Interpreting Comorbidity Groups via Risk Trajectories in the Health Record


In this paper, we aim to explore patient trajectories in time that evolve according to their risk of developing comorbidities. For our analysis, we use a probabilistic latent variable model, which aligns patients through time according to their risk of developing conditions within discovered comorbidity groups. We report our findings on a large population of patients and their in-hospital admissions through time (300,000 patients, over a span of 24 years overall).

Neural Information Processing Systems Machine Learning for Health Workshop
Bharat Srikishan
PhD Student