(I-638) Multilinear Regression Modeling of Smartphone-Based Accelerometry Data Predicts Post-Operative Mobility Status in Lumbar Spine Surgery Patients
Medical Student Perelman School of Medicine at the University of Pennsylvania, United States
Introduction: Improving functional ability is one of the main treatment goals of both patients and surgeons in lumbar spine surgery; however, traditional subjective outcome assessments in spine surgery cannot directly quantify patient mobility. Recently, patient activity data extracted from patient smartphones has been investigated as a high-fidelity and objective method of assessing patient functional status before and after surgery. In this study, we investigated whether smartphone activity data can be used to predict post-operative mobility status in lumbar spine surgery patients.
Methods: Patients who received either lumbar decompression or lumbar fusion were retroactively consented and enrolled. Activity data (steps-per-day) recorded in the Apple Health (Apple Inc., Cupertino, CA) across 2 years peri-operatively was classified into 6 temporal epochs representing distinct functional states, including pre-operative baseline, acute pre-operative decline, post-operative recovery, and secondary decline after post-operative recovery. Multilinear regression was conducted to determine associations between epoch periods and the presence of secondary mobility decline.
Results: A total of 41 patients (18 LD and 23 LF) were included. Periods of acute pre-operative decline and post-operative recovery had a mean duration of 146 (standard deviation [SD] = 167) and 216 (SD=190) days, respectively. 34.1% of the patients experienced a secondary functional decline after initial sustained increase in physical activity after surgery. Regression tests demonstrated a significant negative association (p = 1.4*10-9) between the length of immediate post-operative recovery and the subsequent occurrence of secondary functional decline. None of the patients whose initial post-operative recovery lasted longer than 300 days experienced a post-recovery decline.
Conclusion : Our results demonstrate an inverse linear relationship between the duration of immediate post-operative recovery and the occurrence of a post-recovery decline in physical activity, suggesting that longer initial recovery periods are protective against subsequent secondary functional declines. These data-driven analytics can potentially identify patients who are more likely experience adverse long-term outcomes during the pre- and early post-operative period. Future studies should investigate other clinical variables that can stratify patient outcomes and potentially inform prognostic algorithms.
