Hajira Dambha-Miller
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Hajira Dambha-Miller has a total of 291 co-authors. Here are the top co-authors by number of shared publications:
- Lucy Smith
Primary Care Research Centre, University of Southampton, Southampton, UK
22 shared publications - Glenn Simpson
University of Southampton
20 shared publications - Beth Stuart
Queen Mary University of London
12 shared publications - Nazrul Islam
Primary Care Research Centre, Faculty of Medicine, University of Southampton, Southampton, UK
12 shared publications - Andrew Farmer
Nuffield Department of Primary Care Health Sciences, University of Oxford, Oxford, United Kingdom
10 shared publications - Lysanne Veerle Michels
University of Southampton, UK
10 shared publications
Latest Publications
- Open Access
- 6 October 2026
- Published by Springer Nature in BMC Psychiatry
Mapping equity-relevant inputs of AI mental health tools to routine primary care data: a population-based study of 7.19 million adults with multiple long-term conditions in England
Background: Artificial intelligence-enabled tools are increasingly used to support mental health diagnosis, triage and clinical decision-making. However, limited evidence exists on whether these tools incorporate equity-relevant population characteristics, particularly among people living with multiple long-term conditions (MLTC), who often experience complex health needs and structural inequalities. Methods: We conducted a cross-sectional analysis using linked Clinical Practice Research Datalink Gold and Aurum primary care data for 7,186,890 adults aged ≥ 18 years with MLTC in England between 1987 and 2020. Five artificial intelligence-based mental health diagnostic or triage tools identified through a scoping review were operationalised using routinely collected primary care variables. Many symptom-level psychiatric inputs central to the original models could not be captured in routine primary care data and were excluded. Reported tool inputs were mapped to sociodemographic, clinical and social-care factors, including deprivation, ethnicity, age, sex, disability, place of residence and social care involvement, informed by the CORE20PLUS5 and PROGRESS-PLUS frameworks. We calculated an Equity-Variable Coverage Index to quantify the proportion of equity-relevant variables each tool includes as inputs, a measure of input coverage rather than demonstrated fairness. To illustrate the empirical importance of these characteristics, we fitted multivariable logistic regression models for hospital admission, a proxy outcome influenced by system-level and clinical factors, with leave-one-out removal of each equity-related variable. Results: Equity domain coverage varied substantially across tools, ranging from 13.0% to 56.5%. Clinical characteristics were more consistently represented than sociodemographic or social-care factors. Drug and alcohol misuse was included in all five tools, whereas age, ethnicity, deprivation, disability status and social care involvement were inconsistently incorporated. Removing ethnicity and age produced the largest reductions in model explanatory performance, followed by outpatient attendance, Accident and Emergency attendance, and drug and alcohol misuse, indicating that several under-represented equity-relevant variables were associated with hospital admission among adults with MLTC. Conclusions: Current artificial intelligence-enabled mental health tools show substantial and variable gaps in their coverage of equity-relevant population characteristics available in routine UK primary care data. Embedding equity considerations into tool design and routinely evaluating model inputs and performance using population-scale data will support fair and responsible deployment of mental health artificial intelligence.
- 6 October 2026
- Published by Springer Nature in BMC Psychiatry
Open AccessMapping equity-relevant inputs of AI mental health tools to routine primary care data: a population-based study of 7.19 million adults with multiple long-term conditions in England
Background: Artificial intelligence-enabled tools are increasingly used to support mental health diagnosis, triage and clinical decision-making. However, limited evidence exists on whether these tools incorporate equity-relevant population characteristics, particularly among people living with multiple long-term conditions (MLTC), who often experience complex health needs and structural inequalities. Methods: We conducted a cross-sectional analysis using linked Clinical Practice Research Datalink Gold and Aurum primary care data for 7,186,890 adults aged ≥ 18 years with MLTC in England between 1987 and 2020. Five artificial intelligence-based mental health diagnostic or triage tools identified through a scoping review were operationalised using routinely collected primary care variables. Many symptom-level psychiatric inputs central to the original models could not be captured in routine primary care data and were excluded. Reported tool inputs were mapped to sociodemographic, clinical and social-care factors, including deprivation, ethnicity, age, sex, disability, place of residence and social care involvement, informed by the CORE20PLUS5 and PROGRESS-PLUS frameworks. We calculated an Equity-Variable Coverage Index to quantify the proportion of equity-relevant variables each tool includes as inputs, a measure of input coverage rather than demonstrated fairness. To illustrate the empirical importance of these characteristics, we fitted multivariable logistic regression models for hospital admission, a proxy outcome influenced by system-level and clinical factors, with leave-one-out removal of each equity-related variable. Results: Equity domain coverage varied substantially across tools, ranging from 13.0% to 56.5%. Clinical characteristics were more consistently represented than sociodemographic or social-care factors. Drug and alcohol misuse was included in all five tools, whereas age, ethnicity, deprivation, disability status and social care involvement were inconsistently incorporated. Removing ethnicity and age produced the largest reductions in model explanatory performance, followed by outpatient attendance, Accident and Emergency attendance, and drug and alcohol misuse, indicating that several under-represented equity-relevant variables were associated with hospital admission among adults with MLTC. Conclusions: Current artificial intelligence-enabled mental health tools show substantial and variable gaps in their coverage of equity-relevant population characteristics available in routine UK primary care data. Embedding equity considerations into tool design and routinely evaluating model inputs and performance using population-scale data will support fair and responsible deployment of mental health artificial intelligence.
- Open Access
- 28 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Can Healthcare Afford to Innovate? Rethinking Sustainability in the Age of Artificial Intelligence (Preprint)
UNSTRUCTURED Artificial intelligence (AI) is increasingly integrated into healthcare, with evaluation extending beyond model performance to clinical utility, safety, fairness, economic value and implementation. Environmental sustainability has received comparatively less attention despite the computational, energy, water and hardware requirements associated with AI. Recent approaches have begun to quantify these direct environmental impacts, including lifecycle-informed methods for reporting energy consumption, carbon emissions, water use and embodied hardware emissions. We argue that direct measurement alone is insufficient for healthcare AI because environmental impact also depends on how a technology changes the delivery of care. Evidence from autonomous AI for diabetic eye examination, for example, suggests that pathway-level emissions may differ substantially when patient travel and specialist activity are incorporated alongside computational emissions. AI may similarly reduce resource use by preventing investigations, consultations or hospital care, but may increase it where alerts, monitoring and additional clinical activity are added without replacing existing processes. Environmental evaluation of healthcare AI should therefore consider both the resources required to develop and deploy a technology and its measured consequences for healthcare delivery at the intended scale of implementation. Sustainability should form part of responsible healthcare AI alongside clinical utility, safety, fairness and economic value, with greater attention to real-world resource use during evaluation, implementation and procurement.
- 28 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessCan Healthcare Afford to Innovate? Rethinking Sustainability in the Age of Artificial Intelligence (Preprint)
UNSTRUCTURED Artificial intelligence (AI) is increasingly integrated into healthcare, with evaluation extending beyond model performance to clinical utility, safety, fairness, economic value and implementation. Environmental sustainability has received comparatively less attention despite the computational, energy, water and hardware requirements associated with AI. Recent approaches have begun to quantify these direct environmental impacts, including lifecycle-informed methods for reporting energy consumption, carbon emissions, water use and embodied hardware emissions. We argue that direct measurement alone is insufficient for healthcare AI because environmental impact also depends on how a technology changes the delivery of care. Evidence from autonomous AI for diabetic eye examination, for example, suggests that pathway-level emissions may differ substantially when patient travel and specialist activity are incorporated alongside computational emissions. AI may similarly reduce resource use by preventing investigations, consultations or hospital care, but may increase it where alerts, monitoring and additional clinical activity are added without replacing existing processes. Environmental evaluation of healthcare AI should therefore consider both the resources required to develop and deploy a technology and its measured consequences for healthcare delivery at the intended scale of implementation. Sustainability should form part of responsible healthcare AI alongside clinical utility, safety, fairness and economic value, with greater attention to real-world resource use during evaluation, implementation and procurement.
- Open Access
- 21 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Implementing Clinical Artificial Intelligence as a Continuous Lifecycle (Preprint)
UNSTRUCTURED Clinical artificial intelligence (AI) requires sustained integration within health care systems. AI performance and use may change following model updates, shifts in population characteristics and data distributions, changes to clinical pathways, and adaptation by health care professionals and patients. Implementation therefore extends across the clinical AI lifecycle. We identify four priorities for sustainable implementation: socio-technical integration within clinical workflows; lifecycle governance and accountability; continuous generation of evidence on clinical and system value; and longitudinal assessment of equity. These priorities require defined clinical objectives, accountable governance, prospective evaluation, post-deployment monitoring, and mechanisms for adaptation or withdrawal. Implementation science can support this process by examining interactions between AI technologies, users, organizations, and wider health care systems over time. A lifecycle approach provides a basis for integrating clinical AI safely, effectively, equitably, and sustainably within routine health care.
- 21 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessImplementing Clinical Artificial Intelligence as a Continuous Lifecycle (Preprint)
UNSTRUCTURED Clinical artificial intelligence (AI) requires sustained integration within health care systems. AI performance and use may change following model updates, shifts in population characteristics and data distributions, changes to clinical pathways, and adaptation by health care professionals and patients. Implementation therefore extends across the clinical AI lifecycle. We identify four priorities for sustainable implementation: socio-technical integration within clinical workflows; lifecycle governance and accountability; continuous generation of evidence on clinical and system value; and longitudinal assessment of equity. These priorities require defined clinical objectives, accountable governance, prospective evaluation, post-deployment monitoring, and mechanisms for adaptation or withdrawal. Implementation science can support this process by examining interactions between AI technologies, users, organizations, and wider health care systems over time. A lifecycle approach provides a basis for integrating clinical AI safely, effectively, equitably, and sustainably within routine health care.
- Open Access
- 18 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
From software to hardware: The next frontier of AI interoperability in healthcare (Preprint)
UNSTRUCTURED Artificial intelligence (AI) has accelerated progress in healthcare interoperability, largely through improved exchange of information across electronic health records, digital platforms, and clinical systems. Emerging developments suggest a potential next stage: interoperability between AI systems and physical hardware. Recent advances, including model-agnostic standards that enable AI agents to coordinate diverse devices through common interfaces, raise the possibility of healthcare ecosystems in which AI not only interprets information but also orchestrates actions across connected technologies. In this viewpoint, we explore the implications of this transition from software-to-software interoperability toward AI-enabled hardware–software interoperability. We discuss how such architectures could reshape National Health Service (NHS) workflows, reduce fragmentation across care pathways, support management of multimorbidity, and lower barriers to innovation. We also examine important challenges, including patient acceptability, workforce readiness, digital literacy, health inequalities, commercial concentration, accountability, and safety. We argue that interoperability should be understood not only as a technical capability but also as a sociotechnical transformation that redistributes clinical work, decision-making authority, and responsibility across patients, professionals, organisations, and technology providers. As AI systems increasingly connect information, reasoning, and physical action, conventional approaches to evaluation and regulation of individual software products or devices may become insufficient. Healthcare organisations, regulators, and researchers should therefore begin preparing for a future in which the primary challenge is not whether individual components are safe and effective, but whether the interconnected systems created by combining them remain clinically useful, equitable, resilient, transparent, and accountable
- 18 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessFrom software to hardware: The next frontier of AI interoperability in healthcare (Preprint)
UNSTRUCTURED Artificial intelligence (AI) has accelerated progress in healthcare interoperability, largely through improved exchange of information across electronic health records, digital platforms, and clinical systems. Emerging developments suggest a potential next stage: interoperability between AI systems and physical hardware. Recent advances, including model-agnostic standards that enable AI agents to coordinate diverse devices through common interfaces, raise the possibility of healthcare ecosystems in which AI not only interprets information but also orchestrates actions across connected technologies. In this viewpoint, we explore the implications of this transition from software-to-software interoperability toward AI-enabled hardware–software interoperability. We discuss how such architectures could reshape National Health Service (NHS) workflows, reduce fragmentation across care pathways, support management of multimorbidity, and lower barriers to innovation. We also examine important challenges, including patient acceptability, workforce readiness, digital literacy, health inequalities, commercial concentration, accountability, and safety. We argue that interoperability should be understood not only as a technical capability but also as a sociotechnical transformation that redistributes clinical work, decision-making authority, and responsibility across patients, professionals, organisations, and technology providers. As AI systems increasingly connect information, reasoning, and physical action, conventional approaches to evaluation and regulation of individual software products or devices may become insufficient. Healthcare organisations, regulators, and researchers should therefore begin preparing for a future in which the primary challenge is not whether individual components are safe and effective, but whether the interconnected systems created by combining them remain clinically useful, equitable, resilient, transparent, and accountable
- Open Access
- 18 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Are Large Language Models amplifying bias in medical science? (Preprint)
UNSTRUCTURED Large Language Models (LLMs) offer unprecedented efficiency across medical research, yet current evaluation focuses almost exclusively on output accuracy, neglecting critical issues of fairness, bias, and global representation. Because LLMs retrieve from and synthesize an evidence base already skewed toward high-income populations, their widespread adoption creates a feedback loop of recursive amplification that repeatedly surfaces dominant data while filtering out novel hypotheses and evidence from under-researched groups. Even if individual outputs remain factually accurate, this process risks creating an artificial scientific consensus that fails to reflect global health needs. Because standard disclosure rules cannot track how AI alters the broader literature landscape, evaluation paradigms must shift from checking study-level accuracy to measuring field-level representation. Establishing longitudinal literature-surveillance frameworks to monitor distributional skew, source diversity, and demographic representation, alongside strict reporting standards for model usage, is essential to prevent LLMs from exacerbating structural health inequities and undermining scientific validity.
- 18 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessAre Large Language Models amplifying bias in medical science? (Preprint)
UNSTRUCTURED Large Language Models (LLMs) offer unprecedented efficiency across medical research, yet current evaluation focuses almost exclusively on output accuracy, neglecting critical issues of fairness, bias, and global representation. Because LLMs retrieve from and synthesize an evidence base already skewed toward high-income populations, their widespread adoption creates a feedback loop of recursive amplification that repeatedly surfaces dominant data while filtering out novel hypotheses and evidence from under-researched groups. Even if individual outputs remain factually accurate, this process risks creating an artificial scientific consensus that fails to reflect global health needs. Because standard disclosure rules cannot track how AI alters the broader literature landscape, evaluation paradigms must shift from checking study-level accuracy to measuring field-level representation. Establishing longitudinal literature-surveillance frameworks to monitor distributional skew, source diversity, and demographic representation, alongside strict reporting standards for model usage, is essential to prevent LLMs from exacerbating structural health inequities and undermining scientific validity.
- Open Access
- 17 September 2026
- Published by JMIR Publications Inc. in Journal of Medical Internet Research
- Vol. 28, e100654
When Innovation Leaves People Behind: Reframing Accountability in Commercial Digital Health
Digital technologies are increasingly integrated across health care systems; however, accumulating evidence shows that these tools often perform unevenly across population groups. Biases in AI-enabled tools can reinforce existing health inequities, particularly when systems are developed and validated using datasets that exclude underserved populations. Studies demonstrate systematic underestimation of illness severity, diagnostic inaccuracies, and measurement bias affecting ethnic minority groups, rural populations, people with disabilities, people experiencing homelessness, and other groups with limited access to health care. These disparities highlight structural vulnerabilities in the data that inform machine learning systems, where socially patterned access to care shapes what is recorded and therefore learned. Commercial adoption patterns further exacerbate inequities due to socioeconomic gradients in digital engagement and a lack of accountability measures for adherence to regulatory frameworks within National Health Service procurement processes. In this Viewpoint, we argue that ensuring accountability in digital health care requires transparent reporting of subgroup performance, representative development of datasets, and ongoing monitoring of distributional impacts. Achieving equitable and reliable digital health care demands regulatory and methodological standards that prioritize fairness and generalizability within the commercialization sector and are monitored across the pre- and postdeployment life cycle.
- 17 September 2026
- Published by JMIR Publications Inc. in Journal of Medical Internet Research
- Vol. 28, e100654
Open AccessWhen Innovation Leaves People Behind: Reframing Accountability in Commercial Digital Health
Digital technologies are increasingly integrated across health care systems; however, accumulating evidence shows that these tools often perform unevenly across population groups. Biases in AI-enabled tools can reinforce existing health inequities, particularly when systems are developed and validated using datasets that exclude underserved populations. Studies demonstrate systematic underestimation of illness severity, diagnostic inaccuracies, and measurement bias affecting ethnic minority groups, rural populations, people with disabilities, people experiencing homelessness, and other groups with limited access to health care. These disparities highlight structural vulnerabilities in the data that inform machine learning systems, where socially patterned access to care shapes what is recorded and therefore learned. Commercial adoption patterns further exacerbate inequities due to socioeconomic gradients in digital engagement and a lack of accountability measures for adherence to regulatory frameworks within National Health Service procurement processes. In this Viewpoint, we argue that ensuring accountability in digital health care requires transparent reporting of subgroup performance, representative development of datasets, and ongoing monitoring of distributional impacts. Achieving equitable and reliable digital health care demands regulatory and methodological standards that prioritize fairness and generalizability within the commercialization sector and are monitored across the pre- and postdeployment life cycle.
- Open Access
- 5 September 2026
- Published by Springer Nature in Air Quality, Atmosphere & Health
- Vol. 19 (9) , 201
Application of Artificial Intelligence in analysing meteorological data for health research: A rapid review
Introduction: Artificial intelligence (AI) has rapidly advanced as a key analytical tool for processing complex datasets across disciplines, including environmental and health research. Meteorological data is increasingly used to understand and mitigate health risks, often linked to climate change. This rapid review aims to synthesise studies involving AI for meteorological data applications in health research to better understand its use. Methods: PubMed, Web of Science, and Scopus were systematically searched from 2020 to 2025 following a standardised framework in line with PRISMA-RR guidelines. Eligible studies included any empirical design involving human-related health research where AI techniques were applied to meteorological data. Two reviewers independently screened studies, extracted data, and synthesised findings narratively. Results: Twelve studies met eligibility criteria. Despite heterogeneity in study design and sample size, most examined the impact of extreme pollution levels and temperature variations on the prevalence and severity of respiratory, bacterial, and other diseases. AI methods primarily included Random Forest models, as well as time-series and clustering analyses. Model performance was commonly evaluated using sensitivity; however, methodological justification was often insufficiently recorded. Conclusion: Overall, findings suggest that incorporating meteorological variables enhances the prediction of health outcomes, although detailed population characteristics were frequently underreported. This restricts the generalisability and applicability of AI-driven health models incorporating meteorological data, with stronger methodological rigour and clearer reporting standards needed to support reliable future development.
- 5 September 2026
- Published by Springer Nature in Air Quality, Atmosphere & Health
- Vol. 19 (9) , 201
Open AccessApplication of Artificial Intelligence in analysing meteorological data for health research: A rapid review
Introduction: Artificial intelligence (AI) has rapidly advanced as a key analytical tool for processing complex datasets across disciplines, including environmental and health research. Meteorological data is increasingly used to understand and mitigate health risks, often linked to climate change. This rapid review aims to synthesise studies involving AI for meteorological data applications in health research to better understand its use. Methods: PubMed, Web of Science, and Scopus were systematically searched from 2020 to 2025 following a standardised framework in line with PRISMA-RR guidelines. Eligible studies included any empirical design involving human-related health research where AI techniques were applied to meteorological data. Two reviewers independently screened studies, extracted data, and synthesised findings narratively. Results: Twelve studies met eligibility criteria. Despite heterogeneity in study design and sample size, most examined the impact of extreme pollution levels and temperature variations on the prevalence and severity of respiratory, bacterial, and other diseases. AI methods primarily included Random Forest models, as well as time-series and clustering analyses. Model performance was commonly evaluated using sensitivity; however, methodological justification was often insufficiently recorded. Conclusion: Overall, findings suggest that incorporating meteorological variables enhances the prediction of health outcomes, although detailed population characteristics were frequently underreported. This restricts the generalisability and applicability of AI-driven health models incorporating meteorological data, with stronger methodological rigour and clearer reporting standards needed to support reliable future development.
- Open Access
- 18 August 2026
- Published by National Institute for Health and Care Research
- Vol. 6, 119
Developing and optimising an intervention prototype for addressing health and social care need in multimorbidity
Research Aim The objective of this project was to translate artificial intelligence (AI) data models into a practical, low-burden primary care intervention. Specifically, the project aimed to co-design and refine an early-stage tool that enables healthcare teams to easily identify and address the social care needs of patients living with multiple long-term conditions (MLTCs). Background An increasing number of people live with multimorbidity—the presence of two or more chronic health conditions such as diabetes, heart disease, or depression. These illnesses frequently intersect with complex socio-economic challenges, including financial strain, unstable housing, and social isolation. While these social care needs profoundly impact clinical outcomes and overall quality of life, they are rarely identified during standard, time-constrained medical consultations. This project builds directly on the foundational, NIHR-funded AIM programme. The AIM study utilised machine learning to analyse anonymous health records, successfully identifying four distinct population clusters based on shared health and social care patterns. While stakeholders recognised the value of this AI data, they emphasised that algorithms must support, rather than replace, human clinical interactions. Furthermore, a comprehensive scoping review of nearly two thousand published studies revealed that existing digital tools focus almost exclusively on medical metrics like symptom tracking, often increasing the treatment burden on vulnerable patients. Consequently, a clear policy and clinical gap exists for an intervention that addresses broader life priorities. What We Did To bridge the gap between data science and clinical practice, the research team used a multi-stage, qualitative approach: Stakeholder Interviews: In-depth qualitative interviews were conducted with 24 individuals living with MLTCs and 20 health and social care providers to understand real-world boundaries, resource constraints, and systemic expectations. Prototype Co-Design: Synthesising these insights with literature reviews, the team generated a non-digital paper prototype. The prototype consisted of a one-page consultation prompt for clinicians and a self-navigation priority checklist for patients. Iterative Usability Testing: The prototype underwent "think-aloud" usability testing with 16 patients and 15 primary care professionals. Feedback was analysed dynamically throughout the fieldwork, allowing for rapid, recorded adjustments to the prototype's layout, flow, and terminology to minimise cognitive load. What We Found The research confirmed that patients view trusted primary care settings as appropriate hubs for initiating social care conversations, provided the approach is non-judgemental. However, clinicians expressed concern over severe time constraints and limited local referral pathways. The iterative testing crystallised six foundational user requirements for integrating health and social care tools: Patient Autonomy: Individuals must retain control over which social priorities to address first. Transparent Intent: The purpose and limitations of the tool must be stated explicitly up front. Low Treatment Burden: Visual layouts and wording must remain highly simplified and calming. Validation of Past Experiences: The tool must sensitively account for patients who have previously struggled to access care. Clear Operational Roles: Primary care staff must have clearly defined boundaries regarding their social care responsibilities. System Integration: Any final intervention must align seamlessly with existing NHS electronic health records and standard appointment durations. Patient and Public Involvement and Engagement (PPIE) Patient and public involvement was fundamental to ensuring the ethical application of AI-derived insights. PPI contributors collaborated closely with the research team to design all participant-facing study literature. Utilising a digital-first, visually accessible methodology, public advisors reviewed and refined the prototype's language. This direct involvement ensured that the wording was supportive, accessible to individuals with varying levels of literacy, and entirely free of institutional or stigmatising jargon. Conclusion By anchoring advanced algorithmic modelling in human-centred design, this project successfully created a viable framework to address social care needs within primary care. The research demonstrates that complex population data can be translated into clear, actionable, and low-burden clinical tools. This approach offers an equitable pathway toward integrated, holistic care that improves patient wellbeing without exacerbating NHS staff shortages or consultation pressures. Next Steps and Future Plans Backed by extended funding from the original NIHR grant, the project has entered a commercial partnership to translate the paper prototype into a live, NHS-compatible digital application named IndependAble. Future work will focus on securing NIHR Research for Social Care funding to launch a robust, multi-site evaluation. This upcoming trial will formally measure IndependAble's clinical utility, its impact on social care identification rates, and its operational sustainability within busy primary care services.
- 18 August 2026
- Published by National Institute for Health and Care Research
- Vol. 6, 119
Open AccessDeveloping and optimising an intervention prototype for addressing health and social care need in multimorbidity
Research Aim The objective of this project was to translate artificial intelligence (AI) data models into a practical, low-burden primary care intervention. Specifically, the project aimed to co-design and refine an early-stage tool that enables healthcare teams to easily identify and address the social care needs of patients living with multiple long-term conditions (MLTCs). Background An increasing number of people live with multimorbidity—the presence of two or more chronic health conditions such as diabetes, heart disease, or depression. These illnesses frequently intersect with complex socio-economic challenges, including financial strain, unstable housing, and social isolation. While these social care needs profoundly impact clinical outcomes and overall quality of life, they are rarely identified during standard, time-constrained medical consultations. This project builds directly on the foundational, NIHR-funded AIM programme. The AIM study utilised machine learning to analyse anonymous health records, successfully identifying four distinct population clusters based on shared health and social care patterns. While stakeholders recognised the value of this AI data, they emphasised that algorithms must support, rather than replace, human clinical interactions. Furthermore, a comprehensive scoping review of nearly two thousand published studies revealed that existing digital tools focus almost exclusively on medical metrics like symptom tracking, often increasing the treatment burden on vulnerable patients. Consequently, a clear policy and clinical gap exists for an intervention that addresses broader life priorities. What We Did To bridge the gap between data science and clinical practice, the research team used a multi-stage, qualitative approach: Stakeholder Interviews: In-depth qualitative interviews were conducted with 24 individuals living with MLTCs and 20 health and social care providers to understand