Comparison of back pain prognostic risk stratification item sets

Back pain outcomes may be improved and costs lowered through risk-stratified care, but relative performance of alternative item sets for predicting back pain outcomes has not been well characterized. We compared alternative prognostic item sets based on STarT Back and Chronic Pain Risk screeners in a cohort of patients initiating primary care for back pain. The STarT Back item set was brief and relied on binary responses, whereas the Chronic Pain Risk item set employed scaled responses and assessed pain persistence and diffuse pain. Patients (N = 571) were assessed soon after their initial visit and 502 (88%) were reassessed 4 months later. Items sets based on STarT Back and Chronic Pain Risk prognostic screeners, as well as a combination of items from both, were used to predict Chronic Pain Grade II-IV back pain at 4 months. The area under the receiver operating characteristic curve estimates (95% confidence intervals) were .79 (.74-.83) for items based on the STarT Back, .80 (.75-.83) for items based on Chronic Pain Risk, and .81 (.77-.85) for a composite item set. Differences in prediction were modest. Items from 2 prognostic screeners, and both combined, achieved acceptable and similar prediction of unfavorable back pain outcomes.

Given comparable predictive validity, choice among prognostic item sets should be based on clinical relevance, number of items, ease of administration, and item simplicity.