Severe postpartum hemorrhage (sPPH) remains a leading cause of preventable maternal morbidity. We evaluated whether a combined early postpartum model incorporating Shock Index (SI), lactate, and fibrinogen improves 24 h sPPH prediction compared with any single marker.

In this single-center cohort study (January 2019–December 2024), we included deliveries at ≥24 weeks’ gestation in which the first postpartum heart rate and systolic blood pressure (to calculate SI), lactate, and Clauss fibrinogen were all measured within 2 h after placental delivery. The primary outcome was sPPH within 24 h, defined as quantified blood loss ≥1000 mL or bleeding accompanied by hypovolemia. We developed a ridge-penalized logistic regression model using SI, lactate, and fibrinogen and performed internal validation with 1000 bootstrap resamples. We reported discrimination (AUC), calibration (slope and intercept), and Brier score. Single-marker AUCs were compared with the combined model using paired DeLong tests with false discovery rate (FDR) control. Decision-curve analysis evaluated clinical utility at 1%, 2%, and 5% risk thresholds.

Among 12,242 deliveries, 172 sPPH events occurred (1.4%). The median time from placental delivery to sampling was 54 min (IQR 40–78). The combined model achieved an AUC of 0.82 (95% CI 0.77–0.87), with good calibration (slope 0.95; intercept −0.01) and a Brier score of 0.0066. Single-marker AUCs were 0.69 for SI, 0.73 for lactate, and 0.77 for fibrinogen; all were significantly lower than the combined model (FDR-adjusted q ≤ 0.01). At 1%, 2%, and 5% thresholds, sensitivity/specificity were 90.1%/60.0%, 76.2%/82.0%, and 55.2%/93.0%, with net benefit 0.0023, 0.0017, and 0.0002, respectively.

A simple early postpartum model combining SI, lactate, and fibrinogen provided better discrimination than any single marker, with good calibration and positive clinical utility at low-risk thresholds. Multicenter external validation and prospective impact evaluation are warranted before clinical implementation.