AI-enabled health profiling, disability estimation and forecasting using linked health data

Abstract

The growing availability of linked healthcare big data creates major opportunities to improve the profiling of chronic diseases and their consequences at population level. This session examines how artificial intelligence, machine learning, and advanced epidemiological methods can be used jointly to turn routinely collected data into meaningful health profiles for public health action. The presentation will focus on the use of administrative databases, electronic health records, and disease registries to identify chronic conditions more accurately, reduce under-ascertainment, and address misclassification and missing data. It will then show how these same data sources, analysed through neural networks and related AI approaches, can be used to estimate disability burden, model disease trajectories, and capture broader dimensions of population health, including self-perceived health. By integrating disease identification with burden estimation and trajectory forecasting, the session will highlight the value of AI-enabled analytics for more comprehensive health profiling in complex, heterogeneous populations. Particular attention will be given to methodological robustness, including data linkage, phenotyping algorithms, validation strategies, and the treatment of uncertainty in longitudinal prediction. Neurological conditions, including Alzheimer’s disease, other dementias, and multiple sclerosis, will be used as the main case studies. These examples will illustrate how AI-supported analysis of routine health data can improve surveillance, reveal unmet needs, and generate policy-relevant evidence for service planning and prioritisation.