Postpartum Depression: Overlap Between Symptom‐Based and Register‐Based Measures

INTRODUCTION: Postpartum depression (PPD) can be measured in various ways, including questionnaire-based symptom scores or through administrative health databases. We examined the overlap between symptom-based and register-based PPD definitions and assessed the consistency in exposure-outcome associations. METHODS: We linked Danish nationwide health registers to PPD screening records (Edinburgh Postnatal Depression Scale, EPDS) at 2 months postpartum from women screened between January 1, 2015, and December 31, 2021. We defined symptom-based PPD as an EPDS ≥ 11, and register-based PPD as an antidepressant prescription or hospital depression diagnosis within 1 year postpartum. We estimated the overlap between women with register-based and symptom-based PPD, as well as between continuous EPDS scores and register-based treatment indicators. We evaluated consistency between the two definitions using logistic regression analyses for socioeconomic, obstetric, demographic- and health-related exposures from which we obtained odds ratios (ORs) for each PPD definition and calculated ratio of odds ratios (ROR). RESULTS: Among 157,193 mothers (132,593 unique), 11,193 (7.1%) had symptom-based PPD and 2409 (1.5%) had register-based PPD. Of those with symptom-based PPD, 8.8% also had register-based PPD, while 40.8% of those with register-based PPD had symptom-based PPD. We observed a higher overlap with higher EPDS scores. Consistency was highest for obstetric variables but varying for demographic- and health-related exposures (ROR register-based vs. symptom-based 1.84 [95% CI 1.67-2.03] for psychiatric history and 0.75 [95% CI 0.68-0.83] for primiparity). CONCLUSION: We observed substantial differences and limited overlap between symptom-based and register-based PPD. Register-based measures capture hospital-based diagnoses and pharmacological treatment only, but not all treatment within the healthcare system, and differences may reflect a combination of severity, symptom transience, misclassification, and timing of measurement. Consistency in exposure-outcome associations varied, particularly for certain exposures. Our findings underline the importance of considering both the nature and implications of PPD definitions in epidemiological research.