Abstract
IMPORTANCE: Sleep plays a critical role in postoperative recovery, influencing immune function, pain perception, neurocognitive performance, and wound healing. Although surgical interventions are known to disrupt sleep, the extent and trajectory of these disruptions across varying procedural risks remain poorly characterized. OBJECTIVE: To characterize postoperative alterations in sleep stages and determine how these trajectories vary across different surgical procedures. DESIGN, SETTING, AND PARTICIPANTS: This retrospective cohort study used wearable device-derived sleep data linked to electronic health records from US participants in the prospectively maintained All of Us Research Program database. Adults undergoing surgery from January 2012 to December 2024 with 90 days or more of preoperative and 30 days or more of postoperative wearable sleep data were included. Data analyses were performed from July to November 2025. EXPOSURE: Surgical intervention stratified by procedural risk levels. MAIN OUTCOMES AND MEASURES: The primary outcome was sleep metrics, including total sleep and rapid eye movement (REM), deep, light, and wake-stage durations, compared across 9 postoperative epochs and stratified by surgical risk (low, intermediate, or high). Linear mixed-effects models were fitted for each metric with opioid exposure, dose, and patient age as covariates. Association between each sleep metric and postoperative complications was assessed using multivariate logistic regression analysis. RESULTS: A total of 634 unique surgical procedures in 512 patients were included in the analytic cohort; median (IQR) patient age at the time of surgery was 59 (46-67) years, and 558 patients (88.0%) were female. High-risk procedures were associated with significant and sustained mean (SD) reductions in REM and deep sleep through postoperative day 7 (deep: Δ, -18.7 [48.6] minutes; P < .001; REM: Δ, -12.4 [30.1] minutes; P < .001), while total sleep was reduced acutely (Δ, -19.4 [145.3] minutes; P = .004). Light sleep decreased transiently; wake-stage duration increased significantly (mean [SD] Δ, +13.9 [31.8] minutes; P < .001). Intermediate-risk procedures were associated with milder disruptions, primarily limited to days 0 through 3. In contrast, low-risk procedures were not associated with significant changes in any sleep metric at any postoperative epoch. Opioid exposure was associated with increased light (β, approximately +5.1 minutes; P < .001) and wake-stage (+10.2 minutes; P < .001) durations but suppressed deep sleep (-5.1 minutes; P < .001). Age was similarly linked to greater wake-stage sleep (+0.30 minutes/y) and reduced restorative sleep (-0.05 minutes/y). Finally, each 10-minute reduction in change in total sleep duration was associated with higher odds of Clavien-Dindo grades I and II complications (adjusted odd ratio, 1.13 per 10 minutes; 95% CI, 1.04-1.24; P = .006). CONCLUSIONS AND RELEVANCE: Per the results of this cohort study, sleep architecture is differentially disrupted following surgery, with the greatest and most prolonged impairments occurring after high-risk procedures, with REM and deep sleep being most affected.