Defining and predicting outcomes of a work-injuries rehabilitation program through multivariate baseline phenotypes and longitudinal modeling
Mohammad Bayattork, Tom Carter, David M. WaltonObjective
To characterize short-term pain and health-related quality-of-life (HRQoL) recovery trajectories in injured workers undergoing standardized musculoskeletal rehabilitation. We also examined whether multidimensional baseline phenotypes derived from brief self-report measures predicted these trajectories and subsequent recovery.
Methods
Adults with work-related musculoskeletal injuries were enrolled at entry into a community-based rehabilitation program (N = 146), of whom 134 contributed to pain and 138 to HRQoL trajectory analyses. Participants were classified into one of 4 latent profiles at intake from a prior analysis: Localized/Resilient (57.0%), Mood-Dominant (21.8%), Neurosensory-Dominant (12.7%), and Complex/Central (8.5%). Pain (Numeric Pain Rating Scale, NPRS) and HRQoL (EQ-5D-5L index) were measured at baseline, 4, and 8 weeks; self-rated global recovery, self-reported return-to-work status, and administrative discharge/ongoing-care codes were collected at 8 weeks. Growth mixture modeling defined NPRS and EQ-5D-5L trajectories. Odds ratios (ORs) quantified associations between baseline classes and “No Recovery/No Improvement” trajectories, as well as point outcomes.
Results
Three pain trajectories (Rapid Recovery 34.1%, Incomplete Recovery 42.2%, No Recovery 23.7%) and two QoL trajectories (Good QoL 86.2%, No Improvement 13.8%) were identified. Compared with the Localized/Resilient baseline class, odds of NPRS No Recovery were higher in the Mood-Dominant (OR 3.97, 95%CI 1.32–11.92), Neurosensory‑Dominant (OR 10.99, 3.20–37.38), and Complex, Central (OR 22.67, 4.76–107.90) classes, with similarly graded increases for EQ‑5D‑5L No Recovery (ORs 9.00, 7.50, and 150.00, respectively), although estimates for the small Complex/Central subgroup were imprecise. The Complex /Central class also had higher odds of self-reported “not fully returned to work” (OR 4.18, 1.05–16.64) and being coded as “requires ongoing care” (OR 5.74, 1.93–17.11) relative to Localized/Resilient, whereas self-rated global recovery showed no significant between-class differences.
Conclusions
Brief, fully self-reported, multivariate phenotypes obtained at program entry stratify acutely injured workers into likely rehabilitation outcomes in the short term. These findings support the potential value of integrating mechanism-informed phenotyping with trajectory modeling to inform stratified, risk-guided musculoskeletal rehabilitation.