Deep learning architectures are increasingly proposed for patient trajectory modeling in electronic health records (EHRs), yet their advantage over simpler, more interpretable models is rarely subjected to rigorous empirical scrutiny in real-world clinical settings. We present a comprehensive patient timeline pipeline applied to elderly patients in CPRD Aurum, incorporating 260 clinical conditions classified via a three-tier automated framework including specialised detection logic for 17 complex conditions. Using this infrastructure, we benchmark Temporal Graph Convolutional Neural Networks (TG-CNN) against Logistic Regression with LASSO regularisation and Random Forests for predicting 12-month all-cause emergency hospitalisation risk, motivated by (but not filtered to) the elevated risk of adverse drug reactions. Under cross-validation, TG-CNN achieves a marginally higher mean AUC-ROC than LASSO (0.712 vs. 0.705), whereas on the held-out test set LASSO achieves the highest discrimination of three models (AUC-ROC 0.733, versus 0.710 for Random Forest and 0.702 for TG-CNN). We show, that discrimination alone is an incomplete criterion for clinical deployment: after Platt calibration, LASSO is the only model with an acceptable calibration slope (0.817), while Random Forest (0.759) and, TG-CNN (0.391) remain substantially miscalibrated. We argue that LASSO, not the highest-discriminating model, is the model best suited to direct clinical deployment. We present lessons for the machine learning and healthcare community regarding data infrastructure, model selection, and value of calibration and interpretability in high-stakes decision support.

BACKGROUND:Structured medication reviews (SMRs) are an essential component of medication optimization, especially for patients with multimorbidity and polypharmacy. However, the process remains challenging due to the complexities of patient data, time constraints, and the need for coordination among health care professionals (HCPs). This study explores HCPs' perspectives on the integration of artificial intelligence (AI)-assisted tools to enhance the SMR process, with a focus on the potential benefits of and barriers to adoption. OBJECTIVE:This study aims to identify the key user requirements for AI-assisted tools to improve the efficiency and effectiveness of SMRs, specifically for patients with multimorbidity, complex polypharmacy, and frailty. METHODS:A qualitative study was conducted involving focus groups and semistructured interviews with HCPs and patients in the United Kingdom. Participants included physicians, pharmacists, clinical pharmacologists, psychiatrists from primary and secondary care, a policy maker, and patients with multimorbidity. Data were analyzed using a hybrid inductive and deductive thematic analysis approach to identify themes related to AI-assisted tool functionality, workflow integration, user-interface visualization, and usability in the SMR process. RESULTS:Four major themes emerged from the analysis: innovative AI potential, optimizing electronic patient record visualization, functionality of the AI tool for SMRs, and facilitators of and barriers to AI tool implementation. HCPs identified the potential of AI to support patient identification and prioritizing those at risk of medication-related harm. AI-assisted tools were viewed as essential in detecting prescribing gaps, drug interactions, and patient risk trajectories over time. Participants emphasized the importance of presenting patient data in an intuitive format, with a patient interface for shared decision-making. Suggestions included color-coding blood results, highlighting critical medication reviews, and providing timelines of patient medical histories. HCPs stressed the need for AI tools to integrate seamlessly with existing electronic patient record systems and provide actionable insights without overwhelming users with excessive notifications or "pop-up" alerts. Factors influencing the uptake of AI-assisted tools included the need for user-friendly design, evidence of tool effectiveness (though some were skeptical about the predictive accuracy of AI models), and addressing concerns around digital exclusion. CONCLUSIONS:The findings highlight the potential for AI-assisted tools to streamline and optimize the SMR process, particularly for patients with multimorbidity and complex polypharmacy. However, successful implementation depends on addressing concerns related to workflow integration, user acceptance, and evidence of effectiveness. User-centered design is crucial to ensure that AI-assisted tools support HCPs in delivering high-quality, patient-centered care while minimizing cognitive overload and alert fatigue.

