Comparative Analysis of Obstetric Hemorrhage Risk Prediction Tools

Abstract

Objective Hemorrhage risk prediction tools were developed in response to rising rates of obstetric hemorrhage (OBH). The California Maternal Quality Care Collaborative (CMQCC) risk prediction tool classifies patients as low, medium, and high risk for OBH based on individual risk factors. At our institution, Magee-Womens Hospital (MWH), a unique OBH risk prediction tool was derived from the CMQCC tool that differs through its use of weighted risk factors and distinctive laboratory value cutoffs. Our objective is to compare this enhanced institution-specific tool to the CMQCC tool. Study Design This study was a retrospective cohort analysis of delivery admissions from a single health care network. Admission OBH risk scores were assigned to each patient using both the MWH and CMQCC scores. Cohen's kappa estimated agreement. Scoring systems and maternal outcomes were compared using chi-square test. Composite morbidity included transfusion, hysterectomy, uterine artery embolization, and intensive care unit admission. Results A total of 21,843 delivery admissions were included. A moderate association was observed between scoring systems (kappa 0.41, p < 0.001). The CMQCC tool categorized 16,184 (74%) patients as low risk, 4,664 (21%) as medium risk, and 995 (5%) as high risk. The MWH tool categorized 13,137 (60%) patients as low risk, 8,113 (37%) as medium risk, and 593 (3%) as high risk. The MWH score recategorized CMQCC low-risk patients to a higher stratum 26% of the time. CMQCC high-risk patients were recategorized to a lower stratum 82% of the time. Both the MWH and CMQCC tools were able to differentiate OBH-related morbidity across risk strata. The MWH tool independently predicted risk of composite morbidity within each stratum of the CMQCC score. Conclusion Both the MWH and CMQCC tools independently distinguish risk of composite morbidity. Adding weighted values to individual risk factors further discriminates risk of morbidity. This suggests it may be reasonable to adapt the CMQCC tool to reflect institutional populations and resources. Key Points