OBJECTIVES. The purpose of this study was to examine factors predictive of duration of work-related disability. METHODS. Multivariate survival analysis techniques were used to conduct a population-based, retrospective cohort study on a random sample of 28,473 workers' compensation claims from Washington State filed for injuries occurring in 1987 to 1989. The principal outcome measure was length of time for which compensation for lost wages was paid, used as a surrogate for duration of temporary total disability. RESULTS. The findings suggest that, even after adjusting for severity of injury, older age, female gender, and a diagnosis of carpal tunnel syndrome or back/neck sprain significantly predict longer duration of disability. Other predictors that were stable and significant, but involved lower magnitudes of effect included divorced marital status, firm size of fewer than 50 employees, higher country unemployment rates, and construction and agricultural work. CONCLUSIONS. Greater disability prevention efforts targeting these higher risk subgroups could have significant economic and public health effects. The greatest impact may be on claimants who remain disabled at 6 months after an injury that did not require hospitalization.