Perioperative Wearable Device Features are associated with Moderate-to-Severe Chronic Pain after Surgery in the "All of Us" Research Program.

Abstract

Chronic pain is a common and disabling condition affective over 50 million adults in the United States. Surgery may offer a critical opportunity to prevent chronic pain, as 10-35% of surgical patients develop new or worsening pain. However, prevention and treatment of chronic pain among surgical patients is hindered by a lack of reliable and clinically actionable biomarkers. Digital health technologies offer novel opportunities to develop and validate digital biomarkers of chronic pain among surgical patients. In this study, we leveraged data from the "All of Us" Research Program, which linked Electronic Health Record (EHR) data with participant's own Fitbit data. We identified participants who: 1) had a surgical procedure as indicated in the EHR; 2) had Fitbit data available 30 days before and/or after surgery; and 3) completed the Overall Health Questionnaire assessing pain intensity 3 months to 5 years after surgery. Our final cohort included 302 surgical patients, 29% of whom reported moderate-to-severe pain approximately 1.5 years after surgery. Among the domain-specific models, sleep features provided the best predictive performance, achieving an AUC of .722 in a held-out test set. The lowest AIC was observed in the stepwise model based on the subset of participants whose Fitbits provided sleep stage data (n = 244, AIC = 178.75, AUC = .649, sensitivity = 0.37, specificity = 0.87). Younger age (OR = 0.97; p = 0.048), lower preoperative step variability (OR = 0.56; p = 0.009), and higher preoperative variability in REM sleep proportion (OR = 1.62; p = 0.023) were associated with greater risk of moderate-to-severe pain approximately 1.5 years after surgery. Digital biomarkers derived from consumer wearable device data appear highly promising for improving prediction and understand of chronic postoperative pain.