Presentation (CTI)

OBJECTIVES: Routine depression screening in maternal healthcare can aid in preventing and controlling perinatal depression which is linked to adverse obstetric outcomes. However, the effectiveness of the 10-item Edinburgh Postnatal Depression Scale (EPDS) for screening could be limited in contexts that require large-scale, rapid administration, and particularly when used across multi-ethnic populations. Therefore, this study aims to equip healthcare workers at primary level with a brief, practical tool to improve screening efficiency.
METHODS: In May 2022, EPDS was used to screen 1,191 women from pregnancy to one year postpartum in an ethnic-minority autonomous county of Yunnan Province. Data were randomly divided into training set and test set in a ratio of 7:3. Risk-Calibrated Supersparse Linear Integer Model (RiskSLIM) was utilized for developing short-form EPDS models on the training set with Python 3.13. Models performance were assessed with the Area Under the Receiver Operative Characteristic Curve (AUC) and R-squared (R²). The cultural adaption of models were validated across the datasets of Han majority, Zhuang, Miao, Yao, and other minorities. The optimal cut-off value was determined by Restricted Cubic Splines.
RESULTS: Short-form models of EPDS from 1 to 9 items were constructed with RiskSLIM. Item 2 (can not look forward with enjoyment to things), Item 5 (scared or panicky), item 7 (sleep difficulties), and item 8 (sad or miserable) were selected into the 4-item version EPDS (EPDS-R4), which exhibited consistently high and stable AUC (0.980-0.998) and R² values (0.721-0.921) across multi-ethnic population. The cut-off score of ≥ 10 was recommended.
CONCLUSIONS: EPDS-R4 (cut-off value ≥10) is an efficient and practical screening tool for perinatal depression, which might be adopted in population-based screening program.