Background/Objective: The purpose of this study was to compare patient-reported and passive-sensor-based measures of sleep in patients with a psychiatric diagnosis.

Design: This was a retrospective study of patients being treated for psychiatric disease who received the Tempus nP pharmacogenomics assay and used the TempusPRO application (app). We extracted patient-reported sleep scores from standard psychiatric assessments and a one-question daily score (“check-in”). Sleep from the two weeks prior to an assessment was summarized with average sleep duration and variance. The association between sleep and cardiac sensor data was tested using Pearson correlation.

Results: We analyzed a subset of 1,191 patients that used TempusPRO between January 2021 and June 2022. Of this subset, 1,096 patients completed daily check-in surveys, and 544 completed assessments reviewing the prior two weeks’ sleep. Self-reported sleep difficulty on assessments was negatively associated with patients’ average duration of sleep and positively associated with the variation in sleep duration over the prior two weeks (n=177). Daily sleep quality scores were positively correlated with sleep duration (Spearman estimate 0.08, p=0.002). More sleep was associated with higher heart rate variability and lower resting heart rate (r2: 0.1–0.15).

Conclusion: We found expected associations between passive-sensor data and self-reported sleep data, as well as between sleep and cardiac sensor data. Self-reported data measurements are more sparse and often more difficult to capture, compared to passive-sensor data. Mobile sensor data can complement patient-reported sleep quality and potentially be used in the absence of self-reported data.

Funding/financial disclosures: All the authors are employees of Tempus Labs, Inc.

VIEW THE PUBLICATION

VIEW THE POSTER