Abstract
Purposes and Background
Musculoskeletal disorders are leading causes of work disability. Our purpose was to develop a predictive model in a cohort from 2012 and validate the model in 2016 data.
Methods and Results
Prospectively collected data was used to identify inception cohorts in 2012 (n=1652) and 2016 (n=199). Data from back pain claimants receiving treatment in physiotherapy clinics and the Ontario workers' compensation database were linked. Patients were followed for 1 year.
Variables from a back pain questionnaire and clinical, demographic and administrative factors were assessed for predictive value. The outcome was cumulative number of calendar days receiving wage-replacement benefits.
Cox regression revealed 8 significant predictors of shorter time on benefits in the 2012 cohort: early intervention (HR=1.51), symptom duration < 31 days (HR=0.88), not in construction industry (HR=1.89), high Low Back Outcome Score (HR=1.03), younger age (HR=0.99), higher benefit rate (HR=1.00), intermittent pain (HR=1.15), no sleep disturbance (HR=1.15). The 2012 model c-statistic was 0.73 with a calibration slope of 0.90 (SE=0.19, p=0.61) in the 2016 data, meaning not significantly different. The c-statistic in the 2016 data was 0.69. Median duration on benefits of those with a high risk score was 129 days in 2012 and 45 days in 2016.
Conclusion
The 2012 model had a good fit and calibrated well in the 2016 data. Changes in case management by the WSIB might have resulted in significant differences between the 2012 and 2106 cohort. The difference in duration on benefits could affect treatment decision making in future.
Conflicts of interest: IWH is funded by the Ministry of Labour in Ontario
Sources of funding: Funding was provided by the Canadian Institute for Health Research
