Background Communication in a family's primary language can support safe care. Vital steps within the care delivery process are contingent on successful communication, including reporting symptoms, clinical decision-making, informed consent, discharge communication and follow-up coordination. The importance of effective information exchange is particularly pronounced in paediatric emergency care, and complex interactions may arise as parents or carers advocate on behalf of children. This scoping review aimed to identify and map existing research indicating where along the care journey communication-related risks for safety lie during paediatric emergency care and what strategies exist to mitigate them.Methods We searched MEDLINE, Embase, CINAHL, Scopus, Web of Science and Cochrane Library for studies which examined the influence of language barriers on patient safety in paediatric emergency care as well as studies that evaluated interventions. Bibliographic database searches were executed on 18 December 2024; retrieved records were independently screened by two authors at title and abstract level followed by full text level. Data on study objectives, population characteristics, study design and their key findings were extracted.Results 1578 articles were identified, of which 33 were included and mapped according to (i) studies reporting safety risks linked to language barriers in paediatric emergency care (n=24) and (ii) existing interventions designed to mitigate these risks (n=9). Studies highlighted that language barriers can influence safety at multiple stages of the emergency care pathway, with discharge most frequently reported as a point of risk for paediatric patient safety. Interventions focused primarily on usage, uptake and documentation of professional interpreter services.Conclusion Addressing misunderstandings around follow-up and home-care advice during medical safety netting are priority areas for intervention. Future research should involve carer and clinical perspectives in exploring whether technology-enabled tools, including artificial intelligence, can safely mitigate language barriers in these situations.

Testicular cancer is the most common malignancy in males age 15-40 years and one of the most curable cancers, with a cumulative 10-year survival rate exceeding 90%. Management strategies depend on the histologic subtype, stage at diagnosis, sites of disease, tumor markers, and risk classification. Germ cell tumors, including seminomas and nonseminomas, constitute the majority of testicular cancers and require distinct therapeutic approaches. For localized disease, radical orchidectomy remains the cornerstone of treatment, followed by active surveillance, chemotherapy, or primary retroperitoneal lymph node dissection, depending on the histology and the risk of relapse. Seminomas are highly curable, with low-stage patients often managed through surveillance or postoperative single-agent carboplatin. By contrast, nonseminomas typically require adjuvant multiagent chemotherapy, such as bleomycin, etoposide, and cisplatin, particularly in higher-risk patients. For metastatic disease, chemotherapy remains the standard of care, achieving excellent cure rates even in patients with bulky tumors. Surgical resection of residual masses is especially critical in nonseminomatous germ cell tumors to remove viable cancer or teratoma components. The treatment of refractory or relapsed disease frequently involves second-line standard-dose or high-dose chemotherapy with autologous stem-cell transplantation, ideally performed in specialized high-volume centers. Before, during, and after treatment, multidisciplinary care is essential to addressing psychosocial challenges, optimizing fertility preservation, and enhancing quality of life. After curative treatments, long-term management involves regular follow-up to monitor for recurrence, late toxicities, and secondary malignancies, with survivorship programs playing a crucial role in meeting patients' ongoing needs. Advances in molecular diagnostics for early relapse detection and the introduction of targeted therapies continue to improve outcomes, particularly in resistant patients.

