What is the best screening test for depression in chronic spinal pain patients?

BACKGROUND CONTEXT

High prevalence rates of depression have been found in patients with chronic spinal disorder (CSD). The biopsychosocial model has become widely adopted and, with it, the role of psychopathology in the development and/or exacerbation of CSD has become increasingly recognized. Standardized diagnostic criteria, such as the Diagnostic and Statistical Manual of Mental Disorders (DSM), have been used to diagnose major depressive disorder (MDD). Many measures of MDD (and depressive symptom inventories) have been developed during the past 50 years, but their comparative utility in CSD populations is still unclear.

PURPOSE

To systemically compare the performance of depression screening questionnaires in detecting MDD among a large sample of patients with CSD.

STUDY DESIGN/SETTING: Prospective cohort study comparing the screening ability of four popular depression measures for diagnosing MDD against the "gold standard" Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I), using a receiver operating characteristic (ROC) analysis in a CSD population.

PATIENT SAMPLE

A consecutive cohort of 546 patients with CSD admitted to an interdisciplinary functional restoration program.

OUTCOME MEASURES

Sensitivity, specificity, ROC curves, area under the curve (AUC), and optimal cutoff points that are most closely related to the prevalence rates of MDD, with balanced sensitivity and specificity analysis.

METHODS

Using the SCID-I diagnosis as a "gold standard," the ability of four screening measures in detecting MDD were compared. These included: the Beck Depression Inventory (BDI); Hamilton Rating Scale for Depression (HRSD); 9-Item Patient Health Questionnaire Depression Module (PHQ-9); and the Short Form-36 (SF-36).

RESULTS

Of 542 CSD patients, 331 (61.1%) were diagnosed with MDD by the SCID-I. Results of the ROC analysis revealed that the BDI (AUC 0.768), HRSD (AUC 0.796), and PHQ-9 (AUC 0.768) have similar abilities to discriminate between depressed and nondepressed patients in this population. These depression measures outperformed the two mental health scales derived from the SF-36 (Mental Component Summary score/5-Item Mental Health Index; AUC 0.679-0.715). The optimal cut-off scores of 15 (for the BDI), 17 (for the HRSD), and 10 (for the PHQ-9) were also determined. Although the greatest overall accuracy (sensitivity of 81.3% and specificity of 65.4%) was obtained with the HRSD, it is the only clinician-administered instrument. Self-report measures of depression (the BDI and PHQ-9) showed comparable abilities to detect depression, only slightly less than the HRSD.

CONCLUSIONS

Compared to the HRSD, both BDI and PHQ-9 are relatively short and easy to self-administer. The cut-off scores established in this study may be used to reliably determine whether a person should be evaluated more thoroughly for an MDD diagnosis. Using an acknowledged "gold standard," the HRSD, BDI and PHQ-9 showed similar validity to recommend their use for future clinical and research purposes. The SF-36 is less appropriate for diagnosing MDD.