Classification and Prediction of Post-Trauma Outcomes Related to PTSD Using Circadian Rhythm Changes Measured via Wrist-Worn Research Watch in a Large Longitudinal Cohort

Abstract

Post-Traumatic Stress Disorder (PTSD) is a psychiatric condition resulting from threatening or horrifying events. We hypothesized that circadian rhythm changes, measured by a wrist-worn research watch are predictive of post-trauma outcomes.

Approach: 1618 post-trauma patients were enrolled after admission to emergency departments (ED). Three standardized questionnaires were administered at week eight to measure post-trauma outcomes related to PTSD, sleep disturbance, and pain interference with daily life. Pulse activity and movement data were captured from a research watch for eight weeks. Standard and novel movement and cardiovascular metrics that reflect circadian rhythms were derived using this data. These features were used to train different classifiers to predict the three outcomes derived from week-eight surveys. Clinical surveys administered at ED were also used as features in the baseline models.

Results: The highest cross-validated performance of research watch-based features was achieved for classifying participants with pain interference by a logistic regression model, with an area under the receiver operating characteristic curve (AUC) of 0.70. The ED survey-based model achieved an AUC of 0.77, and the fusion of research watch and ED survey metrics improved the AUC to 0.79.

Significance: This work represents the first attempt to predict and classify post-trauma symptoms from passive wearable data using machine learning approaches that leverage the circadian desynchrony in a potential PTSD population.

Fig. 1.

AURORA Freeze 2 Dataset overview and number of participants in each outcome group that is used in this research. Outcome surveys applied at week eight (PCL-5, PSQIA-PanicSleep, and PROM-Pain4a) were used to create the outcome groups. ED surveys included PDI, MCEPS and PCL-5 administrated at ED department following trauma. Top row of the tables indicates the number of participants that answered the outcome surveys, which is the maximum number available for the analysis. The rows below the first row indicate if the participants shared other modalities in addition to the outcome surveys.

Fig. 2.

Percentage of hours with actigraphy and derived heart rate (HR) data in the eight-week study period. If no samples are captured in a given clock hour, that hour is marked as empty.

Fig. 3:

Timeline of data collection and clinical surveys. In the bottom plots, actigraphy and RR Interval data collected with the research watch is illustrated.

Fig. 4.

Detection of rest and activity regions from actigraphy data. Lighter colors indicate higher intensity movements. Deviations from the typical pattern are seen on days 40–56 in this example participant.

Fig. 5.

Feature importance for logistic regression models (window size=56 days). Highest five average absolute feature coefficients across folds are illustrated for each outcome.

Fig. 6.

AUC of the logistic regression models with different window size selection. Subplot (a) shows the AUC for the PTSD outcome, subplot (b) shows PTSD-Panic Sleep/Anx. outcome, and subplot (c) shows PTSD-Pain. Int. outcome over the days.

Fig. 7.

Trajectories of pnn50, HF, and RMSSD features for participants who develop PTSD and who do not, as determined by PCL-5 survey at week eight. Mean of features are shown with solid lines and 95% confidence intervals are shown with the shaded regions.