Assessing the performance of a new depression screener for primary care (PC-SAD)

As many as 50% of patients with major depression seen in primary care settings are not diagnosed. To facilitate efficient identification of primary care patients with depression, we developed a new patient-administered depression screening instrument (PC-SAD) that produces a DSM-IV diagnosis, and compared its performance to other screeners that yield DSM-IV diagnoses. To assess validity, the diagnostic accuracy of the PC-SAD was compared with the Inventory to Diagnose Depression (IDD) and the PRIME-MD-PHQ (PHQ) in a convenience sample (N = 312) of health plan members, primary care outpatients, and psychiatric patients (with diagnoses). The screeners were compared with each other and with psychiatric diagnoses to assess their relative performance. Disagreement among the three screeners was formally tested using a triangulation approach that incorporates a statistical likelihood model. Of patients diagnosed as depressed using the IDD, 84.2% were also depressed by the PC-SAD (sensitivity). Of patients not diagnosed as depressed by the IDD, 94.7% were not depressed by the PC-SAD (specificity). Using the triangulation method the sensitivities were 87.2% (PC-SAD), 88.4% (IDD), and 60.7% (PHQ). The specificities were 95.0% (PC-SAD), 92.7% (IDD), and 98.3% (PHQ). The performance of the PC-SAD and the IDD was comparable. The PHQ was less sensitive than either of those. The PC-SAD respondent burden strikes a balance between the very short PHQ, and the longer IDD, and has the lowest (easiest) Flesch-Kincaid reading level. Investigators, clinicians, and health plans that want a DSM-IV-based depression screener can choose from among these three instruments, with known tradeoffs in sensitivity, respondent burden, and readability.