Unidimensional or multidimensional? Revisiting the psychometrics of PHQ-9 and EPDS using bifactor model and item response theory in 2939 Chinese perinatal women

Abstract

BackgroundPerinatal depression often goes undetected and untreated in low- and middle-income countries like China. Reliable screening tools can improve this situation. The Patient Health Questionnaire-9 (PHQ-9) and the Edinburgh Postnatal Depression Scale (EPDS), two widely used tools, often exhibit inconsistent factor structures, leading to debates regarding their unidimensionality versus multidimensionality and casting doubts on their psychometric properties. MethodsOur study aimed to assess the utility of PHQ-9 and EPDS in Chinese perinatal women and to address the debate by employing the bifactor model and item response theory (IRT). We enrolled 2939 perinatal women from a maternity and infant health hospital serving all 16 districts of Shanghai. The bifactor model was used to examine the factor structure of PHQ-9 and EPDS, while IRT analysis evaluated the psychometric properties. ResultsThe indices derived from the bifactor model indicated that both PHQ-9 and EPDS should be used as unidimensional measurements. All items in PHQ-9 and EPDS showed adequate discriminative ability and difficulty, but certain items require further refinement. PHQ-9 demonstrated better measurement precision at high levels of latent depression than EPDS. LimitationsThese findings might not generalize to perinatal women in impoverished areas. The absence of clinical diagnoses limited the exploration of sensitivity and specificity. ConclusionsPHQ-9 and EPDS are effective tools for detecting depression in Chinese perinatal women and should be used as unidimensional tools. Our study expands upon existing psychometric findings related to PHQ-9 and EPDS, offering valuable insights for their application in research and clinical settings.