Measuring depression severity in general practice: discriminatory performance of the PHQ-9, HADS-D, and BDI-II

Background
The UK Quality and Outcomes Framework (QOF) rewards practices for measuring symptom severity in patients with depression, but the endorsed scales have not been comprehensively validated for this purpose.

Aim
To assess the discriminatory performance of the QOF depression severity measures.

Design and setting
Psychometric assessment in nine Scottish general practices.

Method
Adult primary care patients diagnosed with depression were invited to participate. The HADS-D, PHQ-9, and BDI-II were assessed against the HRSD-17 interview. Discriminatory performance was determined relative to the HRSD-17 cut-offs for symptoms of at least moderate severity, as per criteria set by the American Psychiatric Association (APA) and NICE. Receiver operating characteristic curves were plotted and area under the curve (AUC), sensitivity, specificity, and likelihood ratios (LRs) calculated.

Results
A total of 267 were recruited per protocol, mean age = 49.8 years (standard deviation [SD] = 14.1), 70% female, mean HRSD-17=12.6 (SD = 7.62, range = 0-34). For APA criteria, AUCs were: HADS-D = 0.84; PHQ-9 = 0.90; and BDI-II = 0.86. Optimal sensitivity and specificity were reached where HADS-D =9 (74%, 76%); PHQ-9 =12 (77%, 79%), and BDI-II =23 (74%, 75%). For NICE criteria: HADS-D AUC = 0.89; PHQ-9 AUC = 0.93; and BDI-II AUC = 0.90. Optimal sensitivity and specificity were reached where HADS-D =10 (82%, 75%), PHQ-9 =15 (89%, 83%), and BDI-II =28 (83%, 80%). LRs did not provide evidence of sufficient accuracy for clinical use.

Conclusion
As selecting treatment according to depression severity is informed by an evidence base derived from trials using HRSD-17, and none of the measures tested aligned adequately with that tool, they are inappropriate for use.