Open Access Research article

Utility and limitations of PHQ-9 in a clinic specializing in psychiatric care

Takeshi Inoue*, Teruaki Tanaka, Shin Nakagawa, Yasuya Nakato, Rie Kameyama, Shuken Boku, Hiroyuki Toda, Tsugiko Kurita and Tsukasa Koyama

Author Affiliations

Department of Psychiatry, Hokkaido University Graduate School of Medicine, North 15, West 7, Kita-ku, Sapporo, 060-8638, Japan

For all author emails, please log on.

BMC Psychiatry 2012, 12:73  doi:10.1186/1471-244X-12-73

Published: 3 July 2012



The Patient Health Questionnaire-9 (PHQ-9), despite its excellent reliability and validity in primary care, has not been examined for administration to psychiatric patients. This study assesses the accuracy of PHQ-9 in screening for major depressive episode and in diagnosing major depressive episode in patients of a psychiatric specialty clinic.


We compared operational characteristics of PHQ-9 as a screening and diagnostic instrument to DSM-IV-TR diagnosis by a trained psychiatrist as a reference standard. The reference criteria were “current major depressive episode” or “current major depressive episode with major depressive disorder”. PHQ-9 was used with two thresholds: diagnostic algorithm and summary scores (PHQ-9 ≥ 10). The optimal cut-off points of PHQ-9 summary scores were analyzed using a receiver operational characteristics (ROC) curve.


For “current major depressive episode”, PHQ-9 showed high sensitivity and high negative predictive value at both thresholds, but its specificity and positive predictive value were low. For “current major depressive episode with major depressive disorder”, PHQ-9 also showed high sensitivity and high negative predictive value at both thresholds, but the positive predictive value decreased more than that for “current major depressive episode”. The ROC analysis showed the optimal cut-off score of 13/14 for “current major depressive episode”.


PHQ-9 is useful for screening, but not for diagnosis of “current major depressive episode” in a psychiatric specialty clinic.