Objective: People with coronary heart disease (CHD) are at heightened risk of depression, and this co-occurrence of conditions is associated with poorer outcomes including raised mortality. This study compares the diagnostic accuracy of two depression case finding instruments in CHD patients relative to a diagnostic standard, the revised Clinical Interview Schedule (CIS-R).
Methods: The Patient Health Questionnaire (PHQ-9), the Hospital Anxiety and Depression Scale depression subscale (HADS-D) and the CIS-R depression module were administered to 803 patients identified from the CHD registers of GP practices in Greater London.
Results: Of 730 recruited patients without previously identified depression, 32 (4.4%) met ICD-10 depressive episode criteria according to the CIS-R. For the PHQ-9 and HADS-D lower cut-points than those routinely recommended were associated with improved case identifying properties. The PHQ-9 appeared the superior instrument using a cut-point of ≥8 (sensitivity=94%; specificity=84%). Using categorical scoring the PHQ-9 was 59% sensitive and 95% specific. For the HADS-D using cut-point ≥5, sensitivity was 81% and specificity was 77%. Areas under the curves (AUC) (standard error) were 0.95 (0.01) and 0.88 (0.02) for the PHQ-9 and HADS-D, and 0.91 (0.02) for PHQ-9 using the categorical algorithm. Statistically significant differences between AUCs of the PHQ-9 and the HADS-D favoured the former. Severity ratings compared across measures indicated inconsistency between recommended bandings: the PHQ-9 categorised a larger proportion of participants with mild and moderate depression.
Conclusion: This is the first large-scale investigation of the accuracy of these commonly used measures within a primary care CHD population. Our results suggest that although both scales have acceptable abilities and can be used as case identification instruments for depression in patients with CHD, the PHQ-9 appeared diagnostically superior. Importantly, optimal cut-off points for depression identification in this population appear to differ from standard values, and severity ratings differ between these measures.