A data-driven exploration of hypotheses on disease dynamics

Abstract

Unsupervised learning is often used to obtain insight into the underlying structure of medical data. In this paper, we show that unsupervised methods, in particular hidden Markov models, can go beyond this by guiding the generation of clinical outcome measures and hypotheses, which play a crucial role in medical research. The usage of the data-driven approach facilitates selecting which hypotheses to further investigate. We demonstrate this by using clinical trial data for psychotic depression treatment as a case study. The discovered latent structure and proposed outcome are shown to provide new insight into the heterogeneity of psychotic depression in terms of predictive symptoms.