A practical overview and decision tool for analyzing recurrent events in mental illness: A review

Abstract

Mental illnesses are chronic conditions in which an individual will often experience recurrent outcomes such as hospitalization, symptomatic relapse or self-harm behaviours. Most clinical research in psychiatry considers only the first event, and does not analyze subsequent recurrent events. Methods exist to analyze recurrent events; however, these methods are underused in the psychiatric research literature. This review identifies that recurrent events can be analyzed using a time homogenous or time-to-recurrent-event (TTRE) framework. The TTRE framework is underutilized in psychiatric research; however, this framework allows for longitudinal observations that are more congruent with the chronic nature of psychiatric illness than typical first event analyses. There are several readily available statistical models using the TTRE framework extending the standard Cox proportional hazards model. Our decision tool outlines four aspects of a research question to consider when selecting a TTRE model: (1) importance of event timing, (2) explanatory vs predictive, (3) common vs event-specific hazard, and (4) correlation of events within an individual. Analyzing recurrent events in psychiatric research provides an opportunity to address research questions aimed at understanding the longitudinal course of a chronic condition. These approaches may provide novel insights into risk factors or interventions for psychiatric illness, and ultimately improved outcomes for these chronic conditions.