A novel postpartum hemorrhage prediction model for women with hypertensive disorders of pregnancy in secondary hospitals of middle-income settings

Background This study aimed to develop and validate a risk-stratification tool to aid clinical referral decision-making for postpartum hemorrhage among women with hypertensive disorders of pregnancy. The tool targets secondary hospitals with basic laboratory capacity in middle-income settings. Methods This was a retrospective study. 2,761 women diagnosed with hypertensive disorders of pregnancy formed the development set (training and internal validation sets), and 592 women from the same hospital constituted a temporal validation set to evaluate model performance. LASSO regression was applied for optimal feature selection, and the prediction model was constructed using multivariable logistic regression. Model performance was evaluated via the area under the receiver operating characteristic curve, calibration curves, decision curve analysis, and clinical impact curves. Results Key predictors, including placental abruption, placenta previa, placenta accreta, macrosomia, hemoglobin level, international normalized ratio, D-dimer concentration, and parity, were incorporated into a nomogram. The model achieved areas under the curve of 0.733 (training set), 0.756 (internal validation set), and 0.877 (temporal validation set). The apparent and bias-corrected calibration curves deviated from the ideal line at low predicted probabilities and gradually approached it as predicted PPH risk increased. Decision curve analysis indicated that the model yielded a higher net benefit compared with treat-all and treat-none strategies, whereas the clinical impact curve characterized the counts of predicted high-risk women and observed PPH events across risk thresholds. Conclusion The model demonstrated moderate discrimination and may be suitable for stratifying postpartum hemorrhage risk among women with hypertensive disorders of pregnancy.