Smartwatch-derived digital biomarkers distinguish episodic pain phenotypes in chronic low back pain

Abstract This study evaluated smartwatch-derived digital biomarkers for chronic low back pain phenotyping. Temporal features extracted from six months of smartwatch data (resting heart rate, heart rate variability, step counts) in 261 patients discriminated episodic from non-episodic pain phenotypes with 82% accuracy. Frequency-domain and local variability measures outperformed summary statistics, with combined heart and activity features providing superior classification. These findings demonstrate feasibility of passive wearable monitoring for objective pain phenotyping.