Smartwatch-derived heart rate and activity distinguish episodic pain subgroups in chronic low back pain

Purpose To determine whether temporal features of six-month smartwatch-derived resting heart rate (RHR), heart rate variability (HRV), and step count trajectories discriminate episodic from non-episodic pain trajectories in chronic low back pain (cLBP). Methods Adults with cLBP (n = 261) from a multicenter observational cohort classified their pain over the preceding six months as episodic or non-episodic using the Visual Trajectories Questionnaire for Pain. Daily smartwatch RHR, HRV, and step count for the same period were retrieved retrospectively. Summary statistics were compared between groups, and between-patient trajectory heterogeneity was quantified with functional principal component analysis. Six hundred time-series features per signal were extracted. Stability selection and elastic-net logistic regression with nested cross-validation were used to evaluate four feature sets: RHR, HRV, RHR + HRV, and RHR + HRV+steps. SHAP values quantified feature importance. Results Of the 261 participants, 195 (74.7%) reported episodic and 66 (25.3%) non-episodic pain. No summary statistic differed between groups for any signal. Explaining 90% of trajectory variance required 6, 12, and 34 functional components for RHR, HRV, and step count, respectively. Nested AUCs were 0.72 (RHR), 0.67 (HRV), 0.78 (RHR + HRV), and 0.82 (RHR + HRV+steps). Frequency-domain coefficients and local variability measures, including day-to-day RHR change and weekly HRV autocorrelation, were the most important features. Conclusion Temporal features of consumer smartwatch data discriminate episodic from non-episodic cLBP, whereas summary statistics of the same signals do not. Combining heart rate and activity features yields the highest discrimination. These findings highlight the potential of biophysical monitoring to better understand and predict recurrent cLBP episodes.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-19
DOI
https://doi.org/10.21203/rs.3.rs-11355046/v1
Primary Topic
Musculoskeletal pain and rehabilitation
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Smartwatch-derived heart rate and activity distinguish episodic pain subgroups in chronic low back pain

Karim Khattab, Jeffrey C. Lotz, Jeannie F. Bailey, Abel Torres‐Espín et al.
Research Square
Musculoskeletal pain and rehabilitation
preprint

Smartwatch-derived heart rate and activity distinguish episodic pain subgroups in chronic low back pain

Karim Khattab, Jeffrey C. Lotz, Jeannie F. Bailey, Abel Torres‐Espín, Gregorij Kurillo, Aniket Pratapneni, Trisha F. Hue, MATTHEW KAUFMAN, Patricia Z. Zheng
preprint en

Abstract

Purpose To determine whether temporal features of six-month smartwatch-derived resting heart rate (RHR), heart rate variability (HRV), and step count trajectories discriminate episodic from non-episodic pain trajectories in chronic low back pain (cLBP). Methods Adults with cLBP (n = 261) from a multicenter observational cohort classified their pain over the preceding six months as episodic or non-episodic using the Visual Trajectories Questionnaire for Pain. Daily smartwatch RHR, HRV, and step count for the same period were retrieved retrospectively. Summary statistics were compared between groups, and between-patient trajectory heterogeneity was quantified with functional principal component analysis. Six hundred time-series features per signal were extracted. Stability selection and elastic-net logistic regression with nested cross-validation were used to evaluate four feature sets: RHR, HRV, RHR + HRV, and RHR + HRV+steps. SHAP values quantified feature importance. Results Of the 261 participants, 195 (74.7%) reported episodic and 66 (25.3%) non-episodic pain. No summary statistic differed between groups for any signal. Explaining 90% of trajectory variance required 6, 12, and 34 functional components for RHR, HRV, and step count, respectively. Nested AUCs were 0.72 (RHR), 0.67 (HRV), 0.78 (RHR + HRV), and 0.82 (RHR + HRV+steps). Frequency-domain coefficients and local variability measures, including day-to-day RHR change and weekly HRV autocorrelation, were the most important features. Conclusion Temporal features of consumer smartwatch data discriminate episodic from non-episodic cLBP, whereas summary statistics of the same signals do not. Combining heart rate and activity features yields the highest discrimination. These findings highlight the potential of biophysical monitoring to better understand and predict recurrent cLBP episodes.

Research Square
University of California, San Francisco (US)
Musculoskeletal pain and rehabilitation
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon