk for high crisp consumption [relative risk (RR) = 0.75 (95% CI 0.63 to 0.89)] and similar or lower fizzy drink consumption [RR = 0.77 (95% CI 0.58 to 1.03)], but baseline diet imbalances cast doubt over these findings. In the intervention group, there was a higher proportion [RR = 1.42 (95% CI 0.99 to 2.03)] of students meeting the 60-minute daily activity guideline than in the control group, yet other activity outcomes showed no differences. Compliance challenges also arose: LifeLab day attendance varied, and app engagement was low, with only 228 students using the app an average of 2.2 sessions with a duration per session of 17.7 minutes. There was no way of assessing the fidelity of the HCS component of the intervention or of the LifeLab delivery in schools. The COVID-19 pandemic was the backdrop to this research, and its effects on the findings are hard to assess. Quantitative data from the process evaluation of the main trial showed that 87.9% of eligible students attended the LifeLab facility and 42.5% of eligible students downloaded the smartphone application. Students and teachers had positive experiences of LifeLab and the teacher professional development, despite the COVID-19 pandemic preventing in-person delivery for some schools. Parents had positive views of research and were enthusiastic about the EACH-B intervention, but most said they did not know about the study from their children or school. Students valued learning about their own health and bodies, and teachers were enthusiastic about supporting their students to make healthier choices, but opportunities to talk to students about health during the school day were limited. Conclusions The EACH-B programme successfully developed and tested an intervention to improve diet and PA among 12- to 14-year-olds. Stakeholder engagement and evidence synthesis informed programme design, and the programme produced a range of outputs, including 18 peer-reviewed publications, 47 conference presentations and 4 doctoral theses. Engaging Adolescents in Changing Behaviour highlights the challenges of cluster-randomised trials in schools, particularly baseline imbalances affecting data interpretation. The study aimed to recruit 50 schools and 2300 students; 49 were recruited and randomised, and 45 provided data at both baseline and follow-up, and 2065 students were seen at baseline. Only 1127 (55%) students provided data at follow-up, however. Dropout rates were higher among disadvantaged students and exacerbated by an opt-in consent process, which limited participation. Engagement with two of the intervention components was low. Despite this, the EACH-B programme has advanced understanding of youth health interventions and influenced policy, underscoring the need for engagement-driven, adaptable and collaborative approaches. Key insights from the programme highlight that interventions must support young people’s autonomy, competence and relatedness to be effective. The programme’s impact extends beyond research, notably through the UK Research and Innovation-funded Pathways to Health Consortium, which integrates young people into local policy decision-making. Another major output is the NxtGen Researchers Training Programme, which equips young people with research and advocacy skills. This led to the Young People’s Manifesto for Change, containing 12 policy recommendations to be endorsed by the Hampshire and Isle of Wight Integrated Care Board. NxtGen Researchers continues to run and is in the process of being accredited by the Royal Society for Public Health. Trial registration This trial is registered as ISRCTN 74109264. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0216-20004) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 8. See the NIHR Funding and Awards website for further award information.

Background Children and young people with epilepsy are significantly more likely to experience multiple mental health problems, including anxiety, depression and behavioural difficulties, compared to youth without chronic physical health needs. Yet the majority of mental health problems go undetected and untreated in this population and can seriously impact social, occupational and educational functioning into adulthood. Currently, mental and physical health care is not integrated in paediatric epilepsy. Existing evidence-based psychological interventions typically target only one problem area and do not meet the specific needs of young people with epilepsy. Objectives The aim of this research programme was to transform the treatment of mental health disorders in young people with epilepsy. Our objectives were to: adapt and personalise an existing modular cognitive–behavioural treatment for multiple mental health disorders to address the particular needs of young people with epilepsy integrate the intervention into paediatric epilepsy services evaluate the clinical and cost-effectiveness of the intervention; and understand the experience of the intervention from the perspectives of patients and families. Design and methods Four complementary work packages linked to our objectives comprised: (1) multimethod improvement science work to develop the Mental Health Intervention for Children with Epilepsy treatment, (2) training and supervising staff working with young people with epilepsy to deliver the treatment with competence and fidelity,(3) a multicentre randomised controlled trial to compare the clinical and cost-effectiveness of the treatment in addition to assessment-enhanced usual care with assessment-enhanced usual care alone and (4) longitudinal qualitative study of young people and their families to explore patient experience. Setting The programme took place in National Health Service paediatric epilepsy services across England and Northern Ireland. Participants Participants were aged 3–18 years with epilepsy and at least one common mental health disorder. Many participants had multiple mental health problems, were neurodivergent and/or had intellectual disabilities. Interventions Participants in the active treatment arm of the trial received up to 20 weekly sessions of the personalised treatment, which was delivered remotely by professionals with limited prior experience in mental health. Main outcome measures Our primary outcome measure in the randomised controlled trial was the Strengths and Difficulties Questionnaire, a parent-reported standardised measure of severity of youth mental health symptoms at 6 months post randomisation. Secondary outcome measures included impact scales of mental health symptoms, health-related quality of life and parental mental health. Results Using improvement science methods, we personalised an existing modular psychological intervention. We coproduced a core module addressing epilepsy-specific issues in mental health, and three optional modules to target stigma, parental mental health and the transition to adulthood. We trained staff from paediatric epilepsy settings to deliver the intervention with competence and fidelity over 6 months. We conducted the first randomised controlled trial of a modular mental health treatment for young people with epilepsy, with excellent rates of retention and follow-up. Participants in the Mental Health Intervention for Children with Epilepsy group reported a significantly greater reduction in emotional and behavioural symptoms at 6 months post randomisation compared to the assessment-enhanced usual care control group (Cohen’s d = 0.3). These gains were maintained at 12 months’ follow-up (Cohen’s d = 0.4). Parental mental health also significantly improved in the treatment group, compared to a deterioration in the control group. The average cost of the Mental Health Intervention for Children with Epilepsy (including delivery, supervision and training) was approximately £1466 per participant, and when youth and parent quality-adjusted life-years were combined, the Mental Health Intervention for Children with Epilepsy treatment was cost-effective at the National Institute for Health and Care Excellence-preferred threshold for cost per quality-adjusted life-year (incremental cost per quality-adjusted life-year for Mental Health Intervention for Children with Epilepsy vs. control = £6311). Longitudinal qualitative interviews highlighted positive changes in feelings and behaviours of participants following therapy, independent of the child’s age, gender or level of learning needs. Limitations Limitations of our work include the lack of an active treatment control group in the randomised controlled trial, reliance on parent-reported outcome measures due to challenges in the availability and suitability of youth-report measures in this population, limitations in measurement of health-related quality of life for young people impacting on economic analyses, and a lack of power to explore the effects of therapist competence on outcome. Conclusions The multiple mental health needs of young people with epilepsy can be identified and successfully treated from within existing paediatric epilepsy services, with a personalised modular cognitive–behavioural treatment. This model of integrated health care has implications for other chronic physical health conditions. Future work Further research is needed to establish the optimal factors in providing training and supervision to professionals to promote the long-term integration of mental and physical health care in paediatric epilepsy. Trial registration The trial is registered as ISRCTN57823197. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0616-20007) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 9. See the NIHR Funding and Awards website for further award information. Plain language summary Our research aimed to improve the treatment of mental health problems in young people with epilepsy. At least half of young people with epilepsy have mental health problems like depression, anxiety and behavioural problems, and many have more than one difficulty. Mental health problems impact on all areas of life, including family, friendships and education. However, epilepsy services are separate from mental health services, and few young people with epilepsy access mental health services when they need them. We adapted a psychological treatment for anxiety, depression and behavioural problems so that it met the unique needs of young people with epilepsy. The treatment was flexible, delivered by telephone and could address more than one mental health problem. We trained staff who did not have special mental health training to deliver this treatment within epilepsy services. We compared this treatment to the usual care that young people with epilepsy receive. We talked to young people and families throughout the research to understand how their mental health changed during the study, what the treatment was like and how to improve it. We found that young people who received the treatment had fewer mental health symptoms after 6 and 12 months, compared to young people who received usual care. The mental health of caregivers of young people who received the treatment also improved. Young people and families were positive about the treatment and found it helpful. Staff in epilepsy services also liked the treatment and were able to use it appropriately when they had training and support. The treatment may also be good value for money, particularly when we looked at outcomes for both young people and their parents. Overall, this research showed that the treatment can be used within epilepsy services to treat mental health difficulties that young people with epilepsy experience. Scientific summary Background Epilepsy is the most common significant neurological disorder affecting children and young people. There is a high incidence of mental health conditions, such as anxiety, depression and behavioural difficulties, among children and young people with epilepsy compared to those without a chronic physical illness. Many young people with epilepsy experience multiple coexisting mental health disorders and also have additional needs, including autism spectrum disorder or intellectual disabilities. Mental health disorders in young people with long-term physical health conditions are associated with reduced quality of life (QoL), poorer physical health, worse mental health in parents and caregivers, and increased chance of death. Yet, despite the high level of need in this population, the majority of mental health disorders go unrecognised and are inadequately treated. Treatment for mental health disorders in paediatric epilepsy is significantly limited because specialist child and adolescent mental health services are not integrated with physical healthcare interventions so treatments are hard to access and are delivered late in the care pathway. Furthermore, standard evidence-based interventions typically do not address multiple coexisting mental health disorders and are not personalised to meet the specific needs of young people with epilepsy. Objectives The overarching aim of the programme was to transform the treatment of mental health disorders in young people by identifying and treating mental health problems from within epilepsy services to enable early detection and intervention. We had four main objectives, organised around four inter-related work packages (WPs) and underpinned by significant patient and public involvement (PPI): Optimise a modular cognitive–behavioural treatment for common mental health problems for use in children and young people with epilepsy. We aimed to adapt and personalise the existing materials to: address epilepsy-specific issues, including the relationship between epilepsy and mental health; provide epilepsy-specific examples; and account for the learning difficulties often experienced by young people with epilepsy. Train and supervise NHS staff from within paediatric epilepsy services to ensure that the personalised, modular psychological Mental Health Intervention for Children with Epilepsy (MICE) developed in WP1 could be delivered with competence, fidelity and flexibility in a sustainable manner. Conduct a parallel-group, multicentre, open-label randomised controlled trial (RCT) to compare the clinical and cost-effectiveness of the MICE treatment in addition to assessment-enhanced usual care (UC) with assessment-enhanced UC alone. Extend the quantitative findings by providing a rich and detailed qualitative analysis of therapeutic change and experience of the MICE. Methods Work package 1 We personalised and adapted the MICE over 12 months, taking a mixed-methods concurrent iterative approach to ensure thorough development and optimisation of the treatment. We conducted a systematic literature search, 6 iterative focus groups of 5 young people and 10 caregivers, 6 iterative focus groups of 12 professionals, and consulted with experts in mental health and epilepsy to personalise and adapt the treatment materials. Twelve young people (five male; seven female) received the adapted intervention, with ongoing development and adjustment based on feedback and evaluation through Plan-Do-Study-Act cycles. Work package 2 Fifteen professionals with limited or no prior experience in delivering mental health interventions underwent a rigorous 6-month training programme aimed at developing their skills to deliver the MICE effectively. This included face-to-face practical workshops, ongoing supervision and completion of a minimum of one training case per therapist. Training included evidence-based psychological techniques, assessment skills, goal-setting and risk management. Additional health professionals also attended the training workshops to support the long-term integration of MICE in epilepsy services. Thirty-four young people aged 3–18 years received the MICE treatment during this training phase, with measures compared at baseline and post treatment to assess outcomes. Supervision was provided regularly, focusing on skill development and monitoring patient progress. Therapists were assessed for competence using the Cognitive Therapy Rating Scale Revised (CTSR) at a threshold aligning with national standards for training of mental health professionals. Qualitative interviews with professionals provided insights into their experiences and perspectives of the intervention and training, and were analysed using thematic analysis. Work package 3 We conducted a parallel-group, multisite, open-label RCT to compare MICE in addition to assessment-enhanced UC with assessment-enhanced UC alone. We recruited participants aged 3–18 years with epilepsy and at least 1 mental health disorder from 13 NHS epilepsy services across England and Northern Ireland. Out of the 1401 young people considered potentially eligible to participate between May 2019 and January 2022, 334 were randomised to receive the MICE in addition to UC, or assessment-enhanced UC alone. The MICE, delivered predominantly by telephone or video conferencing, was administered by 21 competent professionals with limited prior mental health training. Participants received up to 20 weekly sessions over 6 months. UC included referral to local mental health services, and was considered assessment-enhanced as all participants completed a detailed diagnostic assessment prior to randomisation. The primary outcome measure was the parent-reported Strengths and Difficulties Questionnaire (SDQ) total difficulties score at 6 months post randomisation. Secondary measures included the impact score of the SDQ, the Revised Child Anxiety and Depression Scale (RCADS) and measures of seizure severity, paediatric QoL and parental mental health. Therapist competence and adherence to the MICE manual were evaluated using the CTSR and session content analysis. All analyses were conducted on an intention-to-treat basis, and measures were repeated at 12 months post randomisation to examine whether any changes were maintained over follow-up. Follow-up rates were good, with 95% participants completing the primary outcome measure at 6 months’ follow-up. Work package 4 Twenty-five families who participated in WP3 were purposively sampled to take part in two semistructured qualitative interviews conducted immediately after randomisation and at 6 months’ follow-up. Interviews were conducted flexibly to accommodate preferences, addressing topics such as mental health, epilepsy and stigma. Data were analysed longitudinally using a combination of interpretative phenomenological analysis and framework analysis, focusing on participants’ lived experiences and psychological constructs. An analytical framework was developed from the initial six cases, and subsequent interviews were analysed using this framework to identify individual longitudinal themes. This method allowed for a nuanced exploration of barriers and facilitators to change in the therapeutic process. Results Work package 1 Overall, our findings demonstrated that it was not necessary to design an entirely new treatment for young people with epilepsy and mental health disorders. We created a personalised and adapted MICE treatment manual and user guide for therapists, along with developing training materials and a study website. Personalisation and adaptation included: (1) development of a core module to provide psychoeducation about mental health and epilepsy; (2) development of three optional modules to address sigma, parental mental health and transition to adult services and (3) personalisation and addition of epilepsy-specific examples throughout the treatment manual. Our findings emphasised the importance of ongoing training and supervision for staff and identified organisational support needs to overcome barriers to implementation. Work package 2 The study demonstrated the feasibility of training professionals in paediatric epilepsy services to achieve competence in delivering the MICE within 6 months, with 14–15 therapists reaching clinical proficiency. Positive feedback was received regarding the training and intervention, with high satisfaction ratings and therapists recognising the benefits for families and their own professional development. However, concerns were raised about therapist confidence in delivering psychological treatments and the time commitment required. Adjustments were made to the supervision model to minimise therapist burden while maintaining competence standards. Patient outcomes indicated that delivering the MICE to families of children with epilepsy was feasible, with significant improvements in mental health symptoms and QoL observed post intervention. These findings supported the progression to a definitive RCT to establish clinical effectiveness and the cost-effectiveness of the intervention. Work package 3 Participants who received the MICE reported significantly greater reductions in symptoms as measured by the SDQ at 6 months’ follow-up, with treatment gains maintained at 12 months’ follow-up. There were no significant effects of intellectual disability, autism spectrum disorder, age, gender, primary mental health difficulty, seizure severity or parental mental health at baseline on treatment outcome, indicating that the MICE intervention is suitable for patients typically seen in NHS epilepsy services. Participants in the assessment-enhanced UC condition also reported some improvements in symptoms; this may, in part, be due to the comprehensive screening and assessment used in the trial which facilitated access to mental health services. In addition to the improvements in youth mental health, caregivers of young people in the MICE intervention also reported significant reductions in symptoms of anxiety and depression, compared to a deterioration in mental health reported by caregivers in the UC condition. Therapists demonstrated high competence and treatment fidelity, with 91% of sessions adhering to the expected content outlined in the MICE manual. The within-trial economic analyses found that the average cost of the MICE was approximately £1466 per participant, with a cost per session of approximately £80. MICE was cost-effective at the National Institute of Health and Care Excellence threshold of £20,000 per quality-adjusted life-years (QALYs) when the intervention is considered to substitute part of standard care in epilepsy services, and when young person and parent QALYs are combined, although results were inconclusive for cost-effectiveness using the SDQ. Work package 4 Comparative analysis pre and post intervention illuminated cognitive, behavioural and emotional shifts resulting from the MICE tools and materials. Participants demonstrated improvements in the experience and/or understanding of their condition. The qualitative analysis underscored the quantitative outcome changes reported in WP3. This mixed-methods approach offered a profound insight into processes of change and also fostered emotional engagement with stakeholders during the dissemination of our research findings. Conclusions Our research is the first to demonstrate that a modular cognitive–behavioural intervention personalised for epilepsy is superior to assessment-enhanced UC in improving emotional and behavioural symptoms in young people with epilepsy and common mental health disorders. The positive treatment effects were observed in young people across the age range and in those with intellectual disabilities or autism spectrum conditions. Treatment gains were maintained at 12 months’ follow-up and extended to improvements in caregivers' mental health. The MICE is likely to be cost-effective when young person and caregiver QALYs are combined, but given the challenges with measurement of health-related quality of life in this population further economic analyses are required. The MICE research programme has had significant impact and is now being used to deliver routine psychological support in NHS paediatric epilepsy services at pilot sites across England as part of NHS England’s Children and Young People Transformation Programme. It could also serve as a model for integrating mental health care for children and young people with other chronic physical illnesses, advocating for modification of existing interventions and evaluation using a mixed-methods approach. Further research is needed to explore the impact of MICE using young person-reported outcome measures, take a greater focus on parental mental health, analyse the impact of therapist skill on patient outcome, and understand the factors that contribute to the successful integration of mental and physical health care. Trial registration The trial is registered as ISRCTN57823197. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: RP-PG-0616-20007) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 9. See the NIHR Funding and Awards website for further award information.

Background Intensive care unit patients are vulnerable to hospital-acquired and ventilator-associated pneumonias. Immediate antibiotic therapy is warranted. Because: (1) many different bacterial species can be responsible and (2) it takes at least 2 days for conventional microbiological investigation, empirical broad-spectrum agents are given. This creates two hazards. First, these may promote overgrowth of undesirable bacteria in the patient’s gut. Second, the pneumonia may involve bacteria resistant to the antibiotic, precipitating failure. An alternative strategy – explored by the INHALE programme – is to use molecular tests to rapidly characterise the pathogen(s), facilitating early tailoring of antimicrobial therapy. Selecting a molecular test to trial in hospital-acquired and ventilator-associated pneumonias INHALE’s work package 1 used routinely collected respiratory samples (n = 652) from intensive care units at 15 hospitals, comparing the organisms found by two multiplex polymerase chain reaction systems (FilmArray and Unyvero) with those from routine culture. The molecular tests found bacteria in more specimens (74.2–60.4% vs. 44.2%) and found them more quickly (70 minutes to 6 hours vs. > 2 days). Diagnostic accuracy performance was similar, but the FilmArray Pneumonia Panel was swifter, more convenient and better supported. Accordingly, it was carried forward to the work package 3 randomised controlled trial. Defining the epidemiology and management of hospital-acquired and ventilator-associated pneumonias Work package 2 reviewed patient-level data for hospital-acquired/ventilator-associated pneumonias at four intensive care units in England; 142 patients were considered, of whom only 46.5% received appropriate empirical therapy. The cure rate of pneumonia was 62.7%. Enterobacterales, Pseudomonas aeruginosa, Staphylococcus aureus and Haemophilus influenzae predominate, corresponding with international literature. Compliance with local guidelines varied considerably among sites, as did the guidelines’ recommendations. Clinical trial of the effects of rapid microbiology on antibiotic stewardship and clinical cure Intensive care unit hospital-acquired/ventilator-associated pneumonias patients about to receive new or changed antibiotics were randomised to: (1) a FilmArray Pneumonia Panel test, supported by an algorithm translating its outputs into treatment advice, or (2) ‘standard-of-care’, with empirical antibiotics based on the hospital’s guidelines. Fourteen intensive care units participated, recruiting 545 adults and children over 2 years, with a 4-month-interruption owing to coronavirus disease discovered in 2019. Molecular tests were performed in intensive care unit, obviating transport delays. Two coprimary outcomes were considered. First, the proportion of patients on ‘active and proportionate’ antibiotics 24 hours after randomisation. This was substantially higher in the intervention arm (76.5% vs. 55.9%). Second, the proportion of patients with clinical cure of pneumonia 14 days post randomisation. This was marginally inferior in the intervention arm (56.4% vs. 64.7%), with the 95% confidence interval for the difference failing to exclude a preset 13% non-inferiority margin. Among secondary outcomes, there were slightly more deaths in the intervention arm, and slower improvement of clinical [sequential organ failure assessment and paediatric logistic organ dysfunction (PELOD)] scores, without statistical significance. Various interpretations are possible. Worse clinical outcomes may represent chance, given the borderline statistics. Alternatively, hospital pneumonia may, in its early stages, involve mixed bacteria, leading to a benefit from broad-spectrum therapy, which was more often given in the control arm. Behavioural studies INHALE work package 4 used mixed-method designs to examine prescribers’ perceptions of the Pneumonia Panel and its use. Although most recognised the value of rapid results and the importance of antibiotic stewardship, Pneumonia Panel results had a limited impact on prescribing behaviour. Many clinicians were reluctant to avoid or stop initial broad-spectrum in response to test results. The decision to prescribe and maintain broad-spectrum antibiotics often represented an attempt to ‘err on the side of caution’ to protect the patient (and clinician), where the proximal need to protect was more imperative than the more distal threat of antibiotic resistance. Costs and cost-effectiveness of rapid molecular microbiology for hospital-acquired and ventilator-associated pneumonias The additional cost of the Pneumonia Panel (£198) represents a very small component of total intensive care unit costs and are typically below 2% of total costs per intensive care unit patient (base case). We found a reduction in total costs with patients in the intervention arm averaging £33,148, and patients in the control arm averaging £40,951. However, it remains unclear how this arises. There was evidence of the Pneumonia Panel being cost-effective in terms of stewardship, but not for clinical cure; consequently, we cannot conclude an economic advantage for the Pneumonia Panel. However, this does not take account of the potential benefit of improved stewardship in terms of reduced antimicrobial resistance. Limitations The multiplex systems compared in work package 1 are subject to repeated updating, meaning that the better system might change with time. Both work package 1 and work package 3 were limited by the low prevalence of antibiotic resistance in England, precluding rigorous evaluation of resistance gene detection. Last, and critically, the coronavirus disease pandemic substantially changed the types of patients recruited. Conclusions and future work Overall, INHALE demonstrates the potential of molecular testing to swiftly detect pathogens in the respiratory secretions of intensive care unit pneumonia patients and shows that these data can influence prescribing. What remains elusive, needing further work, is translating the results into optimal therapy and improved clinical outcomes. Trial registration (work package 3) This trial is registered as ISRCTN16483855. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref.: RP-PG-0514-20018) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 6. See the NIHR Funding and Awards website for further award information. Plain language summary Intensive care unit patients often develop pneumonia. The laboratory then takes several days to identify the type of bacteria responsible. Meantime, the doctor gives ‘broad-spectrum’ antibiotics to kill the most likely bacteria. This approach carries two risks: undesirable and antibiotic-resistant bacteria may overgrow in the gut second, the pneumonia may involve bacteria resistant to the antibiotic. An alternative approach, explored by INHALE, is to use polymerase chain reaction (as in coronavirus disease testing) to rapidly identify the bacteria, guiding treatment. We compared the bacteria found by two polymerase chain reaction systems with those found by microbiology labs across 15 intensive care units. Polymerase chain reaction was faster and found more bacteria. One polymerase chain reaction system, FilmArray, was better and was used for further research (randomised clinical trial), in 14 English intensive care units. Half of 545 patients received treatment guided by FilmArray. The other half had standard care, with broad-spectrum antibiotics until routine lab results were received. More FilmArray patients (76.5% vs. 55.9%) received appropriate antibiotics within 24 hours, so helped lessen overprescribing of antibiotics, but, surprisingly, their clinical cure outcomes were slightly worse (56.7% cure vs. 64.5%). Other clinical outcomes tended to favour the control arm’s patients, but differences were not statistically significant. We do not know what caused this difference. It may be chance, or it may indicate that intensive care unit pneumonias are more complex than supposed, with patients benefitting from an early broad-spectrum antibiotic. We surveyed intensive care unit clinicians’ views about these tests. Most welcomed their speed, but many were wary about stopping broad-spectrum antibiotics early. They were also torn between wanting to conserve antibiotics and wanting the ‘best’ for their patient, often prioritising the latter. Each test costs £198 per person, but the average total care cost per patient was lower (£33,148 vs. £40,951). The reason for this saving remains unclear. Overall, INHALE showed the potential of polymerase chain reaction tests in intensive care unit pneumonias, and that the results could influence prescribing. How best to translate this into better patient outcomes remains unknown at present. A poem summarising the study and written by the programme’s patient and public involvement group, can be seen here: www.youtube.com/watch?v=7_2bqWlJzFE Scientific summary Background Hospital-acquired and ventilator-associated pneumonias (HAP and VAP) affect one in five intensive care unit (ICU) patients and are responsible for significant excess morbidity, mortality and costs. They can be caused by a range of micro-organisms, including bacteria, viruses and fungi. Aetiological diagnosis is traditionally performed by bacteriological culture, taking 2–4 days. In the meantime, patients receive empirical broad-spectrum antibiotics, which is suboptimal as it frequently overtreats the pathogen present, driving antimicrobial resistance (AMR), or conversely, undertreats, potentially leading to worsened clinical outcomes. We investigated whether rapid molecular diagnostic methods would lead to improved treatment of HAP and VAP. Objectives The objectives of the programme were: To define, across four different ICUs, the accuracy of three molecular diagnostic systems claiming to rapidly identify pathogens and their resistance genes in respiratory specimens from patients with HAP/VAP. Based on (1) above, to select the best-performing test and to conduct a randomised controlled trial (RCT), comparing clinical outcomes and antibiotic utilisation in patients whose treatment is guided by the test versus those managed conventionally with empirical treatment refined by microbiological culture. To measure clinicians’ willingness to adopt molecular diagnostics and treat patients based on their results. To determine whether the outcomes justify the cost. Work package 1: clinical laboratory evaluation Methods Surplus lower respiratory samples were collected from patients with suspected HAP or VAP. Samples were tested on the BioFire FilmArray Pneumonia Panel and the Curetis Unyvero Pneumonia Panel and results compared to routine culture. Analysis was according to a predetermined statistical analysis plan using Stata® (StataCorp LP, College Station, TX, USA) and R (The R Foundation for Statistical Computing, Vienna, Austria). Results Six hundred and fifty-two eligible samples were collected from 15 sites between September 2016 and May 2018. Routine culture generated a result in a median time of 70.2 hours [interquartile range (IQR) 51.1–92.1 hours] and 44.2% of samples were positive for likely pathogens. Considerable site-to-site variation was observed, highlighting shortcomings of routine microbiology as a gold standard. There were 631 eligible test results for the Unyvero, and 632 for the FilmArray. The overall positivity rate was 60.4% for Unyvero and 74.2% for FilmArray, considerably higher than for routine microbiology (p < 0.0001). The range of organisms detected by polymerase chain reaction (PCR) was as expected and broadly similar to routine microbiology. Polymerase chain reaction assay sensitivity was > 95% for most target bacteria, with negative predictive values (NPVs) > 98%. Specificity and positive predictive values (PPVs) were lower, due to the PCR tests detecting more organisms per sample and finding more positive samples than routine microbiology. To address the problem of culture representing a poor gold standard, performance parameters were also obtained using Bayesian latent class (BLC) analyses. This showed routine microbiology was the least sensitive technique. Sensitivity values for the PCR tests remained high; specificity and PPV values for both PCR tests increased. The tests were evaluated for their other characteristics such as speed, size and useability. A final decision-making meeting was held with all stakeholders. The FilmArray Pneumonia Panel was chosen for progression to the randomised controlled trial in work package (WP) 3. Work package 2: epidemiology of the units Methods This work took place at four ICUs participating in WP1. There were two components: (1) collection of patient-level data and (2) collection of unit-level data. Individual participants needed to have an eligible sample included in WP1, in addition to providing consent or assent to participate. Data were collected for up to 21 days after enrolment. The data were used to determine whether antibiotics prescribed were appropriate and proportionate against