Research publications
Browse through all publications from the Institute of Global Health Innovation, which our Patient Safety Research Collaboration is part of. This feed includes reports and research papers from our Centre.
Results
- Showing results for:
- Reset all filters
Search results
-
Journal articleBracey S, Ainsworth B, Alderman J, et al., 2026,
The 'Hippocratic Oath' for AI-based clinical decision support systems
, BMC Medical Informatics and Decision Making, Vol: 26, ISSN: 1472-6947BackgroundThe implementation of Artificial Intelligence assisted Clinical Decision Support Systems (AI-CDSS) shows significant potential to improve healthcare. However, implementing AI-CDSS has many associated challenges. This article introduces the ‘Hippocratic Oath’ for AI which promotes safe and effective AI-CDSS development and implementation.MethodsThis paper summarises discussions which took place during the Turing-Roche Clinical AI Interest Group Joint Workshop. The workshop began with scoping lectures from AI experts, leading into focus group discussions of key themes surrounding AI-CDSS implementation. These include the ethics, trust, evaluation, regulation, human factors and challenges involved with implementing AI-CDSS into healthcare settings. Focus group outcomes, alongside insight from lectures, were used to formulate the arguments in this paper.ResultsThis article presents a consensus definition of AI-CDSS and outlines a comprehensive table of implementation challenges alongside mitigation measures. It introduces the ‘Hippocratic Oath for AI’ and discusses its potential to promote safe and effective AI-CDSS implementation through addressing human factors and explainability.ConclusionsThe ‘Hippocratic Oath for AI’, can be used by AI-CDSS implementers and developers as a framework to mitigate challenges involved with implementing AI-CDSS into healthcare settings. This framework is likely to promote safe and effective implementation and maximise HCP uptake of the AI-CDSS. Through facilitating AI-CDSS use, this oath can transform health care practice via reducing medical errors, healthcare costs and improving patient outcomes.
-
Journal articleAnderson M, O'Neill E, Ljungqvist G, et al., 2026,
Designing eligibility and evaluation criteria for antibacterial pull incentives: a comparative review of implemented and emerging initiatives
, Lancet Regional Health Europe, Vol: 70Background: Pull incentives aim to improve the commercial viability of antibacterials, but the criteria determining which products and companies qualify have not been compared across countries. We compared product and company eligibility and evaluation criteria across these incentives. Methods: We conducted an umbrella review of MEDLINE and Embase for English-language reviews of pull incentives targeting antibacterial innovation and access (2014 to 14 October 2024), supplemented by grey literature searches and the Global AMR R&D Hub dashboard (December 2025). Two reviewers independently extracted criteria from official documentation, using a framework of six product domains (high-priority medical need, relative effectiveness, unmet clinical need, innovative characteristics, health-system impact, and other) and five company domains (antibacterial sustainability, patient access, environmental health, economic criteria, and other). Findings: We identified 28 pull incentives: 20 implemented (UK, Sweden, Germany, France, Italy, US, Japan, and EU) and 8 proposed or in development (US, Japan, Canada, Australia, Switzerland, and EU). As mechanisms overlapped within countries, the 20 implemented incentives were analysed as nine groups. All nine targeted high-priority medical need and seven (78%) included unmet clinical need. Relative effectiveness featured in six (67%), economic criteria in five (56%), and innovative characteristics, health-system impact, patient access, and antibacterial sustainability in four each (44%); environmental health in two (22%). The UK Subscription Model applied all 11 domains and Sweden's Annual Revenue Guarantee eight (73%). Priority pathogen lists were widely referenced but varied in breadth. Interpretation: Eligibility and evaluation criteria vary substantially across countries in scope and specificity. Company-level obligations on stewardship, access, and environmental safeguards are applied inconsistently, and most comprehensive in the
-
Journal articleLi Z, Zhou Z, Lou H, et al., 2026,
A novel pouch-based tension sensor array for soft Cable-Driven Parallel Surgical Robot
, Measurement Journal of the International Measurement Confederation, Vol: 288, ISSN: 0263-2241Endoscopic Submucosal Dissection (ESD) is an advanced, minimally invasive procedure for the removal of tumors or lesions from the gastrointestinal tract. While robot-assisted surgery has enhanced the safety and efficacy of ESD, a significant limitation of current robot systems is the lack of force feedback capabilities. This deficiency increases the risk of excessive force being exerted on tissue. To address this challenge, this study proposes a soft Cable-Driven Parallel Robot (CDPR) integrated with a novel pouch-based sensor array that is disposable and has simple structure. This system utilizes hydraulic pressure sensors and pouch structures to estimate both cable tension and the three-dimensional forces at the end-effector, thereby providing the surgeon with crucial force feedback. During an ex vivo experiment, the mean absolute errors between the estimated forces and the ground truth values were 0.0267 N, 0.0659 N, and 0.0509 N for the x, y, and z axes, respectively. These results demonstrate the potential of the proposed CDPR for future clinical applications.
-
Journal articleIbrahim R, Durand C, Macleod M, et al., 2026,
Implementation, adoption and impact of clinical decision support system for antibiotic prescribing in primary care: a systematic review
, Journal of Antimicrobial Chemotherapy, Vol: 81, ISSN: 0305-7453BackgroundClinical decision support systems (CDSS), computerized tools that assist clinicians in making guideline-based decisions, may support antimicrobial prescribing in primary care.MethodsStudies on CDSS implementation, use and outcomes in primary care on PubMed/MEDLINE and Embase were included up to June 2025.ResultsOf the 64 full-text articles assessed, 40 were included. Most were multicentric (n = 33, 82.5%) and conducted in high-income countries (n = 33, 82.5%), and 19 were randomized controlled trials (47.5%). CDSS mainly targeted respiratory infections (n = 24, 60%) and were integrated into electronic health records in 24 studies (60%). Implementation strategies were assessed in 19 studies with almost all using a multimodal approach, most commonly involving modification of health record systems (n = 18, 45%) and audit and feedback (n = 17, 42.5%). Adoption was highly variable: among the 11 studies reporting use, CDSS were used in more than half of consultations in only five studies. Among the 32 studies assessing clinical outcomes, 17 reported reduced overall antibiotic prescribing. Barriers to CDSS use included perceptions of redundancy, time consumption, language limitations, alert fatigue, technical issues and concerns about negative impacts on patient–doctors trust.ConclusionCDSS may improve guideline-concordant antibiotic prescribing in primary care, but wide variability in implementation and real-world adoption limits generalizability and scalability
-
Journal articleSmith KA, Siafis S, McCutcheon RA, et al., 2026,
Improving access to antipsychotic medications for schizophrenia in Ethiopia, Nigeria, Rwanda, and South Africa: an evidence-based global consensus.
, Lancet Psychiatry, Vol: 13, Pages: 884-896There are disparities in access to antipsychotics for schizophrenia across different country settings. Improving access to a wider and more equitable range of medications in low-income and middle-income countries is a priority. A multidisciplinary team of international experts, including individuals with lived experience, appraised the most relevant and recent information on antipsychotics in schizophrenia and contextualised it to four African countries (Ethiopia, Nigeria, Rwanda, and South Africa) using a validated consensus methodology. We recommended a list of drugs to prioritise to guide clinical implementation and research, and market shaping. We identified key evidence gaps: little of the existing evidence comes from the countries of interest, trials generally involve highly selected populations, and the complexity of real-world settings is not reflected. However, this methodology highlights a route forward to prioritise the best available evidence on pharmacological treatments for schizophrenia at a global scale, which could also be applied to treatments for other mental health conditions.
-
Journal articleSchmidt J, Carter AW, McGuire A, et al., 2026,
A systematic review of economic evidence of artificial intelligence in healthcare.
, Health Policy, Vol: 172BACKGROUND: Ambitious claims suggest AI could save $200-360 billion annually in US healthcare and €212 billion in Europe, though the empirical evidence base supporting these projections remains unclear. OBJECTIVE: This systematic review seeks to synthesise the available economic evidence of AI technologies in healthcare and contextualise the available economic evidence against the broader policy expectations surrounding the economic impact of AI in healthcare. METHODS: We searched MEDLINE, Embase, Global Health, PsycINFO, and Cochrane Central for scientific sources. The JBI dominance ranking matrix was used to compare and interpret the results of the included economic evaluations. Methodological quality was assessed using the JBI Critical Appraisal tool of Economic Evaluations and the CHEERS-AI reporting checklist. RESULTS: We identified 16,430 academic records and 1,593 grey literature records, of which 91 records met the inclusion criteria, representing 98 unique evaluations. Of these, 36% demonstrate a clear health economic preference for the AI technology, which increased to 44% when only considering the 51 high-quality studies. AI interventions were mainly designed for healthcare providers, with ophthalmology and oncology being the most common. CONCLUSIONS: The high-quality studies show definite potential for positive cost-effectiveness and economic impact of AI technologies, though we cannot yet support the ambitious claims that AI technologies can translate to hundreds of billions in cost-savings. When combining our findings with those of randomised controlled trials evaluating AI technologies in clinical practice, it becomes evident that AI technologies frequently yield improvements in terms of clinical and economic outcomes or match the current standard of care.
-
Journal articleGupta A, Prociuk D, Russo A, et al., 2026,
Automatic Conversion of NICE Guidelines to an Executable Computational Model Using Large Language Models.
, Learn Health Syst, Vol: 10INTRODUCTION: The UK National Institute for Health and Care Excellence (NICE) produce guidelines that provide evidence-based recommendations to support clinical care across England and Wales, but remain available in unstructured natural language form. Converting these guidelines into computable, logically coherent representations is an active area of research yet existing approaches typically focus on individual diseases, require substantial manual encoding, and do not scale. Recent advances in large language models offer an opportunity to automate much of this translation process. METHODS: We present an end-to-end approach that automatically converts textual clinical guidelines into an executable model capable of generating explainable patient-specific recommendations. Our approach uses a stepwise LLM-based transformation with in-context examples that can be customized to the guideline of your choice. Each step generates human-inspectable intermediate artifacts, ensuring full transparency and modifiability. We apply the approach to both pancreatic and lung cancer NICE guidelines and use expert human review to assess the alignment of the produced rules as well as evaluating the executable model over 20 pancreatic cancer patient vignettes. RESULTS: Human experts review demonstrated strong alignment between the natural language guidelines and the generated executable models, with the majority of guideline recommendations translated correctly. Most discrepancies involved partial omissions of specific details rather than incorrect logic, and instances of hallucinated or fundamentally incorrect rules were rare. When executed on the vignettes, the resulting executable models produced patient-specific recommendations with an F1 score of 82.5%. CONCLUSION: This work demonstrates that LLMs can be used to automatically transform natural language NICE guidelines into interpretable and executable models. The models preserve guideline structure, allow transparent inspection and
-
Journal articleAnyaibe S, Anderson AK, Domfe CA, et al., 2026,
Eating architecture components and their associations with BMI in urban and rural Ghanaian mothers, fathers, children, and adolescents, assessed using a wearable camera: A cross-sectional study.
, Chronobiol Int, Vol: 43, Pages: 1524-1537Eating architecture - timing, frequency, and size of meals and snacks - affects metabolism, but data from low-middle-income countries (LMICs) are scarce. We examined eating architecture (timing, frequency, and size of eating occasions) among Ghanaian households and its relationship with BMI. Thirty rural and 30 urban Ghanaian households participated. A wearable camera on eyeglasses captured dietary intake over 2 weekdays and 1 weekend day. Custom software was used for nutritional analysis and identifying eating episodes. Meals were distinguished from snacks by time, context, and items consumed. Eating frequencies were 2.4-2.7 x/d, and eating windows were 6.5-9.5 h/d. Children ate more frequently than other household members and had longer eating windows than fathers. Eating and energy intake peaked at 8:00 h and 12:00 h, respectively. Only 56% of urban and 72% of rural members snacked. Meal energy was similar across locations, but snack energy was higher among urban versus rural households, positively associated with BMI among urban mothers. Ghanaian eating architecture varied by location, household member, and differed from high-income countries. Only the size component of eating architecture, namely snack energy, was associated with BMI, and only in urban mothers. Our findings highlight the need to address nutritional disparities in Ghana and other LMICs.
-
Journal articleRahman RMT, Alyacoubi S, Leff DR, et al., 2026,
Objective Measurement of the Learning Curve in Live Laparoscopic Surgery: A Scoping Review of Methods.
, J Surg Educ, Vol: 83OBJECTIVE: This scoping review aims to provide an overview of the methodology used when defining the learning curve (LC) in live laparoscopic surgery. DESIGN: This review was performed in line with the Preferred Reporting Items for Systematic Reviews and Meta Analyses extension for Scoping Reviews (PRISMA-ScR) guidelines. English-language articles were included systematically using Boolean operators for search string combinations. Selected studies were qualitatively analyzed for LC measurement methodology, including statistical approaches. Additional analysis was undertaken by grouping LC metrics into intrinsic, surrogate, and outcome-based. SETTING: Only original publications in the English language were included. No date restrictions were applied. Eligible studies were those that measured LC in laparoscopic cases (i.e., not simulation studies) regardless of their surgical specialty. Robotic-assisted laparoscopic operations were also included. Articles analyzing non-laparoscopic procedures were not included. RESULTS: A total of 87 articles were extracted for review. The majority (45/87) were conducted within general surgery alongside gynecology, vascular, pediatrics, and urology. A total of 50 studies analyzed the LC for a new technology or novel technique. Multiple metrics were used to quantify LC, with the most common being operative time (n = 75), followed by complications (n = 31). Only 3 studies exclusively used intrinsic performance measures for LC analysis. There was significant heterogeneity in the statistical analysis and description of LC. The majority of studies (51/87) grouped cases temporally. Nineteen studies used a Cumulative Sum (CUSUM) statistical analysis. CONCLUSIONS: Measurement of the LC in live laparoscopic surgery has significant potential for training and appraisal of new techniques. Currently, the time-intensive nature of LC measurement limits its widespread adoption. However, newer techniques used alongside real-time measurement of metrics
-
Journal articleLam K, Yiu A, 2026,
Video Analysis in Surgery: Lessons From the Pitch?
, Surg Innov, Vol: 33, Pages: 537-538
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.