Development of prediction models of stress and long-term disability among claimants to injury compensation systems: a cohort study

OBJECTIVES: We sought to develop prognostic risk scores for compensation-related stress and long-term disability using markers collected within 3 months of a serious injury. DESIGN: Cohort study. Predictors were collected at baseline and at 3 months postinjury. Outcome data were collected at 72 months postinjury. SETTING: Hospitalised patients with serious injuries recruited from four major trauma hospitals in Australia. PARTICIPANTS: 332 participants who made claims for compensation for their injuries to a transport accident scheme or a workers’ compensation scheme. PRIMARY OUTCOME MEASURES: 12-item WHO Disability Assessment Schedule and 6 items from the Claims Experience Survey. RESULTS: Our model for long-term disability had four predictors (unemployed at the time of injury, history of a psychiatric disorder at time of injury, post-traumatic stress disorder symptom severity at 3 months and disability at 3 months). This model had good discrimination (R(2)=0.37) and calibration. The disability risk score had a score range of 0–180, and at a threshold of 80 had sensitivity of 56% and specificity of 86%. Our model for compensation-related stress had five predictors (intensive care unit admission, discharged to home, number of traumatic events prior to injury, depression at 3 months and not working at 3 months). This model also had good discrimination (area under the curve=0.83) and calibration. The compensation-related stress risk score had score range of 0–220 and at a threshold of 100 had sensitivity of 74% and specificity of 75%. By combining these two scoring systems, we were able to identify the subgroup of claimants at highest risk of experiencing both outcomes. CONCLUSIONS: The ability to identify at an early stage claimants at high risk of compensation-related stress and poor recovery is potentially valuable for claimants and the compensation agencies that serve them. The scoring systems we developed could be incorporated into the claims-handling processes to guide prevention-oriented interventions.