Pulmonary arteriovenous malformations (PAVMs) larger than 4mm in size are estimated to affect 38 per 100,000 individuals [95% confidence intervals 18-76]. They provide an anatomical right-to-left shunt such that each heartbeat, a proportion of the cardiac output bypasses the pulmonary capillary bed, preventing essential processing functions such as gas exchange and filtration of blood-borne emboli. Although large cohort series were published in earlier decades, more recent data series have been scant. To support modern educational platforms, here we report features of 1149 consecutive patients with imaging-proven PAVMs, reviewed at a single UK centre between 1984-2026, including 813 (71%) with clinical and/or genetically confirmed hereditary haemorrhagic telangiectasia (HHT). The median age was 47y, and 735 (64%) were female. We report 4348 oxygen saturation measurements at presentation and follow-up, and 810 pulmonary artery pressure (PAP) measurements made at angiography prior to treatment of PAVMs by embolisation. Together, these confirm that there is no risk of hypoxic pulmonary hypertension, with PAP measurements higher in patients with higher SaO2. Massive haemoptysis or haemothorax occurred in 18 patients [0.009, 0.023], of which 7/18 [95% CI 0.01, 0.64] events were pregnancy-associated. Ischaemic strokes affected 125 patients [0.09, 0.13], brain abscess 107 patients [0.08, 0.11] patients, and haemorrhagic strokes 29 patients [0.02, 0.03] patients. These data will inform design of future work to evaluate aetiologies, associations and implications for clinical practice. ### Competing Interest Statement The authors have declared no competing interest. ### Clinical Protocols ### Funding Statement The National Institute for Health Research (NIHR) Imperial Biomedical Research Centre (NIHR203323); HHT charitable donations, and received specific funding from the NIHR Imperial BRC Digital Health Theme pilot project scheme ### 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: Ethical approval for the prospective study was provided by the Hammersmith and Queen Charlottes & Chelsea Research Ethics Committee (REC ref. 2000/5764: Hammersmith Hospital patients with pulmonary arteriovenous malformations (PAVMs) and hereditary haemorrhagic telangiectasia) and remains valid. The research database was given favourable ethics approval by the South West - Central Bristol Research Ethics Committee (reference 21/SW/0120; IRAS project ID 282093). 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 Data access is via the iCARE Secure Data Environment (https://www.imperial.ac.uk/medicine/research-and-impact/groups/icare/access-to-data/).

Background Frailty constitutes a growing challenge for health and social care systems around the world. In England, 35% of adults aged 65 years and older live with frailty, with international estimates indicating that almost half of all hospital inpatients within the same age group are frail. This population often experiences multimorbidity and frequent care transitions. Written documentation and verbal handovers may lack the precision and nuance required to understand an older adult’s presentation and support needs. Video recordings of individual patients, capturing aspects of their functional abilities and condition, may help to enhance multidisciplinary team communication and care continuity, yet little is known about their use in the care of older inpatients with frailty. Objective We aimed to evaluate the acceptability, feasibility of implementation, and perceived effectiveness of video-based patient records (the Isla Health Digital Pathway Platform) for supporting the assessment and care of older inpatients with frailty within the acute hospital setting. Methods A nonrandomized mixed methods pilot study was conducted within 3 acute medicine wards for older adults. The video-based patient records intervention, permitting videos to be embedded securely within the electronic patient record, was implemented over a 3-month period alongside usual care. Patient enrollment and retention figures; qualitative interviews with patients, carers, and clinical staff; and video capture and view metrics were used to address the study objectives. The Theoretical Framework of Acceptability of Healthcare Interventions was applied to the framework analysis of interview data, capturing concepts such as intervention ethicality, burden, and coherence. Patient and public involvement and engagement informed each research stage. Results Twenty-nine patients were enrolled (56.9%); 1 patient withdrew before receiving the intervention. Modal reasons given by patients for nonparticipation included not wanting to take part in research (n=8) or feeling too unwell (n=2). Staff identified multiple opportunities for capturing patient videos, including documentation of mobility assessments or seizures. The intervention was considered acceptable on the grounds that safeguards were always in place, including secure data storage and upholding of patient dignity. Implementation barriers and facilitators were identified; factors such as difficulties in capturing videos within busy ward environments and scheduling issues were voiced by participants. Video view metrics and data from interviews collectively suggested low rates of engagement with videos by clinical staff once captured. Potential intervention impacts included perceived enhancements to clinical assessment and person-centered care. Conclusions Our findings suggest that the intervention is largely acceptable to patients, carers, and clinical staff. Conclusions as to intervention feasibility were mixed, with limited engagement with videos suggesting further work is required to promote sufficient uptake among staff. Finally, this research presents promising patient, carer, and clinical opinion as to the potential effectiveness of video-based patient records for improving aspects of patient care. Trial Registration ClinicalTrials.gov NCT06504641; https://clinicaltrials.gov/study/NCT06504641

BACKGROUND: ‘Winter pressures’ in urgent and emergency care (UEC) are widely accepted but have had little empirical attention. Amidst annually increasing demand for UEC and reports of extreme strain during winter, we aimed to understand the extent and nature of seasonal demand by analysing routine data from Emergency Departments (ED) and acute Admitted Patient Care (APC) episodes across England. METHODS: This was a retrospective observational analysis using data from 26 hospitals and 22 EDs between 2021-11-1 and 2022-10-31 comparing emergency attendances and acute admissions between winter (October-March) and summer (April-September). Main outcomes included ED waiting times, length of admissions, the number of investigations, treatments and procedures received, and whether the contact was considered avoidable. Using a novel ‘federated’ approach exploiting local relationships with data providers, regional researchers analysed local data and provided summary statistics and analysis results to the lead site. Aggregation of summary results established a picture of seasonal demand across the country, and an understanding of regional variation in seasonal trends. RESULTS: 1,549,205 ED attendances (775,810 winter; 50.1%) and 747,685 APC admissions (368,910 winter, 49.3%) were analysed. We found no systematic seasonal differences in the number or nature of presentations. While regional variation existed for many outcomes, no nationally consistent effect of winter was found for any measure. CONCLUSIONS: Winter pressures in UEC may not be driven by large differences in the number, avoidability or acuity of ED attendances or APC admissions. Rather, UEC may be operating at or near to capacity all year, meaning small fluctuations in demand or in the complexity of presentations may cause significant strain on an over-burdened system. Focus on managing seasonal demand should be modified to address year-round pressure. Effective policy may require structural reconfiguration to better regulate demand.

Background: Hospital discharge summaries are essential for continuity of care, but fragmented electronic health record data can compromise their quality and safety. Large language models (LLMs) are increasingly proposed for clinical documentation tasks, despite limited evidence regarding their performance in healthcare settings. We evaluated the clinical acceptability of GPT-4o-mini-generated discharge summaries derived from electronic health records compared with routine clinician-written summaries. Methods: Retrospective non-inferiority study comparing LLM-generated with clinician-authored discharge summaries, using admission records from five London hospitals (Sept 1, 2022, to Sept 30, 2023). 1294 summaries (645 LLM-generated; 649 clinician-authored) were evaluated by 12 practising doctors. LLM-generated summaries were produced using GPT-4o-mini (2024-07-18) and a clinician co-designed prompt. The primary outcome was clinician willingness to sign the summary with no or minor edits; a 5-percentage-point non-inferiority margin was prespecified. Secondary outcomes assessed accuracy, completeness, tone, clarity, and potential harmfulness of missing or incorrect information on a 5-point Likert scale. Analysis used generalised linear mixed models adjusting for reviewer and case-level variability. Findings: The absolute difference in predicted probability of sign-off (LLM-generated minus clinician-authored) was 22.2 percentage points (95% CI 19.0-27.5) and non-inferiority was demonstrated. Compared with clinician-authored summaries, LLM-generated summaries received higher ratings for accuracy, completeness, tone, and clarity (adjusted risk differences for score ≥4 were 19.4–29.1 percentage points; all 95% CIs excluded zero). LLM-generated discharge summaries had a lower proportion of cases rated ≥3 for potential harm (17.2-19.4%) compared with clinician-authored summaries (37.5-38.5%). Interpretation: LLM-generated discharge summaries meet the current standard of care and represent a promising early use case for generative artificial intelligence (AI) in healthcare with close clinician oversight. Funding This research was funded by the NIHR Imperial Biomedical Research Centre (NIHR203323). The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. ​​

Objectives Rising demand for emergency care in England is a continuing challenge driven by population ageing and increasing multimorbidity. Ambulatory emergency care (AEC) refers to the provision of same-day acute care for patients who might otherwise require admission. However, the contribution of AEC conditions to demand remains unclear. This study aimed to examine the proportion and nature of patients attending emergency departments (ED) with AEC-related conditions and to describe variation between hospitals in attendances and emergency admissions for AEC conditions. Design and setting A retrospective study of routine data from 21 acute hospitals in England, including adult ED attendances and emergency admissions between 1 November 2021 and 31 October 2022. We used a federated approach to ensure data security, applying established AEC definitions to explore variation by age, socioeconomic status and length of stay. Outcome measures Primary: Proportion of (i) ED attendances and (ii) emergency admissions for AEC conditions. Secondary: (i) Proportion of patients presenting at ED with an AEC condition who were admitted; (ii) proportion of emergency admissions with an AEC condition with a length of stay <2 days. Results We analysed 1 513 480 attendances (median per hospital: 73 125) and 660 105 admissions (median per hospital: 30 425). AEC accounted for 29.6% of attendances and 40.8% of admissions, with substantial inter-hospital variability. Patients aged ≥65 were more likely to present with an AEC, while patients from deprived areas had lower rates. Among AEC-related admissions, 49.3% had a stay of less than 2 days. Conclusions Nearly one-third of attendances and two-fifths of admissions were for conditions potentially manageable in AEC or community settings. Variation between hospitals suggests local factors, including service configuration and primary care access, may influence avoidable acute care use. These findings suggest a need for a more nuanced understanding of the drivers behind AEC, or SDEC Services, to better understand their impact on reducing hospital admissions. Analysing these patterns may inform interventions to reduce avoidable hospital utilisation. Further research is needed to identify drivers of variation and to develop scalable strategies for prevention.

Background A discharge summary should be a clinical report that documents a patient’s hospital stay, including test results, diagnoses, management, and follow-up. Currently, discharge summaries are written by clinicians who manually locate pertinent information across the electronic health record, of which approximately 80% is free text. This process is time-consuming and may be suitable for automation using large language models. Objective This study developed a template-based prompting system that can produce clinically acceptable discharge summaries, specifically the “clinical summary” and “plan and requested actions” sections, from routinely collected electronic patient records. Methods This study used electronic health record data from Imperial College Healthcare National Health Service Trust, a network of 5 hospitals in northwest London. It was conducted within the Imperial Secure Data Environment under Data Access and Ethics Committee approval. In total, 52 inpatient encounters were selected by the clinical team to ensure diversity in clinical specialty, reason for admission, complexity, length of stay, and sociodemographic characteristics; 83% (n=43) of the cases were allocated to the development dataset, and 17% (n=9) comprised the test dataset. The system synthesized clinical notes related to an inpatient hospital encounter and used structured template prompts with OpenAI’s generative pretrained transformer-4 to generate a discharge summary. The prompt was co-designed across 3 iterations. Resident physicians completed an evaluation form to assess the clinical acceptability of the generated summaries, including the primary outcome (global confidence rating) and secondary outcomes (accuracy, completeness, readability, formatting, sociodemographic bias, and potential clinical harm). Sensitivity analyses assessed the effect of length of stay and admission type (emergency department vs other and surgical vs other) on the primary outcome. Results A total of 52 patients (n=32, 62% female) were included, with a mean age of 44.8 (SD 27.1) years and an average length of stay of 15.2 (SD 21.1) days. In the test dataset, 89% (8/9) of generative pretrained transformer-4–generated summaries received a positive global confidence rating (“yes” or “yes, with minor changes”). Secondary outcomes were positive for the “clinical summary” section (8/9, 89% complete and 7/9, 78% accurate) and the “plan and requested actions” section (7/9, 78% complete and 7/9, 78% accurate). Readability, formatting, sociodemographic bias, and potential clinical harm also showed positive results in the test dataset. Sensitivity analyses showed no statistically significant variation in the primary outcome across length of stay or admission type (length of stay: P=.29; surgical admission: P=.99; emergency department admission: P=.15). Conclusions Our results demonstrate the feasibility of the pipeline, but rigorous statistical evaluation in a larger, adequately powered sample is needed.