real-world boundaries, resource constraints, and systemic expectations. Prototype Co-Design: Synthesising these insights with literature reviews, the team generated a non-digital paper prototype. The prototype consisted of a one-page consultation prompt for clinicians and a self-navigation priority checklist for patients. Iterative Usability Testing: The prototype underwent "think-aloud" usability testing with 16 patients and 15 primary care professionals. Feedback was analysed dynamically throughout the fieldwork, allowing for rapid, recorded adjustments to the prototype's layout, flow, and terminology to minimise cognitive load. What We Found The research confirmed that patients view trusted primary care settings as appropriate hubs for initiating social care conversations, provided the approach is non-judgemental. However, clinicians expressed concern over severe time constraints and limited local referral pathways. The iterative testing crystallised six foundational user requirements for integrating health and social care tools: Patient Autonomy: Individuals must retain control over which social priorities to address first. Transparent Intent: The purpose and limitations of the tool must be stated explicitly up front. Low Treatment Burden: Visual layouts and wording must remain highly simplified and calming. Validation of Past Experiences: The tool must sensitively account for patients who have previously struggled to access care. Clear Operational Roles: Primary care staff must have clearly defined boundaries regarding their social care responsibilities. System Integration: Any final intervention must align seamlessly with existing NHS electronic health records and standard appointment durations. Patient and Public Involvement and Engagement (PPIE) Patient and public involvement was fundamental to ensuring the ethical application of AI-derived insights. PPI contributors collaborated closely with the research team to design all participant-facing study literature. Utilising a digital-first, visually accessible methodology, public advisors reviewed and refined the prototype's language. This direct involvement ensured that the wording was supportive, accessible to individuals with varying levels of literacy, and entirely free of institutional or stigmatising jargon. Conclusion By anchoring advanced algorithmic modelling in human-centred design, this project successfully created a viable framework to address social care needs within primary care. The research demonstrates that complex population data can be translated into clear, actionable, and low-burden clinical tools. This approach offers an equitable pathway toward integrated, holistic care that improves patient wellbeing without exacerbating NHS staff shortages or consultation pressures. Next Steps and Future Plans Backed by extended funding from the original NIHR grant, the project has entered a commercial partnership to translate the paper prototype into a live, NHS-compatible digital application named IndependAble. Future work will focus on securing NIHR Research for Social Care funding to launch a robust, multi-site evaluation. This upcoming trial will formally measure IndependAble's clinical utility, its impact on social care identification rates, and its operational sustainability within busy primary care services.
- Open Access
- 10 August 2026
- Published by Springer Nature in BMC Medicine
- Vol. 24 (1)
Remission of long-term conditions among adults with multiple long-term conditions in England primary care: A retrospective cohort study of 4.7 million people
Background: Multiple long-term conditions (MLTC) are increasingly prevalent among UK adults, yet the extent to which remission of long-term conditions is recorded in routine primary care remains poorly understood. Accurate documentation of remission has implications for clinical management, disease trajectories, and the use of electronic health records in research. Examining remission in the context of MLTC is important because remission in one condition may alter overall disease burden and treatment needs in people living with several coexisting conditions. We aimed to describe the frequency and variation of recorded remission across multiple long-term conditions among adults. Methods: We conducted a retrospective cohort study using linked primary care, hospital, and CPRD Aurum data. Adults aged ≥ 18 with MLTC (≥ 2 long-term conditions) between Jan 1, 1987, and Dec 31, 2020, were included. Remission was defined using condition-specific criteria based on remission or resolution codes, relevant clinical measurements meeting predefined thresholds, treatment patterns, and condition-specific clinical events recorded in electronic health records. We examined remission for 10 conditions with operationalizable definitions. Analyses were descriptive, summarising remission frequencies and sociodemographic patterns. Results: Among 4,745,099 adults followed for a median of 8 years, 900,609 (19%) had recorded remission of at least one condition. Remission varied substantially by condition, with the highest rates observed for anaemia (63.0%), asthma (44.71%), and diabetes (15.51%), calculated among people with an ever-recorded diagnosis of each condition. Those with remission were slightly younger and more often male; Asian and Black ethnic groups were under-represented, while deprivation patterns were broadly similar. Conclusions: Recorded remission in MLTC is highly heterogeneous and often rare. Variation between conditions and low rates of sustained remission likely reflect coding practices and clinical recording behaviours. Standardised remission definitions and consistent Systematised Nomenclature of Medicine (SNOMED) coding could improve data quality and support more reliable research on remission and disease trajectories.
