Aim To develop a predictive screener that will accurately predict the future development of postinjury depression or post-traumatic stress disorder (PTSD).

Methods A prospective, longitudinal cohort study of subjects randomly selected from patients requiring an Emergency Department (ED) visit for injury. A theoretically and empirically derived 45-item risk survey was administered soon after injury. A psychiatric diagnostic interview was conducted at 3 and 6 months to determine emergence of depression or PTSD. Multiple logistic regression analyses were performed separately for depression and PTSD, using a best-subset approach to select models at each step by comparing log-likelihood estimates.

Results Of 175 (of 192 enrolled) available for follow-up, 16% developed depression and 2.3% developed PTSD. A logistic model retaining four depression items and five PTSD items emerged. One item predicted both depression and PTSD; hence the final screener is eight items. Receiver operating characteristic for depression and PTSD was 0.81 each. Optimal cut-off scores were >2 (out of 4) for depression and >3 (out of 5) for PTSD. Depression findings: sensitivity 0.81, specificity 0.71, PPV 0.34, NPV 0.95. PTSD findings: sensitivity 1.00, specificity 0.66, PPV 0.07, NPV 1.00.

Conclusions Thirty million seek ED care annually for injury and a significant subset develop psychiatric disorders that that impede recovery. The high NPV for both depression and PTSD supports the role of this new instrument as a short, clinical screener: those who screen negative can be screened out from further monitoring, allowing precious resources to be focused on those at highest risk for the development of depression and PTSD.