Nucleos/tide analogue (NA) drugs are used for long-term treatment of chronic hepatitis B virus (HBV) infection, with treatment eligibility criteria changing rapidly amidst globally evolving clinical guidelines. We aimed to quantify the prescription of NA drugs to date, and to undertake a preliminary assessment of the impact of relaxing treatment eligibility thresholds, leveraging a unique large real-world secondary care dataset. We assimilated longitudinal clinical data, collected between February 1997 and April 2023 from adults with chronic HBV infection from six centres in England through the UK NIHR Health Informatics Collaborative (HIC) Viral Hepatitis and Liver Disease theme. We describe factors currently associated with the receipt of NA treatment and determine the proportion of the population who would become treatment eligible as thresholds change. Across 7558 adults with a mean follow-up of 4.0 years (SD 3.9), NA treatment was prescribed in 2014/7558 (26.6%), and as expected according to guidelines at the time, was associated with HBV e-antigen (HBeAg) positivity and alanine transferase (ALT) above the upper limit of normal (> ULN). Treatment was more likely in males, older adults, in Asian and Other ethnicities (compared to White), and less likely in socioeconomically deprived individuals. The proportion of treatment-eligible individuals was 32.3% based on 2 records of ALT > ULN over 6-12 months, 41.7% based on ALT > ULN and viral load (VL) > 2000 IU/mL, and 95.1% based on detectable VL and either ALT > ULN or age > 30 years. Evolving clinical guidelines will lead to substantial increases in the proportion of individuals living with HBV who are eligible for treatment, underlining the need for services to adapt rapidly to the changing clinical environment.

Technology can augment the interaction between patients and the healthcare system, but its adoption can vary widely across patients due to their personal circumstances or supply-side features. We formulated this study to answer the following variable importance question: which patient, workforce and organisational characteristics are the most important at driving adoption of the online access route to primary care among the population in London? We examined the importance of each candidate treatment variable in separate analyses by contrasting online access adoption under low and high-adoption levels of each candidate variable. We used practice-level data from the population in North West London between 2018 and 2021 and employed the cross-validated targeted minimum loss-based Estimation to obtain variable importance estimates. We found that the top two drivers of online access adoption are the socioeconomic deprivation of the registered patient population and digital maturity of the practice, followed by quality of care measured by an inspection rating, helpfulness of receptionists, internet connectivity at the residential area of patients, practice’s training status and patients’ comorbidity profile. Our results highlight the potential areas of focus when designing interventions to facilitate online access adoption in primary care across the population. These areas of focus include mitigating the impact of patients’ socioeconomic deprivation on using digital technology and improving organisations’ digital maturity by investing in their infrastructure and workforce’s digital capability.

Recent years have seen an emergence of collaborative primary care models in the English National Health Service and other international health systems. Primary Care Networks (PCNs) were introduced in England in July 2019, marking the first time collaboration between general practices was incentivised through a nationwide policy. While participation was not mandatory, nearly all general practices joined a PCN, largely due to strong financial incentives. Our study aim was to estimate the impact of PCNs on emergency hospitalisations using an interrupted time series design. Quarterly data between October 2016 and March 2023 from the North West London Whole Systems Integrated Care dataset was used to construct two primary outcomes: all-cause and ambulatory care sensitive conditions (ACSC) emergency hospitalisations, as well as Accident and Emergency attendances, considered as a secondary outcome. Furthermore, we analysed whether the impact of PCNs varied based on practice characteristics. A reduction in all-cause and ACSC hospitalisations was observed following the PCNs’ introduction, until the start of the COVID-19 pandemic. The analysis also revealed a smaller reduction in ACSC hospitalisations among practices with more deprived patient populations and larger populations of patients with long-term conditions. While PCNs’ implementation appears to have led to a reduction in emergency hospitalisations in North West London, this effect was only observed in the very short term as it stopped with the COVID-19 pandemic. Future studies should examine the effect across England and evaluate their continued impact.