- 10 August 2026
- Published by Springer Nature in BMC Medicine
- Vol. 24 (1)
Open AccessRemission of long-term conditions among adults with multiple long-term conditions in England primary care: A retrospective cohort study of 4.7 million people
Background: Multiple long-term conditions (MLTC) are increasingly prevalent among UK adults, yet the extent to which remission of long-term conditions is recorded in routine primary care remains poorly understood. Accurate documentation of remission has implications for clinical management, disease trajectories, and the use of electronic health records in research. Examining remission in the context of MLTC is important because remission in one condition may alter overall disease burden and treatment needs in people living with several coexisting conditions. We aimed to describe the frequency and variation of recorded remission across multiple long-term conditions among adults. Methods: We conducted a retrospective cohort study using linked primary care, hospital, and CPRD Aurum data. Adults aged ≥ 18 with MLTC (≥ 2 long-term conditions) between Jan 1, 1987, and Dec 31, 2020, were included. Remission was defined using condition-specific criteria based on remission or resolution codes, relevant clinical measurements meeting predefined thresholds, treatment patterns, and condition-specific clinical events recorded in electronic health records. We examined remission for 10 conditions with operationalizable definitions. Analyses were descriptive, summarising remission frequencies and sociodemographic patterns. Results: Among 4,745,099 adults followed for a median of 8 years, 900,609 (19%) had recorded remission of at least one condition. Remission varied substantially by condition, with the highest rates observed for anaemia (63.0%), asthma (44.71%), and diabetes (15.51%), calculated among people with an ever-recorded diagnosis of each condition. Those with remission were slightly younger and more often male; Asian and Black ethnic groups were under-represented, while deprivation patterns were broadly similar. Conclusions: Recorded remission in MLTC is highly heterogeneous and often rare. Variation between conditions and low rates of sustained remission likely reflect coding practices and clinical recording behaviours. Standardised remission definitions and consistent Systematised Nomenclature of Medicine (SNOMED) coding could improve data quality and support more reliable research on remission and disease trajectories.
- Open Access
- 12 June 2026
- Published by SAGE Publications in Journal of Multimorbidity and Comorbidity
- Vol. 16
Development of reporting guidelines for multiple long-term condition research
Background: Incomplete and inconsistent reporting amongst research studies in people with Multiple Long-Term Conditions (MLTC) hinders the comprehensive evaluation, synthesis, and interpretation of study findings for application by clinicians, researchers, patients and policymakers. This limitation leads to heterogeneous findings, duplication of work and restricts the practical application of research outcomes in clinical settings, public health strategies, and policymaking. Given the high prevalence and complexity of MLTC, there is a pressing need for standardised guidelines to promote clarity, consistency, and comprehensiveness in study reporting. Such guidelines can enhance transparency and reproducibility, thereby increasing the impact of research on healthcare decisions and policy development. Methods: We followed a four-stage process of guidelines development: a review of MLTC reporting practices; a workshop with diverse stakeholders to identify and refine items for inclusion; a prioritisation consensus exercise to agree on key items; and pilot-testing to refine interpretation and usability of the guidelines. Results: This work has produced the first set of reporting guidelines addressing the need for standardised reporting in MLTC research. Application of these guidelines has the potential to improve research clarity and reproducibility, enabling better comparisons across studies and shared learning. Improved reporting standards will also facilitate the translation of research findings into effective healthcare strategies and policies, contributing to better health outcomes for MLTC patients. Conclusion: These initial guidelines offer a structured approach to improving the reporting quality of MLTC research. Future evaluations will assess its impact on research transparency and real-world application.
- 12 June 2026
- Published by SAGE Publications in Journal of Multimorbidity and Comorbidity
- Vol. 16
Open AccessDevelopment of reporting guidelines for multiple long-term condition research
Background: Incomplete and inconsistent reporting amongst research studies in people with Multiple Long-Term Conditions (MLTC) hinders the comprehensive evaluation, synthesis, and interpretation of study findings for application by clinicians, researchers, patients and policymakers. This limitation leads to heterogeneous findings, duplication of work and restricts the practical application of research outcomes in clinical settings, public health strategies, and policymaking. Given the high prevalence and complexity of MLTC, there is a pressing need for standardised guidelines to promote clarity, consistency, and comprehensiveness in study reporting. Such guidelines can enhance transparency and reproducibility, thereby increasing the impact of research on healthcare decisions and policy development. Methods: We followed a four-stage process of guidelines development: a review of MLTC reporting practices; a workshop with diverse stakeholders to identify and refine items for inclusion; a prioritisation consensus exercise to agree on key items; and pilot-testing to refine interpretation and usability of the guidelines. Results: This work has produced the first set of reporting guidelines addressing the need for standardised reporting in MLTC research. Application of these guidelines has the potential to improve research clarity and reproducibility, enabling better comparisons across studies and shared learning. Improved reporting standards will also facilitate the translation of research findings into effective healthcare strategies and policies, contributing to better health outcomes for MLTC patients. Conclusion: These initial guidelines offer a structured approach to improving the reporting quality of MLTC research. Future evaluations will assess its impact on research transparency and real-world application.