Psychometric evaluation of the Japanese Edinburgh Postnatal Depression Scale for screening postpartum anxiety

Background Despite the acknowledged importance of addressing postpartum anxiety alongside postpartum depression, standardized screening tools specifically developed for this purpose remain limited. Objective This study aimed to validate the anxiety factor of the Edinburgh Postnatal Depression Scale (EPDS) and to determine optimal cutoff scores for screening postpartum anxiety. Methods EPDS and the State-Trait Anxiety Inventory (STAI) were collected from 100 Japanese women at one month postpartum at Niigata University Medical and Dental Hospital between May 18, 2021, and December 28, 2022, using random convenience and purposive sampling. Confirmatory factor analysis (CFA) was conducted on data from 84 participants to test six previously proposed EPDS factor models, and receiver operating characteristic (ROC) analysis was performed on data from 83 participants to determine area under the curve (AUC) values and cutoff scores. The EPDS anxiety subscales with three items (EPDS-3A) and four items (EPDS-4A) were separately evaluated as predictor variables, with STAI state and trait anxiety as criterion measures. We assessed accuracy, sensitivity, and specificity. Results All models with two-factor and three-factor showed a good fit to the data, with two models with EPDS-3A being superior among other models (comparative fit index = 1.000, root mean square error of approximation = 0.001). ROC analyses indicated good testing accuracy of the EPDS anxiety subscales for detecting both state and trait anxiety. For EPDS-3A, the AUCs were 0.832 (95% CI 0.735–0.930) for state anxiety with an optimal cutoff of ≥3 (sensitivity 79.2%, specificity 79.7%), and 0.912 (95% CI 0.837–0.988) for trait anxiety with an optimal cutoff of ≥4 (sensitivity 82.4%, specificity 84.8%). For EPDS-4A, the AUCs were 0.833 (95% CI 0.736–0.930) for state anxiety with an optimal cutoff of ≥4 (sensitivity 79.2%, specificity 71.2%), and 0.935 (95% CI 0.867–1.000) for trait anxiety with an optimal cutoff of ≥5 (sensitivity 88.2%, specificity 87.9%). Conclusion Both the EPDS-3A and EPDS-4A demonstrated good model fit and screening accuracy for anxiety at one month postpartum. Integrating anxiety screening into routine postpartum care may improve maternal mental health outcomes. Future studies involving clinical settings and larger cohort studies are recommended to improve external validity.