1061 Toward Behavioral Health Risk Monitoring via Wearable Physiological Sensing During Sleep

Abstract Introduction In shipboard military operations, where chronic stress, irregular schedules, and sleep disruption are pervasive, personnel can experience a greater risk of behavioral health disturbances. Wearable sensors may facilitate monitoring of sleep and associated behavioral health symptoms with minimal disruption to operational demands. As a foundational step towards such a solution, we investigated the degree to which wearable-derived sleep outcomes can serve as indicators of changes in depression and anxiety symptoms. Methods Shipboard U.S. Navy personnel (113 men, 83 women) wore Oura Ring (Generation 2 and 3) for 8-26 weeks and completed periodic behavioral health surveys (separated by 20 days) that assessed depression using either the PROMIS or PHQ-8, or anxiety using GAD-7. Physiological markers were computed per sleep period according to Oura algorithms: average heart rate variability (HRV), and calculated sleep summary outcomes [total sleep time (TST), total wake time (TWT), sleep efficiency (SE), sleep latency, and sleep staging durations for light, deep, and REM)]. Linear mixed-effects regression examined associations between these repeated measures across up to two weeks of sleep data prior to reported behavioral health symptoms. The mediating impact of age, BMI, sex, and psychiatric medication intake was also explored. Separate models were fit for each outcome: depression score (DEP), depression severity change (ΔDEP), anxiety score (ANX), and anxiety severity change (ΔANX). Exhaustive best-subset selection minimizing AIC provided inference on indicator associations. Pearson correlation between estimated and actual outcome values characterized fit quality. Results Best-fit models showed robust fit quality with significance: DEP: 0.87 (N=77), ΔDEP: 0.62 (N=70), ANX: 0.87 (N=27), ΔANX: 0.64 (N=27). While sleep latency was associated with both depression outcomes (DEP, p=0.019; ΔDEP, p=0.033), TWT showed the strongest relationship with DEP (p=0.01), whereas SE was the primary correlate for ΔDEP (p = 0.004). For anxiety outcomes, REM duration was the top predictor (ANX, p< 0.001; ΔANX, p=0.027). Two mediating factors showed limited effects as adjustment covariates: Age (DEP) and BMI (ΔDEP). Conclusion Robust associations of wearable-derived sleep metrics with behavioral health outcomes substantiate the potential of using them as indicators of worsening or improving depression and/or anxiety symptoms. Support (if any) Military Operational Medicine Research Program (work unit no. N2010) DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited. This material is based upon work supported by the Department of the Army under Air Force Contract No. FA8702-15-D-0001 or FA8702-25-D-B002. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Department of the Army.