Psychometric properties of the EPDS-3A across pregnancy and postpartum: Factor structure and invariance in a nationwide sample.

Abstract

Although the three anxiety-related items of the Edinburgh Postnatal Depression Scale (EPDS-3A) are widely used to flag anxiety-related distress, their psychometric performance across perinatal phases remains insufficiently established. We evaluated factor structure, measurement invariance, differential item functioning (DIF), score precision and convergence with Generalized Anxiety Disorder-7 (GAD-7) thresholds in a nationwide Italian sample of 10,100 individuals (4,410 pregnant; 5,690 postpartum). Participants completed the EPDS and GAD-7. Analyses included exploratory graph analysis, tau-equivalent ordered-categorical confirmatory factor analysis, multi-group invariance testing, DIF analysis, graded-response item response theory and receiver-operating characteristic curves. The three anxiety-related items formed a distinct community, with stronger clustering postpartum. Tau-equivalent one-factor fit was stronger postpartum and modest during pregnancy; scalar invariance was supported. DIF was negligible, and item response theory analyses indicated ordered thresholds and meaningful item discrimination. The EPDS-3A showed moderate convergence with GAD-7 generalized anxiety symptom thresholds. A score of 5 was the Youden-selected EPDS-3A threshold for distinguishing participants above versus below moderate or severe GAD-7 thresholds in both phases. The EPDS-3A yields a brief, psychometrically coherent anxiety-related score across pregnancy and postpartum, but convergence with worry-related GAD-7 symptoms does not establish diagnostic or broad perinatal anxiety screening accuracy.