Abstract

About 25% of the UK population have two or more long term conditions. This multimorbidity is associated with a reduction in quality of life, increased use of health services and reduced life expectancy. To date, multimorbidity has been seen as a random assortment of diseases, making it difficult to address. However, with new understanding of the impact of various factors (including biological, social, behavioural, environmental and others), multimorbidity can be seen as a series of non-random clusters of disease. Improving the characterisation of these clusters with artificial intelligence and machine learning could have significant benefits to health and social care. In preparation for this development award we have assembled a team of computer and data scientists, statisticians, epidemiologists and clinicians with skills and interests in data analysis and interpretation, to address how to apply artificial intelligence (AI) methodology to develop new insights into the detection and prevention of multimorbid conditions. Our focus will be on detecting long-term trajectories as they evolve over the life course by applying a combination of traditional methods and novel AI approaches under development by our group including probability (fuzzy) clustering and multistate modelling embedded in a deep learning framework. As a paradigm, we will focus our analysis on the role of chronic inflammation in explaining the occurrence of disease across the life course. We will explore the informativeness of three test datasets where relevant long-term data on inflammatory exposures is recorded and test the feasibility of these methods in a set of proof of concept analyses. Chronic inflammation is an established driver of disease risk that is amenable to intervention and understanding its role in the evolution of multimorbidity provides a direct pathway to patient benefit. By establishing a group of experts in clinical medicine, biological sciences, epidemiology, computing sciences and statistics to bring a focus on inflammatory drivers of comorbidity, we expect this development work to demonstrate the feasibility of novel AI technologies to identify inflammation driven clustering of patient disease trajectories over a long time horizon. We have significant research infrastructure partners (MRC Dementias Platform UK, NIHR Applied Research Collaboration Eastern), commercial advisors (British Telecom Health, Aviva Life Analytics, Evergreen Life and Vitality) and will have support and input from the Institute and Faculty of Actuaries. These collaborations will enable us to scale up the analysis to include further local and national research and administrative databases in a future application. By embedding patients in this project, and establishing an AI patient and public engagement group to assist with the understanding of the use of AI in data interpretation, we aim to ensure that the overarching aim of this work is to lead to effective strategies for disease prevention both for the benefit of individuals and for the population.