Depressive and manic episodes within bipolar disorder (BD) and major depressive disorder (MDD) involve altered mood, sleep, and activity alongside physiological alterations that wearables can capture.
We explored whether physiological wearable data could predict: (aim 1) the severity of an acute affective episode at the intra-individual level, (aim 2) the polarity of an acute affective episode and euthymia among different individuals. Secondarily, we explored which physiological data were related to the prior predictions, generalization across patients, and associations between affective symptoms and physiological data.
We conducted a prospective exploratory observational study including patients with BD and MDD on acute affective episodes (manic, depressed, and mixed) whose physiological data were recorded with a research-grade wearable (Empatica E4) across three consecutive timepoints (acute, response, and remission of episode). Euthymic patients and healthy controls (HC) were recorded during a single session (∼48 hours). Manic and depressive symptoms were assessed with standardized psychometric scales. Physiological wearable data included the following channels: acceleration (ACC), temperature (TEMP), blood volume pulse (BVP), heart rate (HR), and electrodermal activity (EDA). For data pre-processing, invalid physiological data were removed using a rule-based filter, channels were time-aligned at 1 second time units and then segmented window lengths of 32 seconds, since those parameters showed the best performances. We developed deep learning predictive models, assessed channels' individual contribution using permutation feature importance analysis, and computed physiological data to psychometric scales' items normalized mutual information (NMI). We present a novel fully automated method for analysis of physiological data from a research-grade wearable device, including a rule-based filter for invalid data and a viable supervised learning pipeline for time-series analyses.
35 sessions (1,512 hours) from 12 patients (manic, depressed, mixed, and euthymic) and 7 HC (age 39.7±12.6; 31.6% female) were analyzed. (aim 1) The severity of mood episodes was predicted with moderate (62%-85%) accuracies. (aim 2) The polarity of episodes was predicted with moderate (70%) accuracy. The most relevant features for the former tasks were ACC, EDA, and HR. Kendall W showed fair agreement (0.383) in feature importance across classification tasks. Generalization of the former models were of overall low accuracy, with better results for the intra-individual models. "Increased motor activity" was associated with ACC (NMI>0.55), "aggressive behavior" with EDA (NMI=1.0), "insomnia" with ACC (NMI∼0.6), "motor inhibition" with ACC (NMI∼0.75), and "psychic anxiety" with EDA (NMI=0.52).
Physiological data from wearables show potential to identify mood episodes and specific symptoms of mania and depression quantitatively, both in BD and MDD. Motor activity and stress-related physiological data (EDA and HR) stand out as potential digital biomarkers for predicting mania and depression respectively. These findings represent a promising pathway towards personalized psychiatry, in which physiological wearable data could allow early identification and intervention of mood episodes.