Data-driven phenotyping of preoperative functional decline patterns in patients undergoing lumbar decompression and lumbar fusion using smartphone accelerometry

Abstract

OBJECTIVE Treatment of degenerative lumbar spine pathologies typically escalates to surgical intervention when symptoms begin to significantly impair patients’ functional status. Currently, surgeons rely on subjective patient assessments through patient-reported outcome measures to estimate the decline in patient wellness and quality of life. In this analysis, the authors sought to use smartphone-based accelerometry data to provide an objective, continuous measurement of physical activity that might aid in effective characterization of preoperative functional decline in different lumbar spine surgical indications. METHODS Up to 1 year of preoperative activity data (steps taken per day) from 14 patients who underwent lumbar decompression and 15 patients who underwent endoscopic lumbar fusion were retrospectively extracted from patient smartphones. A data-driven algorithm was constructed based on 10,585 unique activity data points to identify and characterize the functional decline of patients preceding surgical intervention. Algorithmic estimation of functional decline onset was compared with reported symptom onset in clinical documentation across patients who presented acutely (≤ 5 months of symptoms) or chronically (> 5 months of symptoms). RESULTS The newly created algorithm identified a statistically significant decrease in physical activity during measured periods of functional decline (p = 0.0020). To account for the distinct clinical presentation phenotypes of patients requiring lumbar decompression (71.4% acute and 28.6% chronic) and those requiring lumbar fusion (6.7% acute and 93.3% chronic), a variable threshold for detecting clinically significant reduced physical activity was implemented. The algorithm characterized functional decline (i.e., acute or chronic presentation) in patients who underwent lumbar decompression with 100% accuracy (sensitivity 100% and specificity 100%), while characterization of patients who underwent lumbar fusion was less effective (accuracy 26.7%, sensitivity 21.4%, and specificity 100%). Adopting a less-permissive detection threshold in patients who underwent lumbar fusion, which rendered the algorithm robust to minor fluctuations above or below the chronically decreased level of preoperative activity in most of those patients, increased functional decline classification accuracy of patients who underwent lumbar fusion to 66.7% (sensitivity 64.3% and specificity 100%). CONCLUSIONS In this study, the authors found that smartphone-based accelerometer data successfully characterized functional decline in patients with degenerative lumbar spine pathologies. The accuracy and sensitivity of functional decline detection were much lower when using non–surgery-specific detection thresholds, indicating the effectiveness of smartphone-based mobility analysis in characterizing the unique physical activity fingerprints of different lumbar surgical indications. The results of this study highlight the potential of using activity data to detect symptom onset and functional decline in patients, enabling earlier diagnosis and improved prognostication.

Feasibility and Acceptability of a Preoperative Multimodal Mobile Health Assessment in Spine Surgery Candidates

“…Studies to date have focused on prolonged perioperative activity monitoring as a way of objectively measuring surgical outcomes. 11,12,14,15 These efforts have typically used activity data from smartphones or, rarely, participants' own wearable trackers. 13 Although such efforts have clear value, the prognostic importance of isolated preoperative activity measures and their use in guiding perioperative interventions remain unknown.…”

“…To date, most mHealth studies in spine surgery patients have focused on using activity to objectively measure surgical outcome. [11][12][13][14][15] Nonetheless, growing evidence suggests that the relationship between activity and pain is variable across patients, 16 and it is unknown whether preoperative activity levels alone significantly predict surgical outcome. Consequently, a more personalized approach that integrates physiological, functional, and patientreported data may more effectively predict surgical outcome and guide perioperative management.…”

Digital Biomarkers and the Evolution of Spine Care Outcomes Measures: Smartphones and Wearables

“…Ahmad et al 6 retroactively examined preoperative smartphone accelerometer data in patients who underwent lumbar decompression or lumbar fusion procedures. They developed an algorithm that could predict preoperative functional decline based on activity metrics, which was consistent with reported symptom onset in clinical documentation.…”

“…1 Thus, more recently, there has been a shift toward the development of new objective measures to supplement PROs in the evaluation of outcomes in spine care. 6 Health metrics collected by personal smartphones and wearable devices, so-called "digital biomarkers," are examples of objective data that are accessible, standardized, and personalized. Digital biomarkers have the potential to improve outcome measures and to allow clinicians to continuously monitor individual outcomes and perioperative symptoms.…”

“…1 Thus, more recently, there has been a shift toward the development of new objective measures to supplement PROs in the evaluation of outcomes in spine care. 6…”