Open Science (OS) promises to democratise knowledge and reduce epistemic inequalities. However, a critical analysis reveals the potential of OS to amplify structural vulnerabilities, especially for people and communities already at the margins. With a particular focus on health data, this interdisciplinary essay examines how OS infrastructures perpetuate epistemic harms through the dominance of Eurocentric knowledge norms, legal regimes and corporate capture. Amidst the rapidly evolving health and data landscape, realising the social justice potential of OS, especially in healthcare, demands moving beyond techno-optimism to approaches that centre plural epistemologies, relational accountability and community empowerment.

Background: Breast cancer is the second most common primary tumour to metastasise to the brain. Improvements in the systemic treatment of extracranial metastatic disease have resulted in patients surviving longer with their metastatic disease, which appears to be contributing to the increase in the incidence of cerebral metastases. There is a need to better understand the number of patients affected, their outcomes and treatments in the context of current modern oncological treatment. Finally, it is recognised that inequality can exist with regard to treatment. This study aims to document the burden treatment and survival from CNS disease secondary to breast cancer in England utilising national cancer registry data. Study Design Methods: The study is a retrospective, cohort study using population-based registry data from The National Cancer Registration and Analysis Service (NCRAS) which is managed by NHS Digital (NHS England). Individual data on approx. 28,000 consecutive patients with CNS disease treated between 1995 and 2023 within the English Healthcare System will be utilized. Datasets utilised include: Hospital episode statistics (HES), Hospital Admitted Patient Care Activity (HESAPC), Systemic Anti-Cancer Therapy (SACT) and Radiotherapy Data Set (RTDS). Eligibility criteria: Inclusion Criteria Male or female, aged >16 years, Histologically and/or cytologically confirmed breast cancer with CNS involvement, as defined as having one or more of the following: a) Metastases to the brain parenchyma; b) Metastases to the leptomeninges c) Paraneoplastic Neurological Disorders. There were no formal exclusion criteria. This data release was approved by NHS Digital (ODR2021_030). Study Objectives: The primary objectives are to audit the overall survival from the initial diagnosis of CNS involvement secondary to breast cancer in English centres. With Secondary objectives will include to audit (1) the number of cases of metastatic breast cancer (MBC) involving the CNS presenting per year. (2) the current practice in regarding the diagnosis and management of CNS disease and (3) the outcomes of patients treated for CNS involvement secondary to breast cancer. Dara will be analysed by breast cancer subtype as well as overtime, by geographical location and by deprivation index. Citation Format: Talvinder Bhogal, Kukatharmini Tharmaratnam, Christopher Cheyne, Gary Leeming, Marta Garcia-Finana, Carlo Palmieri. PREMO CNS: PREsentation, Management and Outcomes of patients with CNS disease secondary to breast cancer in England [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P3-12-24.

Trusted Research Environments (TREs) are increasingly used as platforms for secure health data research, but they can also be used for implementing research findings or for action-research (researchers supporting health professionals to solve problems with advanced data analytics). Most TREs have been designed to support analysis of well-structured and coded data, however, with much clinical data recorded as unstructured notes, especially in mental health care, there needs to be a greater variety of tools and data management services available for safe research that includes natural language processing and anonymisation of data sources. The Mental Health Research for Innovation Centre (M-RIC), co-hosted by the University of Liverpool and Mersey Care NHS Foundation Trust, has implemented a novel TRE design that incorporates modern data engineering concepts to improve how researchers access a wider variety of linked data and machine learning tools, to be able to both undertake research and then deploy these tools directly into mental health care.

