Prognostic accuracy of clinical markers of postpartum bleeding in predicting maternal mortality or severe morbidity: a WHO individual participant data meta-analysis

Background Postpartum haemorrhage (excessive bleeding after birth) is a leading cause of maternal mortality and morbidity worldwide. However, there is no global consensus on which clinical markers best define excessive bleeding or reliably predict adverse maternal outcomes. The aim of this study was to assess the prognostic accuracy of clinical markers of postpartum bleeding in predicting maternal mortality or severe morbidity. Methods In this individual participant data meta-analysis, eligible datasets were identified through a global call for data issued by WHO and systematic searches of PubMed, MEDLINE, Embase, the Cochrane Library, and WHO trial registries (from database inception to Nov 6, 2024). Studies were eligible if they included at least 200 participants with objectively measured blood loss or other clinical markers of haemodynamic instability, and reported at least one clinical outcome of interest. Individual participant data were requested for all eligible studies. For each dataset, we computed the prognostic accuracy of each clinical marker to predict a composite outcome of maternal mortality or severe morbidity (blood transfusion, surgical interventions, or admission to intensive care unit). Five clinical markers were assessed: measured blood loss, pulse rate, systolic blood pressure, diastolic blood pressure, and shock index. Results were meta-analysed through two-level mixed-effects logistic regression models, with a bivariate normal model used to generate summary accuracy estimates. Clinical marker and threshold selections were informed by a WHO expert consensus process, which placed emphasis on maximising prognostic sensitivity (preferably >80%) over prognostic specificity (preferably ≥50%). This meta-analysis was registered on PROSPERO (CRD420251034918). Findings We identified 33 potentially eligible datasets and successfully obtained and analysed full data for 12 datasets, comprising 312 151 women. At the conventional threshold of 500 mL, measured blood loss had a summary prognostic sensitivity of 75·7% (95% CI 60·3–86·4) and specificity of 81·4% (95% CI 70·7–88·8) for predicting the composite outcome. The preferred sensitivity threshold was reached at 300 mL (83·9% [95% CI 72·8–91·1]), although at the expense of reduced specificity (54·8% [95% CI 38·0–70·5]). Prognostic performance improved with a decision rule that combined the use of either blood loss thresholds less than 500 mL (≥300 mL to ≥450 mL) and any abnormal haemodynamic sign (pulse rate >100 beats per min, systolic blood pressure <100 mm Hg, diastolic blood pressure <60 mm Hg, or shock index >1·0) or 500 mL or more of blood loss, with sensitivities ranging from 86·9% to 87·9% and specificities from 66·6% to 76·1%. Interpretation Measured blood loss below the conventional threshold, combined with abnormal haemodynamic signs, accurately predicts women at risk of death or life-threatening complications from postpartum bleeding and could support earlier postpartum haemorrhage diagnosis and treatment. Funding The Gates Foundation and UNDP/UNFPA/UNICEF/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction.