Sohan Seth, sohan.seth@ed.ac.uk
Bruce Guthrie, bruce.guthrie@ed.ac.uk
Project Description
Populations around the world are ageing rapidly. It is predicted that by 2050 the proportion of people who are aged 60+ years will nearly double from 12% (2015 estimate) to 22%. Ageing populations drive increasing frailty, multimorbidity and polypharmacy, all responsible for higher mortality, diminished quality of life and increased health and social care needs. Better prediction of individual prognosis (e.g., of mortality, hospital admission and institutionalization), allows targeted early intervention potentially leading to better quality of later life and economic benefit. The availability of electronic health care data facilitates this by providing unprecedented detail of the longitudinal health trajectory of an individual, but more sophisticated methods are needed to deal with the incompleteness of the data, to extract causal reasoning beyond simple association, and to adapt to the dynamic nature of individual and population health. In this project you will explore the explainability and operationalizability of existing prognostic models for later life health and care including multi-state models, random survival forest, deep learning models, causal inference using mixtures etc., and develop models that are explainable and operationalization alongside being accurate, stable, adaptable, transferable and capable of quantifying uncertainty. The project will be using longitudinal survey data from the English Longitudinal Study of Ageing (www.elsa-project.ac.uk) and longitudinal routine healthcare data from the SAIL Databank (saildatabank.com).