Background Patients with cancer are at greater risk of dying from COVID-19 than many other patient groups. However, how this risk evolved during the pandemic remains unclear. We aimed to determine, on the basis of the UK national pandemic protocol, how factors influencing hospital mortality from COVID-19 could differentially affect patients undergoing cancer treatment. We also examined changes in hospital mortality and escalation of care in patients on cancer treatment during the first 2 years of the COVID-19 pandemic in the UK. Methods We conducted a prospective cohort study of patients aged older than 19 years and admitted to 306 health-care facilities in the UK with confirmed SARS-CoV-2 infection, who were enrolled in the International Severe Acute Respiratory and emerging Infections Consortium (ISARIC) WHO Clinical Characterisation Protocol (CCP) across the UK from April 23, 2020, to Feb 28, 2022; this analysis included all patients in the complete dataset when the study closed. The primary outcome was 30-day in-hospital mortality, comparing patients on cancer treatment and those without cancer. The study was approved by the South Central-Oxford C Research Ethics Committee in England (Ref: 13/SC/0149) and the Scotland A Research Ethics Committee (Ref 20/SS/0028), and is registered on the ISRCTN Registry (ISRCTN66726260). Findings 177 871 eligible adult patients either with no history of cancer (n=171 303) or on cancer treatment (n=6568) were enrolled; 93 205 (524%) were male, 84 418 (475%) were female, and in 248 (139%) sex or gender details were not specified or data were missing. Patients were followed up for a median of 13 (IQR 6-21) days. Of the 6568 patients receiving cancer treatment, 2080 (317%) died at 30 days, compared with 30 901 (180%) of 171 303 patients without cancer. Patients aged younger than 50 years on cancer treatment had the highest age-adjusted relative risk (hazard ratio [HR] 52 [95% CI 40-66], p<00001; vs 50-69 years 24 [22-26], p<00001; 70-79 years 18 [16-20], p<00001; and >80 years 15 [13-16], p<00001) but a lower absolute risk (51 [67%] of 763 patients <50 years died compared with 459 [302%] of 1522 patients aged >80 years). In -hospital mortality decreased for all patients during the pandemic but was higher for patients on cancer treatment than for those without cancer throughout the study period. Interpretation People with cancer have a higher risk of mortality from COVID-19 than those without cancer. Patients younger than 50 years with cancer treatment have the highest relative risk of death. Continued action is needed to mitigate the poor outcomes in patients with cancer, such as through optimising vaccination, long-acting passive immunisation, and early access to therapeutics. These findings underscore the importance of the ISARIC-WHO pandemic preparedness initiative.

Abstract Background Covid-19 healthcare worker testing, isolation and quarantine policies had to balance risks to patients from the virus and from staff absence. The emergence of the Omicron variant led to dangerous levels of key-worker absence globally. We evaluated whether using two manufacturers’ lateral flow tests (LFTs) concurrently improved SARS-CoV-2 Omicron detection significantly and was acceptable to hospital staff. In a nested study, to understand risks of return to work after a 5-day isolation/quarantine period, we examined virus culture 5–7 days after positive test or significant exposure. Methods Fully-vaccinated Liverpool (UK) University Hospitals staff participated (February-May 2022) in a random-order, open-label trial testing whether dual LFTs improved SARS-CoV-2 detection, and whether dual swabbing was acceptable to users. Participants used nose-throat swab Innova and nose-only swab Orient Gene LFTs in daily randomised order for 10 days. A user-experience questionnaire was administered on exit. Selected participants gave swabs for viral culture on days 5–7 after symptom onset or first positive test. Cultures were considered positive if cytopathic effect was apparent or SARS-CoV-2 N gene sub-genomic RNA was detected. Results Two hundred and twenty-six individuals reported 1466 pairs of LFT results. Tests disagreed in 127 cases (8.7%). Orient Gene was more likely (78 cf. 49; OR: 2.1, 1.1–4.1; P = 0.03) to be positive. If Innova was swabbed second, it was less likely to agree with a positive Orient Gene result (OR: 2.7, 1.3–5.2; P = 0.005); swabbing first with Innova made no significant difference (OR: 1.1, 0.5–2.3; P = 0.85). Orient Gene positive Innova negative result-pairs became more frequent over time (OR: 1.2, 1.1–1.3; P < 0.001). Of individuals completing the exit questionnaire, 90.7% reported dual swabbing was easy, 57.1% said it was no barrier to their daily routine and 65.6% preferred dual testing. Respondents had more confidence in dual versus single test results. Viral cultures from days 5–7 were positive for 6/31 (19.4%, 7.5%-37.5%) and indeterminate for 11/31 (35.5%, 19.2%-54.6%) LFT-positive participants, indicating they were likely still infectious. Conclusions Dual brand testing increased LFT detection of SARS-CoV-2 antigen by a small but meaningful margin and was acceptable to hospital workers. Viral cultures demonstrated that policies recommending safe return to work ~ 5 days after Omicron infection/exposure were flawed. Key-workers should be prepared for dynamic self-testing protocols in future pandemics. Trial registration https://www.isrctn.com/ISRCTN47058442 (26 January 2022).

Adapting language models (LMs) to novel domains is often achieved through fine-tuning a pre-trained LM (PLM) on domain-specific data. Fine-tuning introduces new knowledge into an LM, enabling it to comprehend and efficiently perform a target domain task. Fine-tuning can however be inadvertently insensitive if it ignores the wide array of disparities (e.g in word meaning) between source and target domains. For instance, words such as chronic and pressure may be treated lightly in social conversations, however, clinically, these words are usually an expression of concern. To address insensitive fine-tuning, we propose Mask Specific Language Modeling (MSLM), an approach that efficiently acquires target domain knowledge by appropriately weighting the importance of domain-specific terms (DS-terms) during fine-tuning. MSLM jointly masks DS-terms and generic words, then learns mask-specific losses by ensuring LMs incur larger penalties for inaccurately predicting DS-terms compared to generic words. Results of our analysis show that MSLM improves LMs sensitivity and detection of DS-terms. We empirically show that an optimal masking rate not only depends on the LM, but also on the dataset and the length of sequences. Our proposed masking strategy outperforms advanced masking strategies such as span- and PMI-based masking.

Background: Population ageing has led to an increase in multimorbidity and polypharmacy. Some medications may need to be stopped, but patient attitudes towards deprescribing are poorly understood. This study explores attitudes towards (de)prescribing in patients with multimorbidity in the UK primary care. Methods: Patients with multimorbidity were invited to complete the Revised Patients Attitudes Towards Deprescribing (rPATD) Questionnaire using the Evergreen Life Personal Health Record App (Manchester, UK). The responses were linked to electronic health records. Anonymised data were analysed in a trusted research environment (University of Liverpool) for group comparisons and using multivariable logistic regression to identify factors associated with satisfaction with current medications. Results: A total 1,019 patients participated in the study (n=365 aged <65, 30% males; n=654 ≥65, 57% males). Most patients were satisfied with their current medications (74% aged <65, 70% aged ≥65) but were willing to stop one or more of their regular medicines if their doctor said it was possible (82%, 68% accordingly). Polypharmacy, use of antihypertensive drugs, and antidepressants were associated with patient-reported burden in taking medicines. Frailty did not influence patient deprescribing attitudes. Patients who were satisfied with current medications had fewer medications. Independent predictors of satisfaction with current medications were higher total involvement and appropriateness scores, and lower total burden score. Conclusions: Most patients with multimorbidity would consider stopping some of their medications, even when they are generally satisfied with the treatments they received. Frailty status does not imply willingness to stop medications. Clinicians should discuss medication deprescribing for shared decision. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement This study/project is funded by the National Institute for Health Research (NIHR) under its Programme Artificial Intelligence for Multiple and Long-Term Conditions (NIHR203986). The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. IB is supported by NIHR as Senior Investigator award (NIHR205131).AW is partly funded by Health and Care Research Wales award (NHS-RTA-21-02) ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: The Newcastle North Tyneside Research Ethics Committee (REC reference:22/NE/0088) granted ethical approval for the DynAIRx study. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes All data produced in the present study are available upon reasonable request to the authors