pathogens found by either PCR and/or culture. Data were analysed according to a predefined analysis plan using Stata and R. In addition to patient-level data, unit-level prescribing and microbiology data were collected from the participating ICUs using hospital information systems. Data were acquired for a period of 15 months, from October 2016 to December 2017. Results One hundred and forty-two participants were recruited. The majority were male (65.5%) with a median age of 57.9 years, 24.6% were children. Fifty-two and one-tenth per cent had VAP and the remainder had HAP. The majority (86%) had received an antibiotic prior to enrolment and over half of prescriptions were broad-spectrum agents. For their episode of HAP or VAP, 51.4% received empirical treatment within the ambit of local guidelines, 32.4% received non-guideline empirical therapy and the remainder received non-empirical therapy based on a susceptibility result. Commonly used regimens were piperacillin/tazobactam ± an aminoglycoside, coamoxiclav, ceftazidime or ciprofloxacin + teicoplanin. At day 21, 62.7% of patients were considered cured of pneumonia when using the cure definition in the WP2 protocol, increasing to 74.6% when the WP3 definition was used. Overall mortality at 21 days was 18.3%. All three methods show a slightly higher cure rate when treatment was judged appropriate as opposed to inappropriate. For routine microbiology, cure rates were reasonably similar across all stewardship categories. However, in the case of PCR, cure rates were noticeably lower when inactive treatment for a positive result was given, at just (17/35) 48.6% compared with (71/103) 68.9% among those with an appropriate treatment based on FilmArray results (χ2 = 4.69, p = 0.03) and (20/37) 54.1% vs. (65/99) 65.7% based on Unyvero results (χ2 = 1.55, p = 0.21). At the unit level, adults had a HAP/VAP rate of almost double that of children, with the rate 22.9 cases/1000 bed-days at University College London Hospital (UCLH) ICU. In terms of microbiology, there were discrepancies in the positivity rate of samples, with UCLH having a ~35% positivity rate compared to ~65% at Cromwell and ~80% at Great Ormond Street Hospital (GOSH). The epidemiology of HAP/VAP was as expected at Norfolk and Norwich University Hospital (NNUH) and UCLH, but at the Cromwell there was a radically different epidemiology, dominated by Pseudomonas aeruginosa and Enterobacterales but with a substantial proportion (14%) of Acinetobacter baumannii. At GOSH, traditionally community-acquired organisms, Haemophilus influenzae and Moraxella catarrhalis, were frequently isolated from lower respiratory tract infection (LRTI) samples. Antibiotic prescribing also differed between the units. Work package 3: randomised controlled trial Methods We ran a pragmatic, multicentre, open-label, parallel-group, RCT to investigate clinical, safety and cost-effectiveness of the Pneumonia Panel test plus trial based prescribing algorithm versus standard care across 14 ICUs. Those who received an antibiotic to treat new or worsening HAP or VAP and provided consent or assent were eligible to participate. Participants randomised to the treatment arm received a Pneumonia Panel test performed at the point of care, alongside an optional treatment algorithm designed to encourage good antibiotic stewardship. Primary outcomes were equivalence of clinical cure of pneumonia at day 14 and improvement in antibiotic stewardship at 24 hours. Clinical cure rate was initially assumed at 70% with a non-inferiority margin of 13%, but cure rate was lowered to 55% following the inclusion of coronavirus disease discovered in 2019 (COVID-19) patients. This sample size required at least 528 participants to achieve 91% power with a significance level still at 5% and was inflated to 552 to allow for 5% attrition. Analyses were based on intention to treat (ITT) and followed a predefined statistical analysis plan using Stata version 17. Results Five hundred and fifty-four participants were recruited, of which 92 were children; primary outcomes were available for 531. Median age for adults was 61 years, and 7.5 months for children. At randomisation, 33.6% had COVID-19 infection. Time to result post randomisation was 1.7 [standard deviation (SD) 0.8 hour] for the Pneumonia Panel compared to 110 (SD 116.2 hours) in the control arm. Intention-to-treat analysis for the coprimary stewardship outcome at 24 hours after randomisation showed that 205/268 (76.5%) intervention arm participants were receiving active and proportionate antibiotics by 24 hours, versus 147/263 (55.9%) in the control arm [estimated difference after accounting for site 21%, 95% confidence interval (CI) 13% to 28%, p < 0.001]. One hundred and fifty-two of 268 (56.7%) intervention arm participants had clinical cure of pneumonia at 14 days, versus 171/265 (64.5%) control patients. The estimated difference, after accounting for site, was −6% with 95% confidence limits of −15% to 2%. These values overlap the non-inferiority margin of 13%, meaning that non-inferiority was not established. Analyses of secondary outcomes supported the primary stewardship results, with improvements consistently apparent for the intervention arm. We saw a 28-day mortality of 31.3% in the intervention arm and 28.2% in the control arm, a non-significant difference. No significant differences between arms were observed for ICU length of stay, ventilator-free days, incidence of septic shock or occurrence of secondary pneumonia. Work package 4: behavioural study Methods Work package 4 used mixed methods across four separate but related studies examining prescribers perceptions of (1) antibiotic prescribing in ICU, (2) the application of molecular diagnostic techniques and (3) their prescribing behaviour in response to Pneumonia Panel results. Study 1 used focus groups and interviews to explore ICU clinician perceptions of antibiotic prescribing using two fictitious vignettes of patients with HAP/VAP. Data were analysed using thematic analysis, applying the Necessity Concerns Framework (NCF). Study 2 used data collected through Study 1, focusing in more detail on clinicians’ beliefs about rapid molecular diagnostic technology ‘in principle’, and its potential role in prescribing. Study 3 used qualitative methods to explore clinicians’ perceptions of using the Pneumonia Panel in practice. Study 4 used ecological momentary assessments (EMA) to examine whether Pneumonia results influenced individual prescribing decisions for specific patients, and to elicit the clinicians’ beliefs that influenced each decision. Results The overarching findings from Studies 1 and 2 provided novel insights into the psychology of prescribing. When asked about the risks of broad-spectrum antibiotics, few clinicians focused on the potential toxic effects of broad-spectrum antibiotics to individual patients. The major concern about using broad-spectrum antibiotics was perceived to be increasing AMR, which clinicians were aware of and concerned about. However, this concern was juxta-positioned against a more immediate and profound concern about the patient in front of them. Therefore, a prescription and continuation of broad-spectrum antibiotics was an attempt to ‘err on the side of caution’ in protecting the patient (and clinician) from the adverse consequences of not prescribing, over-riding concerns about AMR. The immediate necessity to protect fuelled the necessity to prescribe, over-riding more distal concerns about AMR. Study 3 found clinicians’ perceptions about using the ‘Pneumonia Panel’ to treat HAP/VAP patients were nuanced. Although many perceived the value of these tests for supporting antibiotic decision-making and local antimicrobial stewardship (AMS), many harboured concerns about their application in practice (e.g. false negatives/positives). Quantitative data from Study 4 highlighted that hospital guidelines are one of the most influential factors for antibiotic decision-making; however clinicians’ confidence in, and adherence to them, was low (24.9% of decisions were not consistent with local hospital guidelines). Clinicians also reported low confidence in, and adherence to, the WP3 prescribing algorithm (67.6% of decisions were not consistent with the algorithm). Work package 5: health-economic analysis Methods Health-economic data relating to cost of ICU stay were collected for individual participants of WP2, to enable calculation of the costs associated with hospital stays for HAP or VAP. In parallel with WP1, we estimated the cost-per-test for the two PCR diagnostics being evaluated. A review of health-economic models used in previous studies investigating costs of HAP and VAP was conducted. Clinician interviews were performed and a clinical pathway developed. These components were used to develop a bespoke health-economic model of HAP and VAP for this study. Health-economic analyses were conducted alongside WP3 following a predefined health-economic analysis plan. We calculated the additional cost per additional person on active and proportionate antimicrobial therapy within 24 hours of clinical diagnosis (‘stewardship’) as well as the additional cost per additional clinical cure of pneumonia at 14 days post randomisation (‘cure’). Costs were obtained from hospital finance departments and represented hospital income. We also used NHS reference costs. Costs were in 2020–1 Great British pounds, apart from early work to estimate cost of test used in WP1 and WP2 which were in 2018–9 Great British pounds. Results Work alongside WP2 estimated that mean costs for a hospital stay for an adult with HAP or VAP was £42,136. The cost of the Pneumonia Panel was calculated as £189, and this was revised to £198 during WP3. The length of ICU stay was the largest component of cost. In WP3, total costs for the control arm were £40,951 compared to £33,148 for the intervention arm, a difference of £7802 (95% CI −15,696 to 92). For economic evaluation of stewardship and clinical cure, we found estimates of difference in cost for the intervention arm compared to the control arm of −£7373 and −£7147, respectively. For these economic analyses, lower costs for the intervention arm persisted for all sub-analyses, including separate costs for survivors and those that died. Cost-effectiveness of the Pneumonia Panel was demonstrated for antibiotic stewardship but not for clinical cure. Conclusions Rapid molecular diagnosis of the aetiology of HAP and VAP was faster and more sensitive than standard care culture. PCR tests revealed additional pathogens which were not reported by culture, possibly due to reasons of lack of standardisation in interpretation and reporting. The Pneumonia Panel was chosen for further progression to the RCT and placed at the point of care within participating ICUs. This strategy increased uptake of the test and recruitment to the trial. The Pneumonia Panel intervention was successful in improving AMS by 21% in absolute terms. This was despite reservations expressed by clinicians regarding de-escalation of antibiotics for seriously ill patients. The intervention was also associated with a cost saving linked to reduced costs for the ICU stay, although the drivers of this are not yet clear. However, equivalence of clinical cure was not demonstrated. Other clinical outcomes such as mortality and sequential organ failure assessment (SOFA) did not demonstrate significant differences but tracked with clinical cure. It is not currently clear whether the failure to demonstrate clinical cure is related to a small but real effect, another explanation or pure chance. This must be established to confirm the safety of the intervention. Our behavioural work suggests that to realise the potential for molecular diagnostic technologies to improve AMS, we need to employ a ‘technology plus’ approach that acknowledges the challenges that clinicians face when applying technological solutions to the care of individual patients. Future work Future work should focus on whether or not use of the intervention to improve AMS leads to worse clinical outcomes. Such a finding could have broad reaching implications if true and must therefore be investigated further. Work is already underway to carry out a deeper analysis of the existing trial participants and their data. Should these analyses prove inconclusive the next step would be a follow-on trial focusing particularly on clinical outcomes. Future work should also involve the design and development of behavioural materials and interventions to support clinicians with antibiotic decision-making. Another area of interest is the use of broader diagnostic techniques such as metagenomics, which provide a complete overview of the lung microbiome as opposed to focusing on specific pathogens. Trial registration This trial is registered as ISRCTN16483855. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref.: RP-PG-0514-20018) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 6. See the NIHR Funding and Awards website for further award information.

