Longitudinal Aspects of Multimorbidity in Chronic Heart Disease: A Data Science Perspective

Multimorbidity, the simultaneous presence of two or more chronic diseases in the same individual, is an increasing health challenge and will continue to be so in the foreseeable future, particularly among patients with chronic heart disease. Population-wide mappings of chronic diseases based on large databases of patient data offer insights into multimorbidity patterns, informing decision-making processes on prevention and treatment initiatives. However, multimorbidity patterns obtained from mappings are mainly based on cross-sectional data, which does not cover longitudinal aspects.

This thesis studies the trajectories of patients’ portfolios of chronic diseases from a data science perspective. We present three modelling approaches, focusing on different longitudinal aspects of multimorbidity. We apply the models to nationwide registry data from the population of adult Danish chronic heart disease patients. Firstly, a Markov model is proposed for studying the trajectories of disease portfolios, allowing us to map how chronic diseases, age, sex and socioeconomic position affect these. Secondly, we study the impact of disease portfolios on mortality based on a dynamic survival model, providing insights into the consequences of multimorbidity. Thirdly, we present a two-step deep learning based clustering framework for analyzing disease trajectories, considering both age and disease modalities. The framework provides a mapping of how disease portfolios cluster over time, which is informative for possible early-stage targeted preventive interventions.

The model based frameworks in this thesis map longitudinal aspects of multimorbidity in a chronic heart disease population. However, these frameworks can be adapted to study other patient populations, making them broadly applicable.