Introduction Manual investigation of falls incidents for quality improvement is time-consuming for clinical staff. Routine care delivery generates a large volume of relevant data in disparate systems, yet these data are seldom integrated and transformed into real-time, actionable insights for frontline staff. This protocol describes the co-design and testing of a safe mobility and falls informatics platform for automated, real-time insights to support the learning response to inpatient falls.Methods Underpinned by the learning health system model and human-centred design principles, this mixed-methods study will involve (1) collaboration between healthcare professionals, patients, data scientists and researchers to co-design a safe mobility and falls informatics platform; (2) co-production of natural language processing pipelines and integration with a user interface for automated, near-real-time insights and (3) platform usability testing. Platform features (data taxonomy and insights display) will be co-designed during workshops with lay partners and clinical staff. The data to be included in the informatics platform will be curated from electronic health records and incident reports within an existing secure data environment, with appropriate data access approvals and controls. Exploratory analysis of a preliminary static dataset will examine the variety (structured/unstructured), veracity (accuracy/completeness) and value (clinical utility) of the data. Based on these initial insights and further consultation with lay partners and clinical staff, a final data extraction template will be agreed. Natural language processing pipelines will be co-produced, clinically validated and integrated with QlikView. Prototype testing will be underpinned by the Technology Acceptance Model, comprising a validated survey and think-aloud interviews to inform platform optimisation.Ethics and dissemination This study protocol was approved by the National Institute for Health Research Imperial Biomedical Research Centre Data Access and Prioritisation Committee (Database: iCARE—Research Data Environment; REC reference: 21/SW/0120). Our dissemination plan includes presenting our findings to the National Falls Prevention Coordination Group, publication in peer-reviewed journals, conference presentations and sharing findings with patient groups most affected by falls in hospital.

BACKGROUND:Long Covid is a multisystem condition first identified in the Covid-19 pandemic, characterised by a wide range of symptoms including fatigue, breathlessness and cognitive impairment. Considerable disagreement exists in who is most at risk of developing long Covid, driven in part by incomplete coding of a long Covid diagnosis in medical records. OBJECTIVE:To describe the incidence and impact of long Covid. DESIGN:A retrospective observational cohort study. SETTING AND PARTICIPANTS:An integrated primary and secondary care dataset from North West London, covering over 2.7 million patients. Patients with long Covid were identified through clinical terms in their primary care records. MAIN VARIABLES STUDIED:Multivariate logistic regression was used to identify factors associated with having a long Covid diagnosis, while multivariate quantile regression was used to identify factors predicting the time a long Covid diagnosis was recorded. RESULTS:A total of 6078 patients were identified with a long Covid clinical term in their primary care record, 0.33% of the total registered adult population. Women, those aged 41-70 years or of Asian or mixed ethnicity, were more likely to have a recorded long Covid diagnosis, alongside those with pre-existing anxiety, asthma, depressive disorder or eczema and those living outside of the least or most socio-economically deprived areas. Men, those aged 41-70 years, or of black ethnicity, were diagnosed earlier in the pandemic, while those with depressive disorder were diagnosed later. DISCUSSION:Long Covid is poorly coded in primary care records, and significant differences exist between patient groups in the likelihood of receiving a long Covid diagnosis. A recorded long Covid diagnosis is more likely in women, some ethnic minority patients and those with pre-existing long-term conditions. CONCLUSION:The experience of patients with long Covid provides a crucial insight into inequities in access to timely care for complex multisystem conditions and the importance of effective health informatics practices to provide robust, timely analytical support for front line clinical services. PATIENT AND PUBLIC CONTRIBUTION:This study was co-designed, conducted and written in conjunction with people with long Covid.

The English National Health Service (NHS) strives for a fair, diverse, and inclusive workplace, but Black and Minority Ethnic (BME) representation in senior leadership roles remains limited. To address this, a large multi-hospital acute NHS Trust introduced an inclusive recruitment programme, requiring ethnically and gender diverse interview panels and a letter to the Chief Executive Officer (CEO) explaining hiring manager’s candidate choice. This generated large amount of valuable structured and free-text data, but manual analysis to derive actionable insights is challenging, limiting efforts to evaluate and improve such equality, diversity, and inclusion (EDI) recruitment initiatives. Using this routinely collected recruitment data from the programme between September 2021 to January 2024, we used natural language processing artificial intelligence techniques, triangulated with secondary data analysis, to evaluate the programme’s effectiveness in increasing the number of BME appointees to senior leadership roles. Multivariate logistic regression identified recruitment factors that influence the odds of BME candidates applying, being shortlisted or offered a role compared to white candidates. Topic and sentiment analysis revealed thematic trends and tone of candidate assessments, stratified by hiring manager and candidate characteristics. Normalised average interview scores were also compared by job grades and candidate characteristics. The requirement for hiring managers to write a letter to the CEO explaining recruitment decisions raised the odds of a BME candidate being offered a role by 1.7 times [95