Background Raised blood pressure affects 10% of pregnancies worldwide, of which around half develop pre-eclampsia including proteinuria, causing maternal and perinatal morbidity and mortality. Objectives To develop and test interventions for self-monitoring of blood pressure designed to improve the detection and management of hypertension in pregnancy. Additionally, to test the accuracy of self-testing of urine for protein, a key marker of pre-eclampsia. Design and methods Development phase, a pilot trial, two large randomised controlled trials of self-monitoring of blood pressure interventions with integrated economic evaluations, linked qualitative work, a large survey and a diagnostic accuracy study of self-testing for proteinuria, and finally economic modelling. Setting and participants Antenatal clinics in 16 English hospitals. Participants were pregnant women and antenatal healthcare professionals. Patient and public involvement Comprehensive involvement from initial development through to dissemination, with collaboration from both individuals and relevant charities and organisations. Interventions Self-monitoring of blood pressure supported by app to improve the detection (BUMP1) and management (BUMP2) of raised blood pressure in pregnancy (WS3.2.1 and 2). Proteinuria self-testing by pregnant hypertensive women (UDIP, WS4). Main outcome measures Qualitative data informed the development of the BUMP App and trial (WS1). Feasibility of self-monitoring of blood pressure in hypertensive pregnancy (recruitment, retention, adherence and intervention persistence) (WS2). Prevalence of self-monitoring of blood pressure during pregnancy (WS3.1). Time to diagnosis of hypertension defined in routinely recorded clinical data (BUMP1, WS3.2.1) and difference in mean systolic blood pressure recorded by healthcare professionals between randomisation and birth (BUMP2, WS3.2.2). WS3.3 and 4.2 evaluated experiences with the trial and UDIP respectively using inductive and deductive thematic analysis. Within-trial cost–consequence analysis and long-term cost-effectiveness modelling (WS3.4 and WS5). Proteinuria testing accuracy (WS4.1). Results WS1: Areas important to staff included providing clear patient information and supporting them in decision-making in the context of discrepant readings. The intervention was optimised iteratively with pregnant women. WS2: A feasibility trial showed that the self-monitoring of blood pressure intervention was feasible and acceptable. WS3.1: Data from a survey of 5181 women showed that 19% of pregnant women were currently self-monitoring blood pressure but only 482/983 (49%) shared this information with healthcare professionals. WS3.2.1: Self-monitoring of blood pressure in addition to usual care did not lead to an earlier diagnosis of clinic hypertension in routinely recorded clinical data with no evidence of differences in maternal or perinatal outcomes or serious adverse events (BUMP1). WS3.2.2: Self-monitoring of blood pressure in pregnancy hypertension did not improve clinic blood pressure control, with no difference in maternal or perinatal outcomes or serious adverse events (BUMP2). WS3.3: Self-monitoring was generally accepted by women and professionals with differences in views of which blood pressure data to give precedence. Women found self-monitoring empowering, provided health professionals considered home readings in their management. On occasion self-monitoring proved unsettling due to uncertainty. WS3.4: Within trial economic analyses revealed no significant difference in overall total costs between trial arms in either BUMP1 or BUMP2. Women’s health-related quality of life (EQ-5D-5L) was similar between groups in both trials. WS4: Self-testing for proteinuria had a sensitivity of 0.71 (95% confidence interval 0.62 to 0.79) and a specificity of 0.89 (95% confidence interval 0.84 to 0.92) compared to laboratory protein–creatinine ratio testing and this was not clinically or statistically different when compared to healthcare professionals or a colorimetric monitor (UDIP, WS4.1). Self-testing was generally acceptable (WS4.2). WS5: Model frameworks capable of facilitating exploration of the long-term cost-effectiveness of combining self-monitoring of blood pressure with blood pressure treatment and management policies were developed for use in future research projects in the area. Implementation: Self-monitoring was rapidly and widely implemented during the pandemic but has yet to become fully embedded in clinical pathways. Limitations Self-monitoring of blood pressure by women in the control groups; difficulties in testing self-monitoring of blood pressure in clinical pathways in the absence of evidence or accepted treatment thresholds; process evaluation suggested women and professionals privileged different information. Conclusions Self-monitoring of blood pressure during higher risk or hypertensive pregnancy was feasible, acceptable, safe, and no more expensive, but did not improve the detection of hypertension or blood pressure control in those with hypertension when used alongside usual care. During the programme, self-monitoring of blood pressure entered common practice in pregnancy, a process accelerated by the pandemic. Pregnant women can read a dipstick for urinary protein with similar accuracy to healthcare professionals or colorimetric testing, and find this acceptable, suggesting that self-testing could be included in clinical pathways. Future work Future trials of interventions including self-monitoring of blood pressure should test strategies in the context of novel clinical pathways. Study registration This study is registered as Current Controlled Trials ISRCTN16018898. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research programme (NIHR award ref.: RP-PG-0614-20005) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 3. See the NIHR Funding and Awards website for further award information. Plain language summary Background and aims Home monitoring of blood pressure supports the management of raised blood pressure (hypertension) in the general population, but little was known about its use in pregnancy. Similarly, little was known regarding self-testing for protein in the urine, a marker of pre-eclampsia which is a serious condition linked to hypertension in pregnancy. The BUMP programme aimed to develop and test blood pressure home monitoring and protein self-testing to see if these could improve the detection of raised blood pressure and/or proteinuria and management of hypertension in pregnancy. Findings Focus groups and interviews with healthcare professionals and development work (designing the intervention) with pregnant women supported the development of a user-friendly app, trial design and materials. A survey identified that one in five of pregnant women currently home monitor blood pressure, increasing to half of those with hypertension, though did not share these readings with healthcare professionals. The BUMP trials recruited more than 3000 women at higher risk of pre-eclampsia, or those with raised blood pressure. Participating women were randomly allocated to either usual care or usual care plus self-monitoring of blood pressure. Home monitoring of blood pressure did not result in earlier recording of hypertension in clinic. Over half of the women diagnosed with hypertension had raised blood pressure at home. For women with high blood pressure, home monitoring did not improve blood pressure control. Home monitoring of blood pressure was safe and engagement was high, with the majority of women continuing home monitoring throughout pregnancy. No differences in costs or quality of life were found. Pregnant women self-tested for urinary protein with similar accuracy to healthcare professionals and were happy to test finding it, convenient and reassuring. Staff found home readings valuable, although some were less willing to incorporate them into antenatal care. Conclusions Home monitoring of blood pressure during higher risk or hypertensive pregnancy was acceptable, safe, and no more expensive than usual care alone but it did not improve the detection of hypertension or blood pressure control in those with hypertension when used alongside usual care. Self-testing of urine for protein could support remote care of hypertensive women. Scientific summary Background Raised blood pressure (BP) affects 10% of pregnancies worldwide, of which around half develop pre-eclampsia, a leading cause of maternal and perinatal morbidity and mortality. Early detection of raised BP and/or proteinuria and subsequent management of pregnancy hypertension is therefore important and could be improved through self-monitoring whilst empowering women and allowing reduced antenatal visits. However, little evidence was available to guide the utilisation of self-monitoring in pregnancy. Objectives The overall aim of this programme was to evaluate whether self-monitoring of BP (SMBP) could improve the detection of raised BP during pregnancy, whether it was feasible for use in the titration of antihypertensive medication in pregnancy hypertension and whether women with raised BP in pregnancy could accurately test their urine for proteinuria, a key marker for pre-eclampsia. Linked qualitative and economic components examined patient and professional experiences of self-monitoring and responses to it along with cost-effectiveness within the trial and in the longer term. Key research questions WS1: Development What are the best self-monitoring interventions to use? How can SMBP and urine best integrate into current antenatal care pathways? WS2: Blood pressure self-monitoring during pregnancy for anti-hypertensive titration Is titration of antihypertensive medication during and after pregnancy using self-monitoring feasible? What is the participant and professional experience of such monitoring and titration? WS3: Self-monitoring to improve the detection and management of raised blood pressure in pregnancy What is current practice in BP self-monitoring in pregnancy? Can BP self-monitoring improve the detection and management of hypertension during pregnancy? How is BP self-monitoring in pregnancy implemented in daily life and routine clinical practice? Is BP self-monitoring in pregnancy cost-effective? WS4: Self-monitoring of urinary protein in hypertensive pregnancy Can pregnant women with hypertension accurately self-monitor for proteinuria and could this detect pre-eclampsia earlier than usual care? Is self-monitoring of urine practical and acceptable to hypertensive pregnant women, their midwives and obstetricians? WS5: Modelling of the potential long-term costs and consequences Is SMBP and protein in hypertensive pregnancy potentially cost-effective and what are the key parameters affecting this? Methods WS1: Development Focus groups and interviews were undertaken with NHS staff (including obstetricians, community and hospital midwives). We worked iteratively with women talking through their experiences of prototypes of the self-monitoring app and trial materials (Band R, Hinton L, Tucker KL, Chappell LC, Crawford C, Franssen M, et al. Intervention planning and modification of the BUMP intervention: a digital intervention for the early detection of raised blood pressure in pregnancy. Pilot Feasibility Stud 2019;5:153. https://doi.org/10.1186/s40814-019-0537-z; Hinton L, Hodgkinson J, Tucker KL, Rozmovits L, Chappell L, Greenfield S, et al. Exploring the potential for introducing home monitoring of blood pressure during pregnancy into maternity care: current views and experiences of staff – a qualitative study. BMJ Open 2020;10:e037874. https://doi.org/10.1136/bmjopen-2020-037874). WS2: Blood pressure self-monitoring during pregnancy for antihypertensive titration The OPTIMUM feasibility trial It was an unmasked randomised controlled trial (RCT) comparing a SMBP versus usual care for the management of pregnancy hypertension. Women with chronic (CH) or gestational hypertension (GH) from four UK centres were randomised (2 : 1) intervention to control. Primary outcomes were recruitment, retention, adherence and persistence with the intervention (Pealing LM, Tucker KL, Mackillop LH, Crawford C, Wilson H, Nickless A, et al.; OPTIMUM-BP Investigators. A randomised controlled trial of blood pressure self-monitoring in the management of hypertensive pregnancy. OPTIMUM-BP: a feasibility trial. Pregnancy Hypertens 2019;18:141–9. https://doi.org/10.1016/j.preghy.2019.09.018). WS3: Self-monitoring to improve the detection and management of raised blood pressure in pregnancy 3.1 BUMP survey Pregnant women from antenatal clinics in 16 hospitals in England were invited to complete a survey about SMBP. 3.2.1 The BUMP1 trial (Dougall G, Franssen M, Tucker KL, Yu L-M, Hinton L, Rivero-Arias O, et al. Blood pressure monitoring in high-risk pregnancy to improve the detection and monitoring of hypertension (the BUMP 1 and 2 trials): protocol for two linked randomised controlled trials. BMJ Open 2020;10:e034593. https://doi.org/10.1136/bmjopen-2019-034593) It was a multicentre, RCT that recruited pregnant women at higher risk of pre-eclampsia at 20 weeks’ gestation. Women were randomised to BP self-monitoring with telemonitoring and usual care or to usual care alone. The primary outcome was time to the first recorded raised BP taken by a healthcare professional (HCP). Trial registration: NCT03334149. Recruitment 2018–9. Final follow-up April 2020. 3.2.2 The BUMP2 trial It was a multicentre, RCT that recruited women with CH and GH up to 37 weeks’ gestation. Women were randomised to BP self-monitoring with telemonitoring and usual care or to usual care alone. The primary maternal outcome was the difference in mean systolic BP recorded by HCPs between study entry and childbirth. Analyses were by intention to treat (ITT) and stratified by CH or GH. Trial registration: NCT03334149. Recruitment 2018–19. Final follow-up May 2020. 3.3 BUMP trials qualitative process evaluation In-depth interviews were carried out with 39 trial participants and 7 women who declined to take part in the trials to explore their experiences of self-monitoring BP or reasons for choosing not to. Twenty-one HCPs involved in women’s care or in the administration of the trial were interviewed. Interviews were purposively sampled from study sites. A planned ethnographic study was not feasible. Inductive and deductive thematic analysis was carried out on both qualitative data sets. Interviews were analysed using a coding frame that was developed from the research aims and incorporating additional themes that emerged from the data. 3.4 BUMP within-trial economic analyses National Health Service (NHS) perspectives were used for both within trial cost–consequences analyses. Patient-level resource use data were extracted from clinical notes and costed and women’s health-related quality of life was measured during the trials using the EuroQol EQ-5D-5L questionnaire, with responses converted to single index scores. Mean costs and EQ-5D-5L scores were computed and compared between trial arms. Within BUMP2, analyses were conducted separately for CH and GH cohorts. WS4: Self-monitoring of urinary protein in hypertensive pregnancy 4.1 The UDIP study It was a diagnostic accuracy study that recruited 345 pregnant women to self-test for urinary protein using visually read dipsticks. The primary reference test was protein–creatinine ratio (PCR) and secondary index tests included testing by antenatal HCPs and an automated colorimetric reader. Primary outcome measures were sensitivity and specificity. 