Defining and validating a multidimensional digital metric of health states in chronic back and leg pain

Chronic pain (CP) is a debilitating condition that extends beyond persistent pain, influenced by physiological and psychological factors. However, clinical trials often evaluate outcomes solely on self-reported pain amplitude. To address this, we aimed to derive a single metric from multidimensional digital data to comprehensively represent wellness in lower back and leg pain. Daily-reported data were collected for five years (>190 K samples, n = 498, from NCT01719055/NCT03240588), comprised of clinical assessments, digitally-reported symptoms, text responses, and smartwatch-based actigraphy. Clustering analysis of the digital data identified five novel symptom clusters. They were validated by comparing centroid distances to standard assessments, revealing five ordinal best-to-worst states (r = 0.34 to r = -0.51, ps < 0.001), even when pain magnitude was similar. Further, patients' text messages about their status associated better with the clusters than pain reports alone. This solution extends beyond a recapitulation of pain level, yielding non-obvious, meaningful states that serve as an actionable metric in CP care.