Mobile Health Assessment of the Recovery Process in the First Month Following Lumbar and Thoracolumbar Spine Surgery

Introduction: There is sparse research investigating the time-course of early recovery following spine surgery. Instead, most research focusses on cross-sectional assessments, often many months from surgery. longitudinal mobile health (mHealth) assessments integrating symptom dynamics from ecological momentary assessment (EMA) and wearable biometric (Fitbit) data may capture important influences on recovery. Our study aims to evaluate early postoperative recovery using longitudinal mHealth subjective (EMA) and objective (Fitbit) assessment and to assess the utility of preoperative mHealth measures in predicting the postoperative recovery pattern.

Methods: We recruited patients aged 21-85 undergoing lumbar and thoracolumbar surgery for degenerative spine disease from 2021 to 2023. For up to three weeks before surgery and one month after, participants were passively monitored with Fitbit and completed daily EMAs assessing pain, disability, depression, and catastrophizing. First-month postoperative recovery in EMA and Fitbit were clustered using functional principal component analysis. After that, an XGBOOST model predicted patient classification in each domain using preoperative EMA, Fitbit, or both. Models were internally validated.

Results: A total of 129 patients were included; median age was 62 years (IQR:16), with 56% female. All domains were clustered into “better” and “worse” recovery groups. In predicting EMA recovery classification, adding preoperative Fitbit measures significantly improved all the EMA-alone models' performance except for pain intensity. In predicting Fitbit recovery, adding preoperative EMA data significantly improved Fitbit-alone models’ accuracy for activity bout number and maximum 30-minute cadence, but did not improve prediction of step count recovery. Additionally, each assessment type alone predicted the recovery classification of the other type with moderate to very good accuracy.

Conclusion : Our study highlights postoperative recovery dynamics following lumbar spine surgery. We found distinct recovery patterns in all EMA and Fitbit measures. Integrating objective preoperative features enhanced the prediction of postoperative recovery in subjective EMA measures, and vice versa.