4.2 UDIP qualitative study In-depth interviews were carried out with 21 pregnant hypertensive or pre-eclamptic women who took part in the UDIP study, and 18 HCPs who had experience working in antenatal care. Five focus group totalling 15 participants were conducted with HCPs. WS5: Modelling of the potential long-term costs and consequences With no cost or effect differences observed overall or across pre-specified subgroups in the BUMP1 and BUMP2 trials, the need for long-term cost-effectiveness modelling was negated. Instead, model frameworks to facilitate future exploration of the potential long-term costs and effects of combining SMBP with BP management policies for the prevention of hypertension-related complications, were developed. The models cover the pregnancy pathway and a subsequent ten-year period to capture the risks, costs, and consequences of women developing cardiovascular disease. Model parameters are entered as distributions to enable probabilistic sensitivity analysis, and the frameworks facilitate results being presented for women with differing characteristics and for different intervention effect sizes. Results WS1: Development (Band R, Hinton L, Tucker KL, Chappell LC, Crawford C, Franssen M, et al. Intervention planning and modification of the BUMP intervention: a digital intervention for the early detection of raised blood pressure in pregnancy. Pilot Feasibility Stud 2019;5:153. https://doi.org/10.1186/s40814-019-0537-z; Hinton L, Hodgkinson J, Tucker KL, Rozmovits L, Chappell L, Greenfield S, et al. Exploring the potential for introducing home monitoring of blood pressure during pregnancy into maternity care: current views and experiences of staff – a qualitative study. BMJ Open 2020;10:e037874. https://doi.org/10.1136/bmjopen-2020-037874) Focus groups and interviews were conducted with 147 NHS staff at seven different hospital sites. Areas identified as important included: providing clear patient information and supporting staff in decision-making in the context of discrepant readings. Analyses suggested that SMBP would be welcomed by HCPs, while also highlighting potential barriers. This work and iterative development with pregnant women supported the development of a pragmatic and workable trial with user-friendly materials and app. WS2: Blood pressure self-monitoring during pregnancy for antihypertensive titration (Pealing LM, Tucker KL, Mackillop LH, Crawford C, Wilson H, Nickless A, et al.; OPTIMUM-BP Investigators. A randomised controlled trial of blood pressure self-monitoring in the management of hypertensive pregnancy. OPTIMUM-BP: a feasibility trial. Pregnancy Hypertens 2019;18:141–9. https://doi.org/10.1016/j.preghy.2019.09.018; Pealing L, Tucker KL, Fletcher B, Lawley E, Chappell LC, McManus RJ, Ziebland S. Perceptions and experiences of blood pressure self-monitoring during hypertensive pregnancy: a qualitative analysis of women’s and clinicians’ experiences in the OPTIMUM-BP trial. Pregnancy Hypertens 2022;30:113–23. https://doi.org/10.1016/j.preghy.2022.09.006) Women from four UK centres were randomised: 158/222 (71%) of those approached agreed, comprising 86 women with CH (55 SMBP, 31 control) and 72 with GH (49 SMBP, 23 control). Outcome data were available from 154 (97%). The median number of days with home BP readings per week was 5.5 [interquartile range (IQR) 3.1–6.5] for those with CH and 6.1 (4.5–6.7) with GH. Participating women persisted with the intervention for 80% time from enrolment until delivery. Recorded clinic and study BPs were similar for both groups. Interviews showed that the women found SMBP feasible and acceptable and were highly motivated and proactive in their monitoring. They reported greater control and knowledge, which provided reassurance. Most women reported that they responded appropriately for out-of-range readings or symptoms. WS3: Self-monitoring to improve the detection and management of raised blood pressure in pregnancy 3.1 The BUMP survey (Tucker KL, Hodgkinson J, Wilson HM, Crawford C, Stevens R, Lay-Flurrie S, et al. Current prevalence of self-monitoring of blood pressure during pregnancy: the BUMP survey. J Hypertens 2021;39:994–1001. https://doi.org/10.1097/HJH.0000000000002734) Completed surveys were received from 5181/5555 pregnant women (93%). The analysis showed that 983/5181 (19%) were currently SMBP. Around half of those were hypertensive 189/389 (49%) and 794/4792 (17%) were normotensive. However, only 482/983 (49%) of those that monitored their BP reported sharing this information with their obstetric and midwifery team. Comparison to hospital demographic data suggested that respondents were broadly representative.1 3.2.1 The BUMP1 trial (Tucker KL, Mort S, Yu LM, Campbell H, Rivero-Arias O, Wilson HM, et al.; BUMP Investigators. Effect of self-monitoring of blood pressure on diagnosis of hypertension during higher-risk pregnancy: the BUMP 1 randomized clinical trial. JAMA 2022;327:1656–65. https://doi.org/10.1001/jama.2022.4712) A total of 2441 women were randomised to BP self-monitoring plus usual care or usual care alone (n = 1218). Primary outcome data were available from 2346 (96%) women. Baseline characteristics were similar and 15.5% developed hypertension. Time to detection of clinic hypertension was not significantly different between groups: –1.6 days [95% confidence interval (CI) –8.1 to 4.9, p = 0.6]. There was no significant difference in the incidence of severe clinic hypertension [adjusted relative risk 1.2 (0.9 to 1.7), p = 0.3], in maternal and fetal outcomes, or serious adverse events. Most women who developed high BP had self-monitored their BP within a week of diagnosis (73%), and half had raised home BP readings prior to a clinic diagnosis. 3.2.2 The BUMP2 trial (Chappell LC, Tucker KL, Galal U, Yu L-M, Campbell H, Rivero-Arias O, et al.; BUMP 2 investigators. Effect of self-monitoring of blood pressure on blood pressure control in pregnant individuals with chronic or gestational hypertension: the BUMP 2 randomized clinical trial. JAMA 2022;327:1666–78. https://doi.org/10.1001/jama.2022.4726) Eight hundred and fifty pregnant women (454 with CH, 396 with GH) were enrolled into the BUMP2 trial: 430 were randomly allocated to BP self-monitoring (primary outcome available on 416 (96.7%) women) and 420 women to usual care (primary outcome available on 405 (96.4%) women). There was no evidence of difference in the mean systolic BP in those allocated to BP self-monitoring, in either the CH cohort [mean standard deviation (SD) systolic BP: 133.8 (10.3) mmHg in the self-monitoring group compared to 133.6 (11.1) mmHg in those with usual care (adjusted mean difference 0.03; 95% CI –1.73 to 1.79)] or the GH cohort [mean (SD) systolic BP: 137.6 (12.1) mmHg compared to 137.2 (10.8) mmHg in those with usual care (adjusted mean difference –0.03; 95% CI –2.29 to 2.24)]. 3.3 Qualitative process evaluation (Chisholm A, Tucker KL, Crawford C, Green M, Greenfield S, Hodgkinson J, et al. Self-monitoring blood pressure in pregnancy: evaluation of health professional experiences of the BUMP trials. BMC Pregnancy Childbirth 2024;35:88–95.https://doi.org/10.1016/j.preghy.2024.01.134) The majority of trial participants interviewed had positive experiences of self-monitoring, reporting it was reassuring, acceptable, convenient and sometimes led to the earlier detection of hypertension. Having their own series of BP readings could feel empowering but also introduced some uncertainty and new responsibility. Some women described delayed or selective reporting of high BP readings. Some women preferred not to self-monitor due to concerns about anxiety, fears of preoccupation with monitoring, low perceived risk of hypertension, or choosing to have an HCP present for BP measurement. Women’s accounts demonstrated that HCP engagement with BP self-monitoring varied. Healthcare professionals largely trusted home readings from the validated monitors used. Most said such measurements positively affected their clinical encounters and professional roles, amplifying the information on which to base decisions and enriching their relationships with women. Some felt SMBP gave women new responsibilities that required additional support from HCPs. 3.4 BUMP within-trial economic analyses(Campbell HE, Chappell LC, McManus RJ, Tucker KL, Crawford C, Green M, Rivero-Arias O. Detection and control of pregnancy hypertension using self-monitoring of blood pressure with automated telemonitoring: cost analyses of the BUMP randomized trials. Hypertension 2024;81:887–96. https://doi.org/10.1161/HYPERTENSIONAHA.123.22059) In BUMP1 and BUMP2, there were no significant differences between trial arms in EuroQol EQ-5D-5L scores at any time points. In both analyses, healthcare contacts and costs were also similar across resource use categories in each trial arm. In BUMP1, mean (standard error) total healthcare costs with SMBP and with usual care were £7200 (£323) and £7063 (£245) respectively, mean difference (95% CI), £151 (–£633 to £936). For the BUMP 2 chronic hypertension cohort, corresponding figures were £13,384 (£1230), £12,614 (£1081), and £323 (–£2904 to £3549) and for the gestational hypertension cohort were £11,456 (£901), £11,145 (£959), and £41 (–£2486 to £2567). WS4: Self-monitoring of urinary protein in hypertensive pregnancy WS4.1 UDIP (Jakubowski BE, Stevens R, Wilson H, Lavallee L, Brittain L, Crawford C, et al. Cross-sectional diagnostic accuracy study of self-testing for proteinuria during hypertensive pregnancies: the UDIP study. BJOG 2022;129:2142–8. https://doi.org/10.1111/1471-0528.17180) Hypertensive pregnant women were recruited: 335/345 (97%) had sufficient data to be included in the analysis of whom 118 (35.2%) had a positive PCR. Self-testing had a sensitivity of 0.71 [95% CI 0.62 to 0.79] and a specificity of 0.89 [95% CI 0.84 to 0.92] compared to PCR. Sensitivity and specificity of testing by HCPs and the colorimetric reader were similar: sensitivity 0.73 (95% CI 0.64 to 0.80) and 0.78 (95% CI 0.69 to 0.85) respectively; specificity 0.88 (95% CI 0.82 to 0.92) and 0.83 (95% CI 0.78 to 0.88) respectively. WS4.2 UDIP qualitative Associated qualitative work found that self-testing was acceptable to pregnant women and HCPs, and could provide an opportunity for pregnant women to be more involved in their care. WS5: Modelling of the potential long-term costs and consequences The model frameworks developed provide a facility for exploration of the potential cost-effectiveness of future SMBP-guided interventions for different cohorts of women affected by pregnancy hypertension and for differing levels of intervention effectiveness. We present no definitive cost-effectiveness results, instead running a series of hypothetical scenarios to illustrate the capabilities of the models. For example, simulating a hypothetical scenario in which a new SMBP-guided intervention could reduce the risk of a pregnant women developing pre-eclampsia by 10%, the modelling suggested that long-term cost-effectiveness could potentially vary between hypertensive pregnant women and women at risk of pregnancy hypertension. This is because hypertensive women face a greater likelihood of developing associated complications both during and following pregnancy, in turn suggesting a greater absolute level of benefit from a 10% reduction in complications via the mechanism of better BP control. Such hypotheses would of course require empirical testing in practice. Conclusions WS1: Development The BUMP intervention was developed and user tested for implementation in higher risk and hypertensive pregnant women in the BUMP trials. WS2: Blood pressure self-monitoring during pregnancy for antihypertensives titration – pilot trial This randomised feasibility trial of BP self-monitoring during hypertensive pregnancy indicated that a large RCT would be acceptable and feasible. WS3: Self-monitoring to improve the detection and management of raised blood pressure in pregnancy WS3.1: BUMP survey Healthcare professionals should be aware that many women are choosing to self-monitor their BP. They are advised to enquire about this proactively and consider providing information on BP monitoring in pregnancy. WS3.2.1: BUMP1 trial Self-monitoring of BP during higher risk pregnancy appears to be safe. However, it did not improve the detection of hypertension when used alongside usual care. Self-monitoring did provide prior notice of hypertension in many women suggesting it could be useful. Not all women will want to self-monitor meaning that an individualised approach will be needed. Further work is needed to assess the place of SMBP, for example, in remote consultations or alongside self-management. WS3.2.2: BUMP2 trial Blood pressure self-monitoring in pregnancy hypertension was not associated with a change in BP control, as assessed by clinic systolic BP, but appears to be safe, without evidence of harm or unintended deleterious effects on pregnancy outcomes. Not all women will want to self-monitor meaning that an individualised approach will be needed. Furthermore, the addition of further components of self-management may be required in order to achieve improvements in BP control and other pregnancy outcomes. WS3.3: Process evaluation The majority of women and HCPs involved in the trial found SMBP enhanced their experiences of the clinical encounter and the HCP–woman relationship. However, not all women found self-care helpful. Furthermore, selective or delayed reporting of raised readings, along with the pursuit of normal readings, and HCPs’ variable engagement with home readings could have impacted the performance of the BUMP intervention. SMBP by women in the usual care arm may have affected evidence for the intervention’s effectiveness in identifying or managing hypertension. WS3.4: Economic analysis The SMBP intervention as evaluated in the BUMP trials was not associated with changes in health-related quality of life or healthcare resource use or with, pregnancy hypertension. When coupled with findings that SMBP is both safe and acceptable to individuals, this is reassuring for healthcare providers upon whom the recent coronavirus

Background When care homes close, it can be detrimental to older people’s well-being. However, there is little formal evidence to guide services when undertaking such important work. Objectives This study explores what happens when homes close, how best to minimise negative outcomes for older people and families, and key lessons for councils as they manage future closures. Methods Background literature review, national survey of Director(s) of Adult Social Services and analysis of national Care Quality Commission data. Interviews with older people, families, care staff, social workers and broader managers/partners in four case study sites, together with outcomes data (EQ-5D, ICEpop CAPability measure for Older people and outcomes from the literature on what older people value about care services) at initial assessment, 28 days’ review and 1 year. Survey of care staff (Professional Quality of Life survey) before and after closures, supplemented with individual interviews; interviews with commissioners and service providers. Preliminary model-based economic evaluation comparing the costs and consequences of care home closures. Distillation of key messages into a national policy guide, an accessible guide for older people/families and a guide/free training video for care staff. Findings People assume that a care home is a home for life – but this is sadly not always the case. Care homes can close for all kinds of reasons (e.g. to do with funding, buildings, care quality, etc., but also due to broader factors beyond the control of the care sector). There can be significant differences between emergency closures (which can be especially traumatic) and more planned closures (when at least in principle there is more time to work at the pace of individual residents). Local policies vary significantly, and there may be scope for a more evidence-informed, consistent approach. Closures can be extremely difficult for everyone involved – and should perhaps only ever be a last resort. Closures are particularly traumatic for residents, who are losing their home and valued relationships. For some people, this may be similar to a bereavement. The needs of care staff – in terms of meaningful information, emotional and financial well-being and employment support – are often overlooked. This leaves people unsupported and might also reduce their ability to support others. Councils seeking to shape local care markets often lack the levers to be able to do this effectively. Care home closures can create financial pressures on the public sector, costing around £1500 per resident, while families and staff may face varying expenses, such as top-up fees and travel costs. Unplanned and emergency closures are slightly more expensive, and result in poorer outcomes for residents. We have limited long-term data, but it may sometimes be possible to manage closures in a way that minimises negative outcomes for some (especially if existing services were less than optimal and where closures are well planned). Limitations Collecting data from older people, families and staff during care home closures, and in a challenging policy context, is complex, and the amount of data it is possible to collect in such circumstances is inevitably limited in a number of ways. In reflecting on this, we nonetheless draw attention to: The novel nature of the research, filling key gaps in knowledge around such a significant topic. The diverse and multifaceted perspectives which only a programme of research could hope to include. The importance of our policy and practice materials, given the significance of the issues at stake and the lack of previous evidence on which to draw. Conclusions Care home closures can happen for many different reasons, and are always a logical possibility in a ‘care market’, which seeks to use choice and competition to keep costs down and promote quality. This study has identified a series of practical lessons and experiences shared by participants which might help others in future – made available to different audiences via a series of policy, practice and training materials. However, none of this should, in any way, minimise the distress experienced by residents, which can be very significant and may well be long lasting. Moving beyond the specific focus of this research, insights from interviews with local commissioners and providers may have broader implications for the extent to which local authorities have sufficient powers and practical tools to be able to deliver on their ‘market shaping’ responsibilities under the Care Act. Future research on care home closures could usefully focus on: The needs/experiences of people who may have additional or specific needs that might otherwise be overlooked, such as people living with dementia and people from minority ethnic communities or different faith groups. The knock-on effect that care home closures may have on partner agencies. More effective ways of managing emergency closures, and how best to support people after the closure with the trauma they may have experienced. Larger sample sizes might also generate additional insights around individual outcomes and around the implications of particular resident/staff characteristics or types of closure – but this may need to be balanced against the difficulty of conducting such research and the cost to funders of even larger studies. Study registration This study is registered as IRAS project ID: 297258. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: NIHR201585) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 2. See the NIHR Funding and Awards website for further award information. Plain language summary We all want to be cared for with dignity when we are older, and we want the same for our families. Being looked after in a care home is expensive, and we have high expectations of the care that is given. But sometimes care homes close. This might be for a number of reasons, including financial problems, the state of the building or poor-quality care. This can affect the well-being of people in care homes, their families and the people who work there. Despite this, there is very little research to guide this important process. If care homes have to close, we want this to be well managed, so that older people are supported, families are reassured and care staff are helped to find new work and stay in care roles. To help with this, we: Asked Director(s) of Adult Social Services to tell us about what is happening across England and how they support older people at such potentially stressful times. Looked at some of the reasons why care homes might close from national data collected by an organisation called the Care Quality Commission. Worked in four different areas in England where homes are closing to ask older people, families, care staff and social workers about their experiences and how things could be improved. Looked at what impact closure has on older people’s health and well-being. Explored what happens to low-paid care staff after closures and the impact this has on them and the care they can give to others. Explored the cost implications for residents, family members, staff and wider society. We found that: People think that a care home is a home for life. Sadly, this is not always the case. Care homes can close for lots of different reasons. Planned closures can make it easier to support people well. Emergency closures, in particular, can be really upsetting. Local policies can be very different – perhaps there could be a more consistent national approach. Closures can be extremely difficult for everyone. They should only be a last resort. Closures are particularly traumatic for residents. For some people, this might be like a bereavement. The needs of care staff are often overlooked. It is difficult for councils to influence which homes open and close in their area. This might have broader implications for the role of councils beyond this particular topic. Closing a home costs councils money (about £1500 per resident). There can also be costs for residents and families. Even though moving to a new home can be very upsetting, some people settle and do well. We have written some guides to share what we have learnt with others. We have tried to focus on the things that people can do to make things better for residents – not on things they cannot change. There is a guide for older people and families being shared by a national charity called Age UK. Scientific summary Background We all want to be cared for with dignity when we are older, and we want the same for our families. Being looked after in a care home is expensive, and we have high expectations of the care that is given. Despite this, many care homes close every year, whether through an emergency (such as a fire/flood), councils making strategic choices to develop new service models, the cost of maintaining a dilapidated building, a private provider selling up/going bankrupt or a regulatory intervention following the discovery of poor care. In an era of austerity, care markets are increasingly fragile, and the very logic of a ‘market’ implies that the risk of failure has to be real for there to be sufficient incentives to deliver appropriate care at the right price. When care homes close, the received wisdom is that relocation can be detrimental to health and well-being. Despite this, there is little formal evidence to guide closure processes, with councils constantly ‘reinventing the wheel’. Objectives This study builds on a previous pilot in Birmingham in order to explore what happens to older people and care staff when care homes close, how best to manage closures in a way that minimises distress and negative outcomes for older people and families, and key lessons for councils as they seek to manage future closures. Our research questions are: What is the pattern of care home closures nationally? How are they undertaken in different councils, and what do councils consider to be best practice when supporting older people at such potentially stressful times? How do older people experience closures, what impact does closure have on health and quality of life, and how can any negative impacts be reduced? What impact do closures have on care staff and local care markets, and how can negative impacts be reduced? What are the costs and consequences of closures and the key data required to make this estimation? Can we develop a modelling framework to drive appropriate data collection for future home closure prediction to mitigate adverse outcomes? How can future closures be planned and conducted in a more evidence-based manner, so that outcomes for older people are improved and negative impacts reduced? Methods Work package 1 aimed to establish the pattern of care home closures nationally, how they are undertaken in different councils and what might constitute best practice when supporting older people. We carried out a cross-sectional survey targeted at social care leaders and an analysis of care home closure policy documents. All Director(s) of Adult Social Services in 152 councils in England were invited to participate. The survey explored leaders’ experiences of closures, policies, processes, perceived outcomes, challenges and any local evaluations. Participants were invited to share local care home closure policies for analysis. Descriptive statistics were analysed for quantitative survey data. Qualitative survey data and policy documents were analysed thematically. We also analysed routinely collected data from the Care Quality Commission (CQC) to understand the factors that influence care home closures. This combined individual care home characteristics from the CQC with information on council areas, taken from various sources, and data from the Social Care Collection. The CQC data provide us with information on 25,459 care homes between 2010 and 2021. These data are at the care home level and provide information on the type of care home (nursing/residential), the focus of the care home (older people, dementia, disability), size (measured by the number of beds), performance rating and indicators of whether the care home is located in an urban or rural area. The CQC data also provide us with the care home’s geographical location. This information allowed us to link data on other local characteristics (such as the proportion of the population who are over 65, deprivation, residential land prices and care home market competition). We also included variables collected as part of the Social Care Collection. This includes the proportion of self-funders, average daily rates of delayed transfers of care from hospital, total expenditure, satisfaction scores taken from the adult social care survey, proportion of individuals over 65 receiving long-term support and funds attached to carers support. The research used a multilevel logit regression model to study whether care homes close or stay open (a yes or no outcome). By using this approach, the study can account for differences between care homes within the same council, leading to more accurate results about how care home features, local factors, and social care spending affect the chances of a care home closing. Work package 2 explored the experiences of older people, families, care staff, social workers and broader stakeholders/local partners. We worked in four case study councils where homes were closing, including a mix of different locations across England, a mix of urban/rural settings and a mix in terms of socioeconomic/demographic factors. In each site, we undertook semi-structured qualitative interviews with 10–15 stakeholders/broader partners who had key roles in/perspectives on care home closures to explore the rationale for closure, key drivers, anticipated outcomes and impact on other services. We also undertook semi-structured interviews with 96 participants (older people, families, care staff and social workers) to explore the impact of closures, information/support provided, views on the process and areas for improvement. Interviews with older people took place in person with researchers based in the closing care homes (where COVID restrictions permitted). Where people were unable to consent to take part, a consultee was identified under the Mental Capacity Act. Interviews with families, care staff and assessors took place either in person or online, guided by their preference. We collected outcomes data for older people at initial review (before the closure), an early review (often 28 days after relocating) and at a longer-term review (1 year), before, during and after relocation. Twenty-two participants completed up to three paper-based questionnaires: EQ-5D, three-level version, ICEpop CAPability measure for Older people (ICECAP-O) and a 15-question, Likert scale questionnaire, based on a national review of the literature on what older people value about care services. Interviews were transcribed and analysed using the framework approach to identify key themes. Questionnaires were coded and analysed using Stata® (StataCorp LP, College Station, TX, USA). Work package 3 considered the impact of care home closures on care staff. Where access permitted, all care staff in closing homes in our four case study sites were invited to take part in interviews and to complete the Professional Quality of Life survey during and 6 months after closure. The interviews explored how care staff experienced the closure of the care home where they worked, how they prepared for closure, the impact upon themselves and current/future employment, and their insights into the closure process. All employed staff at a care home were given the opportunity to take part, including kitchen staff, cleaners and handypeople, as well as people providing direct care and those working in co-ordinating or care management capacities. In each of our four case study sites, we carried out semi-structured interviews with commissioners and providers (seeking up to eight of each per site). Participants included those in senior roles involved in decision-making or closely connected to care home functioning, as well as those who were indirectly or directly impacted by a care home closing. Interviews explored the relationship between commissioners and providers, the nature and impact of local authority strategy, the responsibilities of care homes and the local care market (including approaches to trying to shape the local market, to ensure stability and to contingency planning). Work package 4 focused on analysing the economic impacts of care home closures on residents, their families, care home staff and councils. Pathway costing employed both quantitative and qualitative methods, organised into three stages. Stage 1 involved analysing stakeholder interviews (from WP2) to identify economic themes related to care home closures. Stage 2 utilised these insights to develop closure pathways for stakeholders, including residents, families and care staff. Stage 3 conducted a full cost analysis to estimate closure costs, including using sensitivity analysis for different stakeholders. This stage involved identifying and quantifying resources needed at each pathway stage. Economic modelling focused on developing a decision tree model to analyse care home closures under three scenarios: planned, unplanned and emergency. This model was informed by interview data from WP2, specifically coded for WP4. The pathways developed in the previous pathway costing served as a foundation and were adapted to reflect the complexities of unplanned and emergency closures. Cost and resource use data were derived from the previous cost analysis, supplemented by literature and expert opinion, especially for unplanned and emergency scenarios. Probabilities within the model were informed by observations from case study sites, literature reviews, qualitative interviews and expert opinions. The analysis was conducted from a public sector perspective using a 12-month decision tree model. This model reports the cost per capability improvement by comparing unplanned and emergency scenarios to a reference case of planned closures. Outcomes were measured using ICECAP-O and EQ-5D data collected in WP2. In work package 5, we summarised findings in a national policy guide sent to all social care leaders in England, a free training video, a guide for older people/families and a guide for care staff. Results Key findings were that: People assume that a care home is a home for life – but this is sadly not always the case. Care homes can close for all kinds of reasons (e.g. to do with funding, buildings, care quality, etc., but also due to broader factors beyond the control of the care sector). There can be significant differences between emergency closures (which can be especially traumatic) and more planned closures (when at least in principle there is more time to work at the pace of individual residents). Local policies vary significantly, with scope for a more evidence-informed, consistent approach. Closures can be extremely difficult for everyone involved – and should perhaps only ever be a last resort. Closures are particularly traumatic for residents, who are losing their home and valued relationships. For some people, this may be similar to a bereavement. The needs of care staff – in terms of meaningful information, emotional and financial well-being and employment support – are often overlooked. This leaves people unsupported and might also reduce their ability to support others. Councils seeking to shape local care markets often lack the levers to be able to do this effectively. Care home closures can create financial pressures on the public sector, costing around £1500 per resident, while families and staff may face varying expenses, such as top-up fees and travel costs. Unplanned and emergency closures are slightly more expensive, and result in poorer outcomes for residents. We have limited long-term data, but it may sometimes be possible to manage closures in a way that minimises negative outcomes for some (especially if existing services were less than optimal and where closures are well planned). Conclusions Overall, we conclude that: Many people assume that moving into a care home means a home for life – but this is sadly not always the case. Homes can close for all kinds of reasons. This can be due to issues around funding, business decisions and/or the quality of care delivered – but can also be influenced by broader factors (such as changes in land values influencing people’s decisions about whether or not to exit the market). There can be very significant differences between emergency closures (which can be especially difficult and traumatic) and more planned closures (when at least in principle there is more time to plan, communicate and work at the pace of individual residents). There has previously been limited evidence to guide decision-makers and practitioners seeking to support and relocate older people. Local policies and protocols vary significantly, and there may be scope for a more evidence-informed, consistent approach (e.g. via a national or regional template). Closures can be extremely difficult and traumatic for everyone involved, from social care leaders and staff, to service providers and care staff, to older people and families – and should perhaps only ever be seen as a last resort. Closures are particularly traumatic for older people, who may be understandably angry, distressed and disorientated. People are losing their home and valued relationships with residents and staff, and the process may well be similar to that of a bereavement. The needs of care staff – in terms of meaningful information, practical details, emotional well-being and employment support – can often be overlooked. This should be a key focus in future, both to better support staff, and because care staff are so significant in terms of supporting older people. Local authorities have a duty to shape their local care markets, but often lack the levers, powers and tools to be able to do this effectively. Too often, they feel as if the ‘wrong’ homes are opening and closing in their areas. Care home closures can create financial pressures on the public sector, costing around £1500 per resident, while families and staff may face varying expenses, such as top-up fees and travel costs. Our findings indicate that unplanned and emergency closures are slightly more expensive and result in poorer outcomes for residents. Well-planned and supportive closures can potentially improve the health-related quality of life of residents over the longer-term period of a year (especially if people are unhappy with care in the original home) but are associated with a brief initial negative impact. Findings highlight the importance of careful preparation, planning and support to minimise disruption and improve overall outcomes. Recommendations for policy, practice and research Above all, there are a series of practical lessons and experiences shared by participants which might help others in future – made available to different audiences via a series of policy, practice and training materials. However, none of this should in any way minimise the distress experienced by residents, which can be very significant and may well be long lasting. We will also seek to work with regulators to learn lessons from one of our emergency closures, which happened at very short notice (hours) and was particularly traumatic for everyone involved. Moving beyond the specific focus of this research, insights from interviews with local commissioners and providers may have broader implications for the extent to which local authorities have sufficient powers and practical tools to be able to deliver on their ‘market shaping’ responsibilities under the Care Act. Future care home closures research could usefully focus on: The needs/experiences of people who may have additional or specific needs during closures that might otherwise be overlooked, such as people living with dementia and people from minority ethnic communities or different faith groups. The knock-on effect that care home closures may have on partner agencies. More effective ways of managing emergency closures and how best to support people after the closure with the trauma they may have experienced. Larger sample sizes might generate additional insights around individual outcomes and the implications of particular resident/staff characteristics or types of closure, but this may need to be balanced against the difficulty of conducting such research/the cost of even larger studies. Study registration This study is registered as IRAS project ID: 297258. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research Programme (NIHR award ref: NIHR201585) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 2. See the NIHR Funding and Awards website for further award information.