IntroductionStructured medication reviews (SMRs), introduced in the United Kingdom (UK) in 2020, aim to enhance shared decision-making in medication optimisation, particularly for patients with multimorbidity and polypharmacy. Despite its potential, there is limited empirical evidence on the implementation of SMRs, and the challenges faced in the process. This study is part of a larger DynAIRx (Artificial Intelligence for dynamic prescribing optimisation and care integration in multimorbidity) project which aims to introduce Artificial Intelligence (AI) to SMRs and develop machine learning models and visualisation tools for patients with multimorbidity. Here, we explore how SMRs are currently undertaken and what barriers are experienced by those involved in them.MethodsQualitative focus groups and semi-structured interviews took place between 2022-2023. Six focus groups were conducted with doctors, pharmacists and clinical pharmacologists (n = 21), and three patient focus groups with patients with multimorbidity (n = 13). Five semi-structured interviews were held with 2 pharmacists, 1 trainee doctor, 1 policy-maker and 1 psychiatrist. Transcripts were analysed using thematic analysis.ResultsTwo key themes limiting the effectiveness of SMRs in clinical practice were identified: 'Medication Reviews in Practice' and 'Medication-related Challenges'. Participants noted limitations to the efficient and effectiveness of SMRs in practice including the scarcity of digital tools for identifying and prioritising patients for SMRs; organisational and patient-related challenges in inviting patients for SMRs and ensuring they attend; the time-intensive nature of SMRs, the need for multiple appointments and shared decision-making; the impact of the healthcare context on SMR delivery; poor communication and data sharing issues between primary and secondary care; difficulties in managing mental health medications and specific challenges associated with anticholinergic medication.ConclusionSMRs are complex, time consuming and medication optimisation may require multiple follow-up appointments to enable a comprehensive review. There is a need for a prescribing support system to identify, prioritise and reduce the time needed to understand the patient journey when dealing with large volumes of disparate clinical information in electronic health records. However, monitoring the effects of medication optimisation changes with a feedback loop can be challenging to establish and maintain using current electronic health record systems.

Objectives To understand severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) transmission risks, perceived risks and the feasibility of risk mitigations from experimental mass cultural events before coronavirus disease 2019 (COVID-19) restrictions were lifted. Design Prospective, population-wide observational study. Setting Four events (two nightclubs, an outdoor music festival and a business conference) open to Liverpool City Region UK residents, requiring a negative lateral flow test (LFT) within the 36 h before the event, but not requiring social distancing or face-coverings. Participants A total of 12,256 individuals attending one or more events between 28 April and 2 May 2021. Main outcome measures SARS-CoV-2 infections detected using audience self-swabbed (5–7 days post-event) polymerase chain reaction (PCR) tests, with viral genomic analysis of cases, plus linked National Health Service COVID-19 testing data. Audience experiences were gathered via questionnaires, focus groups and social media. Indoor CO 2 concentrations were monitored. Results A total of 12 PCR-positive cases (likely 4 index, 8 primary or secondary), 10 from the nightclubs. Two further cases had positive LFTs but no PCR. A total of 11,896 (97.1%) participants with scanned tickets were matched to a negative pre-event LFT: 4972 (40.6%) returned a PCR within a week. CO 2 concentrations showed areas for improving ventilation at the nightclubs. Population infection rates were low, yet with a concurrent outbreak of >50 linked cases around a local swimming pool without equivalent risk mitigations. Audience anxiety was low and enjoyment high. Conclusions We observed minor SARS-CoV-2 transmission and low perceived risks around events when prevalence was low and risk mitigations prominent. Partnership between audiences, event organisers and public health services, supported by information systems with real-time linked data, can improve health security for mass cultural events.

