Using artificial intelligence (AI) to characterize the dynamic inter-relationships between MUltiple Long-term condiTIons and PoLYpharmacy and across diverse UK populations and inform health care pathways (AI-MULTIPLY)

Our overarching question is: "How do multiple long-term conditions (multimorbidity)(MLTC-M) and polypharmacy interact and how is this interaction modified by inequalities?" Background MLTC-M is associated with premature mortality, significant treatment burden, increased healthcare and socio-economic costs. Deprivation contributes to its early onset while inequalities exacerbate it. Polypharmacy is associated with MLTC-M but the inter-relationship is complex and poorly understood. Whereas appropriate polypharmacy may alleviate/prevent disease burden, in other poorly defined circumstances polypharmacy may cause harm. Aims and objectives Overarching aim: to characterise MLTC-M and polypharmacy (MLTC-M-PP) trajectories and define the inter-relationships between MLTC-M clusters, polypharmacy, inequalities and healthcare outcomes over the life-course. Specific aims: Identify early intervention points, including antecedent events prior to tipping points associated with rapidly progressive "bursts" of multimorbidity/polypharmacy. Determine influence of personal and social intersectional factors with the aim of reducing health inequalities. Define "appropriate" and "inappropriate" polypharmacy in a MLTC-M context. Initiate emulation of target trials in our target populations. Demonstrate proof of concept clinical decision support tools based on AI and related technologies to flag MLTC-M-PP risk to health practitioners. Map the emergence of ideas, assumptions, consensus and knowledge creation in the development of a health-related explainable-AI algorithm, with full consideration of Inequalities. Foster a reflexive, interdisciplinary learning and training environment. Data, Methods and Team Our interdisciplinary consortium brings together experts in data access/engineering, epidemiology, pharmacy, clinical medicine, AI, anthropology, trial emulation, PPI, and healthcare relevant to the management of MLTC-M. Our PPI and stakeholder groups have shaped and extended our research questions and methodologies. Our cutting-edge methodologies include: a novel transfer of Bursty Dynamics theory from complex systems research to characterise healthcare event structure; Association Rules Mining to illuminate event sequence; topological data analysis as a novel approach to understanding complex MLTC-M-PP networks; market-basket analysis with topic modelling to define latent associated conditions; ethnographic analysis of interdisciplinary AI-in-health working practices. Quantitative methods will be optimised using well-curated data (UK-Biobank and CPRD), and then validated on routinely collected electronic health record (EHR) NHS data from the North-East of England and East London, with particular focus on intersectional contrasts, including ethnicity and deprivation. Findings will be interpreted in the local context, with PPI support, and validated in carefully selected cohorts in Bradford and Scotland. Timelines for delivery 0-18 months: data access and engineering; 0-24 months application of AI methods to deliver MLTC-M-PP clusters; 24-30 months development of clinical decision support systems; 0-30 months: ethnographic analysis and PPI. Impacts Increased understanding of MLTC-M-PP clusters in the context of inequality factors and wider intersectional drivers. Identification of potential 'tipping points and potential new interventions: For testing in pragmatic trials. To support the development of appropriately designed clinical decision support systems. Enhanced training of early career researchers and capacity building. Development of guidance for future interdisciplinary AI-in-health collaborations. Dissemination Informed by consultation with our PPI groups, important findings will be shared widely. We will facilitate patient, research-partner led discussions across NIHR networks around AI in healthcare including the development of jargon-busting information on AI.