Direct Comparison of Three Postpartum Hemorrhage Risk Assessment Tools: A Retrospective Secondary Data Analysis

Objective The objective was to compare the performance of three postpartum hemorrhage risk screening tools with a known patient data set. Design This is a retrospective secondary data analysis of a parent study whose aim was to develop a novel biomarker for detecting elevated blood loss with childbirth. Setting A single tertiary care hospital in the United States. Participants Inclusion criteria for the parent study were ≥ 37 weeks, ≥ 18 years, and pregnancy with one live fetus (no multiples). Eligible participants were identified upon admission to the labor unit and enrolled after informed consent was obtained. Methods/Main Outcome Measures Statistical analyses were computed using Stata SE, v18. Bivariate associations between binary outcomes and continuous measures were calculated using independent t ‐tests. Bivariate associations between binary outcomes and categorical measures were calculated using chi‐square tests. Results Recruitment occurred June through September 2021 and included n = 525 participants. When using the American College of Obstetricians and Gynecologists’ threshold of ≥ 1000 mL, 13.3% (n = 70) of parturients experienced a hemorrhage. When using the World Health Organization’s definition of postpartum hemorrhage of ≥ 500 mL, the prevalence of hemorrhage was 36.8% ( n = 193). Sensitivity was highest in identifying hemorrhage with the Association of Women’s Health, Obstetric and Neonatal Nurses’ tool. When high‐ and medium‐risk scores were considered, the tool had 86% (95% CI: 83.0%–89.0%) accuracy in identifying cases of hemorrhage. This tool also had the highest negative predictive value (78.7%, CI 75.2%–82.2%). Conclusions The tool with the highest sensitivity and negative predictive values identified 21% of cases as “low risk” that went on to experience blood loss ≥ 500 mL, thus underscoring the need for better predictive models.