Revealing state changes in emotion during treatment for depression using Hidden Markov Models

For depressed individuals, the process of change in emotions during therapy has been shown to occur in heterogeneous, non-linear patterns when studied with intensive longitudinal assessments. Both sudden shifts and more gradual changes are common. This fits the conceptualization of psychopathology as a dynamical system, which leads us to further expect that the changes in emotions and symptoms that occur during therapy may constitute a transition from one state to another – from depression to remission, for instance – or perhaps even a cascading process of intermediate dynamically stable states. By revealing how state changes occur – gradually, abruptly, with intermediate stable states – in individuals’ emotions during treatment, this study may improve our understanding of the clinical observation that emotional instability is part of the process of symptom change and improvement during therapy. Identifying transition periods and discontinuities in emotion change can guide researchers and eventually clinicians to those parts of therapy that are critical for overall change. We aim to examine whether the process of change in symptoms and daily life affect during therapy can be understood as (a series of) transitions between alternative dynamically stable states. We aim to reveal how the process of change, and especially improvement, looks for different individuals. To that end, we will use a Bayesian multilevel Hidden Markov Modeling (mHMM) approach to uncover the underlying states in emotion time series of depressed individuals who are entering psychological treatment.