People who grow up in disadvantaged circumstances tend to develop multiple long-term health conditions—such as diabetes, depression, and heart disease—decades earlier than their better-off peers, and this project will use artificial intelligence to pinpoint exactly when and how to intervene. The problem is that researchers lack the large-scale, lifelong data needed to understand why some people accumulate several chronic illnesses by age 65 while others do not. Current approaches often treat each condition separately, missing the fact that the order in which diseases appear—and the social factors like education and employment that precede them—dramatically alters outcomes. This project fills that gap by linking birth cohort studies spanning decades with millions of electronic health records. If successful, the research will produce a practical toolkit: a set of "burdensomeness" indicators that doctors and public health officials can use to identify high-risk groups in childhood or early adulthood. It will also model what would happen if specific interventions—such as improved housing, mental health support, or workplace adjustments—were applied at critical life stages. The ultimate change would be a shift from treating each disease as it arrives to preventing the entire cascade of illness before it starts, particularly for people from disadvantaged backgrounds.
View original technical description
Background Multimorbidity occurs earlier in the lifecourse among people from more disadvantaged backgrounds. Wider determinants such as education and work influence development of long-term physical and mental health conditions (LTCs) across the lifecourse. Sequence of accrual of conditions varies considerably and influences outcomes. Sentinel conditions (the first LTC to occur in the lifecourse) may play an important role. Understanding how wider determinants, sentinel conditions and LTC accrual sequence affect risk of early-onset burdensome/complex multiple LTC multimorbidity (MLTC-M) is vital to inform when and how to deploy prevention interventions. Data at sufficient scale linking whole lifecourse events, including wider determinants, with clinical information is lacking.Through our Development Award ( MELD ) we established the necessary environment, principles, systems, methods and team to use artificial intelligence (AI) techniques to study lifecourse causes of early-onset MLTC-M. The aim for our Research Collaboration is to safely deliver an Artificial Intelligence (AI)-enhanced epidemiological analytic system in which optimal lifecourse time points and targets for prevention of early-onset, burdensome MLTC-M are identified through multidisciplinary synthesis and analysis of birth cohorts and electronic health records, and disseminated to key stakeholders. Methods - five Work-Packages (WP): Through qualitative evidence synthesis and a consensus study we will develop deeper understanding of what burdensomeness and complexity mean to people living with early-onset (by age 65) MLTC-M, carers and healthcare professionals, and produce a suite of burdensomeness/complexity indicators for use as clustering domains in routine healthcare data.(WP1–lead Fraser) We will provide the safe data environment and readiness for AI analyses across large, representative routine healthcare datasets (Secure Anonymised Information Linkage (SAIL) and Clinical Practice Research Datalink (CPRD)) and birth cohorts (National Child Development Study (NCDS), Aberdeen Children of the 1950s (ACONF), 1970 British Cohort Study (BCS70)), then harmonise specified LTCs across birth cohorts and routine data.(WP2–lead Boniface) Using the WP1 burdensomeness/complexity indicators, we will apply AI methods to identify novel early-onset, burdensome MLTC-M clusters and sentinel conditions in routine data, develop and apply semi-supervised learning to match individuals in birth cohorts into routine data MLTC-M clusters, identify determinants of burdensome clusters using matched datasets, and model trajectories of LTC and burden accrual.(WP3–lead Hoyle) We will characterise clusters of early-life (pre-birth to 18 years) risk factors for early-onset, burdensome MLTC-M and sentinel conditions, define population groups in early life at risk of future MLTC-M, identify critical time points and targets for prevention, and model counterfactual prevention scenarios of interventions acting on combined risk factors at key timepoints.(WP4–lead Alwan). We will investigate the influence of sentinel conditions and sequence of determinant and condition accrual on development of early-onset, burdensome MLTC-M clusters and compare AI and causal inference modelling for potential preventable moments across the lifecourse (WP3/4). Patient and public involvement is embedded throughout. Timeline: 30 months from April 2022 Impact and Dissemination We will identify and engage key stakeholders to explore timepoints and targets to prevent/delay specified sentinel conditions and early-onset, burdensome MLTC-M. Partnering with our PPI Advisory Board, and maintaining stakeholder engagement, we will co-produce public health implementation recommendations based on our findings.(WP5(cross-cutting)–lead Wilkinson).
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know