Background Dementia represents a major public health challenge worldwide. Interventions are required to maintain functional ability and prevent crises. Exercise-based therapy may improve activities of daily living and prevent falls. Objective We aimed to systematically develop an intervention, called Promoting Activity, Independence and Stability in Early Dementia and mild cognitive impairment (PrAISED), and evaluate its clinical and cost-effectiveness. Design A mixed-methods study. Literature review, co-design, patient and public involvement and practical experience were used to develop and refine a therapy intervention. We applied principles of behaviour change to promote engagement and motivation. Three-arm feasibility randomised controlled trial to test research procedures, the practicality of intervention, and establish the need for prolonged supervision. Two-arm, multicentre randomised controlled trial, comparing intervention with control, with 15 months’ follow-up. Process and realist evaluations, including quantitative data, interviews and thematic analyses. Health economic evaluations, including a cost-effectiveness analysis using Markov modelling and social return on investment. Implementation study. Setting Therapy and research were undertaken in participants’ homes and local communities across 5 sites in England. Participants People with diagnosed mild dementia or mild cognitive impairment, Montreal Cognitive Assessment score 13–25, living at home and a family member or carer. In the main trial, there were five sites in England, participants were 98% White ethnicity and 20% had mild cognitive impairment. Interventions A specially designed, dementia-specific, rehabilitation programme focusing on strength, balance, physical activity and performance of activities of daily living, which was tailored, progressive and addressed risk, providing up to 50 therapy sessions over 12 months. The control group received usual care plus a falls risk assessment. Main outcome measures The primary trial outcome was the informant-reported Disability Assessment for Dementia 12 months after randomisation. Secondary outcomes were: self-reported activities of daily living, physical activity, quality of life, frailty, balance, functional mobility, cognition, fear of falling, mood, carer strain and service use (at 12 months) and falls (between months 4 and 15). Results Three hundred and sixty-five people were randomised in the multicentre trial, 183 to intervention and 182 to control. Median age of participants was 80 years (range 65–95), median Montreal Cognitive Assessment score was 20/30 (range 13–26) and 58% were men. Participants received a median of 31 (interquartile range = 22–40) therapy sessions out of a possible maximum of 50. Participants reported completing a mean 121 minutes/week of PrAISED activity. Primary outcome data were available for 149 (intervention) and 141 (control) participants. There was no difference in Disability Assessment for Dementia scores between groups: adjusted mean difference −1.3/100, 95% confidence interval (−5.2 to 2.6); Cohen’s d effect size −0.06 (−0.26 to 0.15); p = 0.5. Upper 95% confidence intervals excluded small to moderate effects on any secondary outcome measures. Between months 4 and 15, there were 79 falls in the intervention group and 200 falls in the control group, and adjusted incidence rate ratio was 0.78 (0.5 to 1.3); p = 0.3. Participants and therapists liked the intervention and thought that it produced health benefits. Tailoring, perceiving benefits, professional supervision and family or carer support were important facilitators. Cognitive and physical impairment, risk aversion and tapering the level of support were barriers. Therapeutic relationship between participant and therapist was important. The cost per quality-adjusted life-year gained was £130,000 over a lifetime horizon. Social return on investment was positive before the onset of the COVID-19 pandemic but was negative after the first pandemic lockdown, largely due to unavailability of access to community facilities. Limitations The multicentre randomised controlled trial was disrupted by the COVID-19 pandemic. The first lockdown occurred when 301 participants had been randomised and 64 participants had completed the trial. Recruitment was suspended, and some therapy and data collection were undertaken remotely. The intervention was diminished compared with in-person delivery, but reported fidelity remained reasonable. Our participant population lacked socioeconomic and ethnic diversity – over 30% lived in the least deprived decile of postcodes. The intervention was very popular with participants and therapists, and it is possible that our predominantly biomedical and functionally orientated outcome variables failed to capture intervention benefits. Conclusions The intensive PrAISED programme of exercise and functional activity training did not improve activities of daily living, physical activity, quality of life, reduce falls or improve any other secondary health status outcomes. Due to the pandemic, the population recruited, and the outcomes chosen, some uncertainty remains about the effectiveness of the intervention. Future work Consider: (1) repeating the trial outside of a pandemic; (2) more psychosocial outcomes, such as social participation, affirming personhood and valuing therapeutic relationships; (3) alternative approaches to risk reduction and ability maintenance in dementia; (4) other models of support to manage problems associated with inevitable progression and decline; (5) that conventional randomised controlled trials may not be the best way to evaluate complex intervention for complex and degenerative conditions; interpretative and realist methods should supplement evaluation; and (6) work to include a wider range of ethnicities and socioeconomic circumstances. Study registration This study is registered as ISRCTN10550694, 15320670. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research programme (NIHR award ref: RP-PG-0614-20007) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 1. See the NIHR Funding and Awards website for further award information. Plain language summary Dementia causes deterioration in memory and thinking abilities. The Promoting Activity, Independence and Stability in Early Dementia (PrAISED) programme aimed to develop and test a physical exercise and activity intervention to improve the ability to do daily activities among older people in the early stages of dementia. We developed a therapy programme specifically designed for people with dementia. We paid particular attention to encouraging participation. Therapy was tailored to participants’ goals, preferences and abilities. We confirmed that we could deliver the intervention and do the research to test it in a small-scale feasibility study. We tested PrAISED by recruiting 365 people with dementia and a family member from five English counties. We randomly assigned them to receive PrAISED therapy or to a control group, who were given advice on falls prevention. The PrAISED group received up to 50 therapy sessions, delivered by trained therapists, and were also encouraged to do exercises on their own. At the start and after 12 months, we measured ability to do everyday activities and other aspects of health, including falls, quality of life, activity and National Health Service and social care use. We did interviews and observations to explain the findings. Those receiving PrAISED therapy did no better on any of our measurements than those in the control group. The therapy programme was popular, and participants described benefits to their lives. Professional supervision and family support were important. However, memory and physical health problems often prevented full participation. The study was disrupted by the COVID-19 pandemic. An economic study showed that PrAISED was not cost-effective. A method which values social outcomes suggested that PrAISED gave a good return before the pandemic but not during it. We conclude that it might be more appropriate to help people manage problems associated with the inevitable decline seen in dementia rather than to try to change the course of the disease. Scientific summary Background The prevalence of dementia is increasing with the ageing population and is expected to double in the next 30 years. About 950,000 people live with dementia in the UK. Dementia causes progressive deterioration in a person’s cognitive and functional abilities. People with dementia are often dependent on other people. Dementia results in high levels of demand on health and social care as well as family and other informal carers. We need therapeutic interventions to reduce the decline in functional abilities so people with dementia can remain independent for longer. Exercise-based activities and functional rehabilitation may improve people with dementia’s activities of daily living (ADL). Objectives The Promoting Activity, Independence and Stability in Early Dementia and mild cognitive impairment (PrAISED) programme aimed to develop and evaluate an exercise and activity intervention to increase independence in ADL for older people living with dementia or mild cognitive impairment (MCI). The programme comprised seven work packages (WPs): WP1 – intervention development: develop, manualise and support the delivery of an evidence-based multicomponent therapy intervention. WP2 – adherence and motivation: develop strategies to support engagement with the intervention and achieve long-term adherence. WP3 – feasibility study: test the feasibility and practicality of delivering the intervention and conducting a randomised controlled trial (RCT). WP4 – process evaluation: conduct process and realist evaluations of the trials. WP5 – multicentre RCT: establish the clinical effectiveness of the PrAISED intervention in a multicentre RCT. WP6 – health economics: establish the cost-effectiveness of the intervention and social return on investment (SROI). WP7 – implementation: understand factors that would affect implementation of the intervention in practice. Methods Work package 1 – intervention development The PrAISED intervention was developed by a team of clinical academics, practitioners and patient and public involvement and engagement representatives using evidence and theory from systematic reviews, interviews, focus groups, empirical studies and expert opinion. This followed work in an Alzheimer’s Society PhD fellowship (Dr Vicky Booth) and a National Institute for Health and Care Research (NIHR) Programme Development Grant. We used theory about how to motivate people with dementia to do exercises. The intervention was described in a manual. We developed practitioner training courses. These were refined following experience in the feasibility study and described using the Template for Intervention Description and Replication (TIDieR) checklist. Work package 2 – adherence and motivation We initially used self-determination theory (SDT) to inform intervention development but later developed a new dementia-specific behaviour change model (PHYT-in-dementia), derived from literature reviews, synthesis and empirical evidence from the feasibility study. It was validated using interview data collected during the RCT process evaluation. PHYT-in-dementia identified factors that mediate behaviour change and maintenance in people living with dementia. These were: characteristics of the person with dementia, support, expectations, goals, carer characteristics, progress, social opportunity, self-efficacy, capability, intervention characteristics, autonomy, control, physical infrastructure, personal history, information, knowledge, characteristics of therapists and personal beliefs. Work package 3 – feasibility study We conducted a three-arm randomised feasibility trial to establish that we could recruit and randomise participants at a sufficient rate, deliver the intervention in participants’ homes across two sites, retain and follow up participants; that the intervention was practical and safe; that we could collect trial data and that our sample size assumptions were reasonable. We explored the level of supervision participants would need to undertake 3 hours of PrAISED exercises and activities a week and sustain this over the duration of the trial. We compared the PrAISED intervention with supervision over 12 months to a shorter intervention comprising nine therapy visits and three telephone calls delivered over 12 weeks and a control group who received a falls prevention assessment and advice. Data were collected at baseline and 12-month follow-up, during face-to-face interviews with two researchers. Health status measures comprised disability in ADL [Disability Assessment for Dementia scale (DAD)], habitual physical activity, quality of life (QoL), frailty, cognition, other intermediate outcomes and carer outcomes. Monthly calendars were completed for falls and activity ascertainment. Work package 4 – process evaluation We investigated implementation of the PrAISED intervention during the trials, the mechanisms of impact and context. We adopted a mixed-methods approach investigating fidelity, adaptations, dose and reach, including quantitative data, interviews and thematic analyses. Mechanisms of impact and context were identified through semistructured qualitative interview with a sample of therapists, participants and carers. Interviews were conducted 6 and 12 months into involvement in PrAISED. Interviews were conducted remotely during the COVID-19 lockdown between May 2020 and September 2020. Work package 5 – multicentre randomised controlled trial Participants were recruited from five sites in England via secondary care memory assessment clinics, general practice registers, dementia support groups and the NIHR Join Dementia Research register. Participants were recruited as patient–carer dyads. The RCT was conducted between September 2018 and June 2022. This included the COVID-19 pandemic period, which impacted recruitment, intervention delivery and data collection. Between March 2020 and September 2020, research and intervention contacts were delivered remotely. We included participants aged over 65 years, with a diagnosis of dementia or MCI, a Montreal Cognitive Assessment (MoCA) score of 13–25 (out of 30), a family member or unpaid carer who knew the participant well and who was willing to participate. Participants had to have mental capacity to consent and be willing to take part in an exercise intervention. Separate consent was taken for the carer. Active intervention comprised a specially designed, dementia-specific, rehabilitation programme focusing on strength, balance, physical activity and performance of ADL, which was tailored, progressive and addressed risk, providing up to 50 therapy sessions over 12 months. The control group received usual care plus a falls risk assessment. The primary outcome was ADL, measured at 12 months by the informant-completed DAD scale. Secondary outcomes included self-assessed ADL (Nottingham Extended ADL scale); cognition (MoCA, animal-naming verbal fluency; Cambridge Neuropsychological Test Automated Battery), balance (Berg Balance Scale); mobility and ability in divided attention [Timed Up and Go (TUG), dual-task (TUG)]; hand grip strength; health and social care resource use for patient and carer Client Service Receipt Inventory; fear of falling; frailty, mood; carer strain, carer and self-assessed health-related quality of life (EQ5D DemQoL scales), physical activity; step count by accelerometer; and apathy. Participants were followed up after 12 months. Between months 1 and 15, self-completed calendars were used to record falls and PrAISED exercise undertaken. A brief postal follow-up questionnaire was completed by the patient’s carer/informant after 6 months. A sample of 368 participants (184 per group), with 23% attrition, had 80% statistical power to detect a change in disability outcome (DAD), with a moderate effect size of 0.5. A secure internet-based system based in a Clinical Trials Unit was used to randomise individuals, 1 : 1, stratified by site, presence of a co-resident and history of previous falls. Blinding of participants and therapists was not possible due to the nature of the intervention. Analysis was conducted blind. An analysis of covariance was conducted for the primary outcome (DAD) at the 12-month follow-up, using group, stratification variables and baseline DAD score as covariates. The analysis was conducted on an intention-to-treat basis. Scaled secondary outcome measures were analysed similarly. Adjusted mean differences, Cohen’s d standardised effect size, 95% confidence intervals (CIs) and p-values were reported. Work package 6 – health economics Cost–utility analysis using a Markov-modelled projection over a 15-year time frame, and a SROI analysis. Work package 7 – implementation We undertook four small-scale implementation studies, using the Consolidated Framework for Implementation Research. We investigated adaptation and adoption of a pilot service in routine practice at one site. We interviewed therapists who delivered the PrAISED intervention, commissioners and service leaders. We explored lack of participation by ethnic minority populations through discussions with community groups and leaders. We developed advice on compiling a business case for commissioning the intervention. Results We developed and refined the PrAISED intervention. This comprised a 12-month, home-based, individually tailored rehabilitation programme, focusing on strength, balance, physical activity and performance of ADL. Tailoring took account of individual history, personality and abilities, problems, interests, family and other resources. Fourteen core principles were defined to guide intervention delivery. A logic model was developed. Delivery was by physiotherapists, occupational therapists and rehabilitation support workers. Participants were encouraged to undertake a total of at least 180 minutes of exercise per week. The programme was progressed by therapists following periodic reassessments. Supervised sessions were tapered over the 12 months (twice-weekly visits in the first 3 months, reducing to monthly in the final 3 months) and community activities were signposted. An intervention manual was published. Therapists were supported throughout the intervention delivery period with training and regular clinical support sessions. We reviewed behaviour change frameworks for older people and people living with dementia. We adapted SDT to support the intervention, which posits the importance of autonomy, relatedness and competence, and 12 practical support approaches. Further reviews led to the development of a new behaviour change framework, PHYT-in-dementia, which was adapted, validated and applied to the intervention for the multicentre RCT. We undertook a two-site, three-arm feasibility RCT. We successfully recruited 60 participants, of whom 45 completed the intervention and provided outcome data. There were no serious, related adverse events (AEs). Missing data rates were satisfactory, apart from some scales that were investigating SDT. We made some other minor adjustments to eligibility criteria and outcome scales. We found that participants were unable to adhere to the programme in the absence of supervision and carried forward higher-intensity supervision but developed an algorithm to plan and gradually reduce intensity, considering ability to undertake activities independently. Analysis of outcomes supported the superiority of the higher-intensity programme and suggested moderate to large benefits in balance, gait speed and disability. We undertook a five-site, two-arm RCT, powered to detect a moderate effect size on the DAD scale. We recruited 365 participants, median age 80 years, 42% female, median MoCA score 20/30, predominantly from less-deprived localities. There were no significant differences in characteristics between groups at baseline. A median of 31 therapy sessions were delivered, interquartile range 22–40, 68% face to face. Fidelity judged from (pre pandemic) video-recorded therapy sessions was good. Intervention group participants reported undertaking an additional mean of 121 minutes of exercise per week. Two hundred and ninety (79%) were followed up. There were no significant differences on the primary outcome, the DAD: adjusted mean difference −1.3 (95% CI −5.2 to 2.6); standardised effect size (d) −0.06, 95% CI −0.26 to 0.15; p = 0.5; or on physical activity, balance, QoL, cognition or a range of other measures. There was a statistically significant small difference in favour of the control group, on the dual-task TUG test and on the self-report DemQoL scale. Upper 95% CIs excluded even small benefits on other scales. Rate of falling was reduced by 22%, but this was not statistically significant. Results did not change in a range of sensitivity analyses. Both service delivery and research were disrupted by the COVID-19 pandemic. Recruitment was delayed, some follow-up was undertaken remotely and some intervention sessions were delivered by telephone or video call, which were consequently much less ambitious than intended. Community facilities and activities became unavailable to vulnerable people. Results were no different for those completing the intervention before the COVID-19 pandemic. One hundred and sixty-seven AEs were recorded: 59 in control and 108 in the intervention groups, involving 68 participants. There were 91 serious adverse events: 29 in control and 62 in intervention, involving 60 participants. None was serious and related to intervention. There was no statistically significant difference between the intervention and control groups for AEs. The process evaluation studied implementation of the intervention, mechanisms of impact and context. Eighty-eight interviews were undertaken with participants, carers and staff. The PrAISED intervention was well received among participants and clinicians. Many gave examples of benefits gained as a result of taking part in PrAISED. However, cognitive impairment, physical comorbidity and fear of falls or getting lost prevented independent engagement. Tapered support was ineffective and acted as a barrier to continued engagement. Family members played a major role in supporting participation. A realist evaluation considered mechanisms behind the social benefits and concluded that participants improved social interactions when therapy activities were tailored to their preferences, when therapy support was maintained and when participants perceived improvements as a result of the intervention. The cost-effectiveness study showed a cost/quality-adjusted life-year of £130,000. SROI suggested benefits, but only in the feasibility and pre-COVID phases of the study. Participants completing the trial after the start of the pandemic had a negative social return, predominantly due to lack of availability of community facilities. We introduced PrAISED into routine practice in a socioeconomically deprived part of Nottingham. Eleven participants were referred and completed a shorter version of PrAISED, less than half what was anticipated or provided for. The intervention was well received by those who participated. We investigated reasons behind poor recruitment of participants from ethnic minority groups, finding feelings of mistrust towards health services and research, and stigma against dementia. We interviewed commissioners to provide guidelines on constructing a business case for implementing exercise and post-diagnostic support programmes. Conclusions We delivered an ambitious programme of research to address whether we can intervene to maintain safe activity and independence after a diagnosis of dementia. We systematically designed a new dementia-specific intervention, and used multiple methods to evaluate it, centred around a multicentre RCT. The intervention was about as intensive as it would be possible to deliver in the UK health and cultural context. Despite positive reception and perceived benefits by participants, we measured no benefits from the programme. Our RCT was significantly disrupted by the COVID-19 pandemic. This, a lack of diversity in the participant population and the fact that the outcome measures used might not have captured the impact of the intervention, leave persisting uncertainties about whether a PrAISED-like intervention might be beneficial. However, our findings suggest that a more ‘supportive’ approach to health care after a diagnosis of dementia may be appropriate, helping to manage problems associated with inevitable functional decline, rather than trying to change the course of disease or to maintain abilities. This would involve developing therapeutic relationships, providing support to live with limitations, minimising intervention burden, maintaining personhood, inclusion and occupation, providing psychological and emotional support and support to family and other carers. Study registration This study is registered as ISRCTN10550694, 15320670. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research programme (NIHR award ref: RP-PG-0612-20004) and is published in full in Programme Grants for Applied Research; Vol. 14, No. 1. See the NIHR Funding and Awards website for further award information.