Background It is unclear what effect the pattern of health-care use before admission to hospital with COVID-19 (index admission) has on the long-term outcomes for patients. We sought to describe mortality and emergency readmission to hospital after discharge following the index admission (index discharge), and to assess associations between these outcomes and patterns of health-care use before such admissions.Methods We did a national, retrospective, complete cohort study by extracting data from several national databases and linking the databases for all adult patients admitted to hospital in Scotland with COVID-19. We used latent class trajectory modelling to identify distinct clusters of patients on the basis of their emergency admissions to hospital in the 2 years before the index admission. The primary outcomes were mortality and emergency readmission up to 1 year after index admission. We used multivariable regression models to explore associations between these outcomes and patient demographics, vaccination status, level of care received in hospital, and previous emergency hospital use.Findings Between March 1, 2020, and Oct 25, 2021, 33 580 patients were admitted to hospital with COVID-19 in Scotland. Overall, the Kaplan-Meier estimate of mortality within 1 year of index admission was 296% (95% CI 291-302). The cumulative incidence of emergency hospital readmission within 30 days of index discharge was 144% (95% CI 140-148), with the number increasing to 356% (349-363) patients at 1 year. Among the 33 580 patients, we identified four distinct patterns of previous emergency hospital use: no admissions (n=18 772 [559%]); minimal admissions (n=12 057 [359%]); recently high admissions (n=1931 [58%]), and persistently high admissions (n=820 [24%]). Patients with recently or persistently high admissions were older, more multimorbid, and more likely to have hospital-acquired COVID-19 than patients with no or minimal admissions. People in the minimal, recently high, and persistently high admissions groups had an increased risk of mortality and hospital readmission compared with those in the no admissions group. Compared with the no admissions group, mortality was highest in the recently high admissions group (post-hospital mortality HR 270 [95% CI 235-281]; p<00001) and the risk of readmission was highest in the persistently high admissions group (323 [289-361]; p<00001).Interpretation Long-term mortality and readmission rates for patients hospitalised with COVID-19 were high; within 1 year, one in three patients had died and a third had been readmitted as an emergency. Patterns of hospital use before index admission were strongly predictive of mortality and readmission risk, independent of age, pre-existing comorbidities, and COVID-19 vaccination status. This increasingly precise identification of individuals at high risk of poor outcomes from COVID-19 will enable targeted support.

Context: The DynAIRx project aims to develop artificial intelligence (AI) tools to support medication reviews for patients with multimorbidity (people with ≥2 chronic conditions), targeting those at greatest risk of medicine-related harm. Challenges faced by healthcare professionals (HCPs) managing multimorbid patients include poor integration of health records across providers and few tools to assist in stratifying patients requiring medication reviews. Objective: To explore how medication reviews are currently being undertaken and how they might be augmented by AI. Study design and analysis: 5 semistructured interviews and 6 focus groups with HCPs (n=26); 2 focus groups with patients (n=10). These were transcribed verbatim and analysed using inductive thematic analysis. Setting: England. Focus groups were conducted via MicroSoft Teams with pharmacists, general practitioners, secondary care clinicians, and policy makers. Patient focus groups were conducted face-to-face. Population studied: The DynAIRx tool will target potentially problematic polypharmacy in 3 key multimorbidity groups: people with mental and physical health problems; those with ≥4 chronic conditions or taking ≥10 drugs; and older patients with frailty. Outcome measures: Report on barriers and facilitators to effective medicine reviews and potential implications for the implementation of new decision support tools. Results: Availability of staff, access to patient information, organisational contracts and patient demographics influenced uptake of medication reviews. Time was a major limiting factor due to the overwhelming density of information in electronic health records, especially for complex patients. Building continuity in medication reviews and adopting a team-based approach to dealing with complex multimorbid patients was emphasised. HCPs welcomed user-friendly digital tools with an intuitive interface that could be used to reduce "detective work" and enable shared decision making with patients. A timeline including diagnoses linked to medicines by indications and previous investigations was viewed as an early potential solution to reduce laborious searching. HCPs were generally positive about using AI tools to aid risk stratification of patients needing medication reviews but emphasised ease of use as important. Conclusions: These findings and those of an observational time-and-motion study will inform development of the DynAIRx prototype. HCPs seem receptive to such a tool.

In recent years, researchers have begun to explore the use of Distributed Ledger Technologies (DLT), also known as blockchain, in health data sharing contexts. However, there is a significant lack of research that examines public attitudes towards the use of this technology. In this paper, we begin to address this issue and present results from a series of focus groups which explored public views and concerns about engaging with new models of personal health data sharing in the UK. We found that participants were broadly in favour of a shift towards new decentralised models of data sharing. Retaining ‘proof’ of health information stored about patients and the capacity to provide permanent audit trails, enabled by immutable and transparent properties of DLT, were regarded as particularly valuable for our participants and prospective data custodians. Participants also identified other potential benefits such as supporting people to become more health data literate and enabling patients to make informed decisions about how their data was shared and with whom. However, participants also voiced concerns about the potential to further exacerbate existing health and digital inequalities. Participants were also apprehensive about the removal of intermediaries in the design of personal health informatics systems.

