The predictive value of self-report scales compared with physician diagnosis of depression in hemodialysis patients

Abstract

The prevalence of depression in end-stage renal disease (ESRD) patients on hemodialysis has not been definitively determined. We examined the prevalence of depression and the sensitivity, specificity, positive, and negative likelihood ratios (+LR and -LR) of self-report scales using the physician-administered Structured Clinical Interview for Depression (SCID) as the comparison. Ninety-eight consecutive patients completed the Beck Depression Inventory (BDI) and the Center for Epidemiological Study of Depression (CESD) scales. A physician blinded to BDI and CESD scores administered the SCID. Receiver/responder operating characteristic curves determined the best BDI and CESD cutoffs for depression. Depressed patients had more co-morbidities and lower quality of life, P<0.05. The prevalence of depression by SCID was 26.5% and of major depression was 17.3%. The CESD cutoff with the best diagnostic accuracy was 18, with sensitivity 69% (95% confidence interval (CI) (51%, 87%)), specificity 83% (95% CI (74%, 92%)), positive predictive value (PPV) 60%, negative predictive value (NPV) 88%, +LR 4.14, and -LR 0.37. The best BDI cutoff was 14, with sensitivity 62% (95% CI (43%, 81%)), specificity 81% (95% CI (72%, 90%)), PPV 53%, NPV 85%, +LR 3.26, and -LR 0.47. Self-report scales have high +LR but low -LR for diagnosis of depression. When used for screening, the threshold for depression should be higher for ESRD compared with non-ESRD patients. Identifying depression using physician interview is important, given the low -LR of self-report scales.