Longitudinal observational studies suggest that patients who present early to secondary care with psoriatic arthritis have a better outcome. Outcome needs to be measured including domains important to patients. To identify disease activity and impact outcomes important to patients with psoriatic arthritis in comparison to existing ones (Patient-Reported Outcome Measurement Study). To assess the validity of new measures of disease activity in psoriatic arthritis derived from patient engagement (ASSESSment of modified composite disease measures study). To explore patients’ experiences of diagnosis in psoriatic arthritis? (Experiences of screening and diaGnosis from the perspective of pAtients with PSoriasis and psoriatic arthritis study). To identify modifiable risk factors for the development of psoriatic arthritis using the Clinical Practice Research Datalink (EPIdemiology of psoriatiC arthritis study). To compare the performance of screening tools for detecting undiagnosed psoriatic arthritis (COMparison of Psoriatic Arthritis scREening tools study). To investigate the clinical effectiveness (Total bUrDen Of psoRiasis trial) and cost-effectiveness (COSt of screening for Psoriatic Arthritis study) of detecting undiagnosed psoriatic arthritis. Mixed methods using data from patient focus groups, a multicentre cohort study, primary care health records (Clinical Practice Research Datalink) and a prospective randomised clinical trial (Total bUrDen Of psoRiasis trial, COMparison of Psoriatic Arthritis scREening tools study, COSt of screening for Psoriatic Arthritis study). Focus groups and cohort studies of 221 patients with psoriatic arthritis including newly diagnosed cases (Patient-Reported Outcome Measurement Study, ASSESSment of modified composite disease measures study, experiences of screening and diaGnosis from the perspective of pAtients with PSoriasis and psoriatic arthritis study) and 2225 patients with psoriasis from primary care not known to have psoriatic arthritis from 135 randomised general practices (COMparison of Psoriatic Arthritis scREening tools study, Total bUrDen Of psoRiasis trial randomised clinical trial, COSt of screening for Psoriatic Arthritis study). Randomised controlled trial comparing the early identification of psoriatic arthritis by annual rheumatological assessment (enhanced surveillance) in people with psoriasis identified in primary care (n = 1123) with standard care (n = 1102). Participants with suspected inflammatory arthritis were referred to the local rheumatology clinic for an assessment of psoriatic arthritis (enhanced surveillance arm: at baseline, 12 and 24 months; standard care arm: at 24 months). Cohort studies of incident psoriasis (n = 90,189) or psoriatic arthritis (n = 6783) patients from the Clinical Practice Research Datalink (EPIdemiology of psoriatiC arthritis study). Cost-effective analysis using data collected in Total bUrDen Of psoRiasis trial. Modelling exercise to extrapolate screening performance from Total bUrDen Of psoRiasis trial to long-term costs and quality-adjusted life-years (COSt of screening for Psoriatic Arthritis study). In Total bUrDen Of psoRiasis trial the primary outcome was the Health Assessment Questionnaire Disability Index at 24 months post registration in participants diagnosed with psoriatic arthritis. Patient-Reported Outcome Measurement Study: Patients rated pain and fatigue as the most important outcomes when receiving treatment for psoriatic arthritis. ASSESSment of modified composite disease measures study: Shortened versions of visual analogue scales performed well in detecting treatment change and responsiveness over 6 months. Experiences of screening and diaGnosis from the perspective of pAtients with PSoriasis and psoriatic arthritis study: Missed opportunities for psoriatic arthritis diagnosis has implications for well-being including mental health, and there is a need for provision of psychological support and self-management guidance. Screening for psoriatic arthritis was generally regarded as a positive experience. EPIdemiology of psoriatiC arthritis study: Bayesian networks identified clusters of symptoms that can accurately predict the development of psoriatic arthritis (area under the curve 0.73, 95% CI 0.70 to 0.75). Obesity is a modifiable lifestyle factor that increases the risk of developing psoriatic arthritis and diminishes linearly over a 10-year period with a reduction in body mass index. Diabetes, cardiovascular disease, uveitis and Crohn’s disease are associated with psoriatic arthritis and may appear early in the disease course. COMparison of Psoriatic Arthritis scREening tools study: The Psoriasis Epidemiology Screening Tool questionnaire remains the preferred screening tool, given its simplicity with no significant differences in performance (sensitivity: 0.625, specificity: 0.757, area under the curve 0.787) compared to COmparisoN of ThreE Screening Tools to detect psoriatic arthritis in patients with psoriasis. Total bUrDen Of psoRiasis trial: The primary analysis population consisted of 87 participants with a positive diagnosis of psoriatic arthritis: 64 in enhanced surveillance, 23 in standard care (overall mean Health Assessment Questionnaire Disability Index score at 24 months was 0.45). The adjusted odds ratio for achieving a Health Assessment Questionnaire Disability Index score of 0 at 24 months post registration in enhanced surveillance compared to standard care was 0.64 (95% CI 0.17 to 2.38), and the adjusted odds ratio of achieving a higher (non-zero) Health Assessment Questionnaire Disability Index score at 24 months post registration in enhanced surveillance relative to standard care arm was 1.12 (95% CI 0.67 to 1.86), indicating no evidence of a difference between the two treatment groups (p = 0.6612). COSt of screening for Psoriatic Arthritis study: The within-trial cost-effectiveness analysis suggests that enhanced surveillance may be associated with higher mean quality-adjusted life-years, although the difference was very small (+0.002, 95% CI –0.03 to 0.03), and lower mean costs (−£120.54, 95% CI –£540.57 to £288.99). Using the cost-effectiveness model based on the sensitivity, specificity and the psoriatic arthritis incidence from the Total bUrDen Of psoRiasis trial, Psoriasis Epidemiology Screening Tool is the most expensive and has the highest lifetime quality-adjusted life-years with an incremental cost-effectiveness ratio of £14,964 per additional quality-adjusted life-year compared to CONTEST, CONTESTjt and no screening using a lifetime time horizon. The Total bUrDen Of psoRiasis trial was underpowered due to a lower proportion of patients developing psoriatic arthritis than expected (87 compared to 148) and disruption to follow-up due to the coronavirus pandemic. COSt of screening for Psoriatic Arthritis study analyses were limited by the extent of missing data. The benefit of early detection of psoriatic arthritis in a primary care psoriasis population remains unproven. Surveillance for undiagnosed psoriatic arthritis generally detects mild cases as measured by physical function. Patients rate pain and fatigue as the most important outcomes. Five-year follow-up of Total bUrDen Of psoRiasis trial participants to investigate longer-term benefits of earlier diagnosis comparing treatment arms and to assess risk factors for psoriatic arthritis development is underway. This trial is registered as Current Controlled Trials ISRCTN38877516. This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research programme (NIHR award ref: RP-PG-1212-20007) and is published in full in Programme Grants for Applied Research; Vol. 13, No. 6. See the NIHR Funding and Awards website for further award information.

Personal narratives describing recovery from mental health problems are widely available to the public. We developed theory on the characteristics and impact of recovery narratives, developed curation procedures for the NEON Collection of 659 recovery narratives and developed and evaluated the NEON Intervention, a theory-informed web application providing access to the NEON Collection. To evaluate the effectiveness and cost-effectiveness of the NEON Intervention as compared to usual care and whether this varies by prior health service usage. Three pragmatic parallel-group randomised controlled trials of the NEON Intervention. Intervention arm participants received immediate access. Control arm participants received access after a 52-week follow-up. The effectiveness analysis was a linear regression model of outcome at 52 weeks. The cost-effectiveness analysis compared the incremental cost-effectiveness ratio to the £20,000–30,000 threshold defined in the National Institute for Health and Care Excellence reference case. All analyses were intention-to-treat and baseline-adjusted, with multiple imputation for missing data. England. All trials recruited people who were aged 18+ years, resident in England, capable of accessing or being supported to access the internet, able to understand written and spoken English and capable of providing online informed consent. NEON Trial participants also had experience of mental health-related distress in the last 6 months, and psychosis in the previous 5 years. NEON-O (i.e. non-psychosis) Trial participants also had experience of mental health-related distress in the last 6 months, but with no psychosis in the previous 5 years. People identifying as informal carers for people affected by mental health problems but not eligible for the NEON Trial or NEON-O Trial were recruited to the NEON-C feasibility trial. All inclusion criteria were self-rated. Recruitment was from March 2020 to March 2021, through public communications by the central study team, and the work of clinical support officers at 11 secondary care research sites. The NEON Intervention has four narrative access mechanisms: theory-informed algorithmic recommendation, random selection, self-selection by narrative category and return to impactful narratives. Participants used the NEON Intervention as much as they wished. Primary outcome: quality of life (Manchester Short Assessment). Secondary outcomes: distress, hope, self-efficacy, meaning in life and health status. For the NEON-O (i.e. non-psychosis) Trial, we found a significant baseline-adjusted difference of 0.13 (95% confidence interval 0.01 to 0.26, p = 0.041) in the Manchester Short Assessment score between intervention and control, and a significant baseline-adjusted difference of 0.22 (95% confidence interval 0.05 to 0.40, p = 0.014) in the presence subscale of the Meaning in Life Questionnaire. The incremental cost-effectiveness ratio was £12,526 per quality-adjusted life-year, lower than a threshold of £30,000 per quality-adjusted life-year used for health service commissioning in England. For participants who had used specialist mental health services at baseline, the intervention appeared to reduce cost, although confidence intervals were wide and results were not statistically significant (–£98, 95% credible interval –£606 to £309). It also improved quality-adjusted life-years (0.0165, 95% credible interval 0.0057 to 0.0273) per participant. Hence, for this subgroup of participants, it dominated usual care. For the NEON Trial, no significant baseline-adjusted differences in outcome were found. An incremental cost-effectiveness ratio of £110,501 was found for the NEON Intervention. A subgroup analysis provided preliminary evidence for greater cost-effectiveness for current mental health service users, with an incremental cost-effectiveness ratio of £35,013. The NEON-C Trial showed acceptability and feasibility for informal carers. It recommended integration of carer narratives and creation of an online carer community. Participants were recruited during a period in which movement and social interaction were widely affected by the COVID-19 pandemic, with the potential to influence generalisability. For the NEON-O Trial, we had an unrepresentative proportion of female-gendered participants (79.3%). Therefore, our NEON-O Trial findings cannot be generalised. This research programme has shown promising findings from the testing of the NEON Intervention. There is further research to do before implementation can be suggested. The NEON Intervention should be evaluated through a randomised controlled trial with people experiencing psychosis and using mental health services. The NEON Intervention should be refined to suit the needs of carers and then evaluated through a randomised controlled trial. The NEON-O Trial should be repeated with narrower mental health populations (e.g. mood disorders, eating disorders) to refine knowledge on effectiveness and cost-effectiveness. This may include refining the narrative collection used with these populations. If the NEON Intervention is implemented on a larger scale for people with non-psychosis mental health problems, then studies should be conducted to monitor benefits, continuously assess safety and documentation implementation processes. Future studies should consider alternative forms for presenting recovery narratives, including through multilanguage or multiculture support, and addressing digital exclusion by providing access through widely available technologies, such as smartphones and text messaging. Longitudinal designs are needed to document the short-term, medium-term and long-term impacts of recovery narratives. This trial is registered as NEON Trial ISRCTN11152837, NEON-O Trial ISRCTN63197153 and NEON-C Trial ISRCTN76355273. This award was funded by the National Institute for Health and Care Research (NIHR) Programme Grants for Applied Research programme (NIHR award ref.: RP-PG-0615-20016) and is published in full in Programme Grants for Applied Research; Vol. 13, No. 9. See the NIHR Funding and Awards website for further award information.