Background:Structured Medication Reviews (SMRs) are intended to help deliver the NHS Long Term Plan for medicines optimisation in people living with multiple long-term conditions and polypharmacy. It is challenging to gather the information needed for these reviews due to poor integration of health records across providers and there is little guidance on how to identify those patients most urgently requiring review. Objective:To extract information from scattered clinical records on how health and medications change over time, apply interpretable artificial intelligence (AI) approaches to predict risks of poor outcomes and overlay this information on care records to inform SMRs. We will pilot this approach in primary care prescribing audit and feedback systems, and co-design future medicines optimisation decision support systems. Design:DynAIRx will target potentially problematic polypharmacy in three key multimorbidity groups, namely, people with (a) mental and physical health problems, (b) four or more long-term conditions taking ten or more drugs and (c) older age and frailty. Structured clinical data will be drawn from integrated care records (general practice, hospital, and social care) covering an ∼11m population supplemented with Natural Language Processing (NLP) of unstructured clinical text. AI systems will be trained to identify patterns of conditions, medications, tests, and clinical contacts preceding adverse events in order to identify individuals who might benefit most from an SMR. Discussion:By implementing and evaluating an AI-augmented visualisation of care records in an existing prescribing audit and feedback system we will create a learning system for medicines optimisation, co-designed throughout with end-users and patients.

The Liverpool Citizens’ Jury was a public consultation on the use of health data to tackle the significant problem of Antimicrobial Resistance (AMR) and is the first step in creating a local AMR network with national and international relevance. The 18 jurors were tasked with learning about AMR as it relates to research and considered how organisations might collect, share and utilise pseudo-anonymised patient data. The overarching aim is to produce a new model supporting societal change focused on Antibiotic Guardianship and to combat the public health challenge of AMR. The model will be implemented in the UK and provided to an international network enabling global knowledge transfer.

Background Dexamethasone was the first intervention proven to reduce mortality in patients with COVID-19 being treated in hospital. We aimed to evaluate the adoption of corticosteroids in the treatment of COVID-19 in the UK after the RECOVERY trial publication on June 16, 2020, and to identify discrepancies in care. Methods We did an audit of clinical implementation of corticosteroids in a prospective, observational, cohort study in 237 UK acute care hospitals between March 16, 2020, and April 14, 2021, restricted to patients aged 18 years or older with proven or high likelihood of COVID-19, who received supplementary oxygen. The primary outcome was administration of dexamethasone, prednisolone, hydrocortisone, or methylprednisolone. This study is registered with ISRCTN, ISRCTN66726260. Findings Between June 17, 2020, and April 14, 2021, 47 795 (75.2%) of 63 525 of patients on supplementary oxygen received corticosteroids, higher among patients requiring critical care than in those who received ward care (11 185 [86.6%] of 12 909 vs 36 415 [72.4%] of 50 278). Patients 50 years or older were significantly less likely to receive corticosteroids than those younger than 50 years (adjusted odds ratio 0.79 [95% CI 0.70-0.89], p=0.0001, for 70-79 years; 0.52 [0.46-0.58], p<0.0001, for >80 years), independent of patient demographics and illness severity. 84 (54.2%) of 155 pregnant women received corticosteroids. Rates of corticosteroid administration increased from 27.5% in the week before June 16, 2020, to 75-80% in January, 2021. Interpretation Implementation of corticosteroids into clinical practice in the UK for patients with COVID-19 has been successful, but not universal. Patients older than 70 years, independent of illness severity, chronic neurological disease, and dementia, were less likely to receive corticosteroids than those who were younger, as were pregnant women. This could reflect appropriate clinical decision making, but the possibility of inequitable access to life-saving care should be considered. Copyright (C) 2022 The Author(s). Published by Elsevier Ltd.