Depression Management Using Electronic Health Record: Individual Progression Prediction

Abstract

Mitigating depression has become a national health priority and is the most common mental illness seen in primary care. Due to the complex dynamics of individual's depression trajectory, how to predict the progression of an individual patient's depression has long been an open problem. In this thesis, by using the electronic Patient Health Questionnaire (PHQ)-9 data, a new nature-history model is proposed to provide individual depression prediction, based on which the PHQ-9 score of a new patient at the next time interval can be predicted by using a multivariate nearness approach. The accuracy of the model is further validated under distinct scenarios by using five-fold validation. A simulation-based monitoring system is further established, with which a visit schedule table can be designed for each patient according to the predicted depression level and a given criteria. The analysis offers important insights into depression prediction and management.