Mortality-weighted severe maternal morbidity: a novel approach to assessing maternal health outcomes

Abstract

Objective: The aim of this study is to develop a robust mortality-weighted severe maternal morbidity (SMM) composite measure that provides increased clinical granularity by measuring the individual and combined risk of inpatient mortality across multiple SMM indicators. The enhanced metric better reflects the complexity gradient of individual patients through a weighted SMM (w-SMM) composite score expressed on a continuous scale. Methods: This retrospective cohort study analyzed 7,174,412 inpatient deliveries from 2016 to 2023 extracted from the Premier Healthcare Database, an all-payer, nationally representative administrative database. Harm weights for each SMM indicator were calculated using the conditional probability of in-hospital maternal mortality given each SMM indicator. The resulting mortality-based harm weights were employed to combine individual SMM indicators into a composite index. The index was scaled to create the in-hospital w-SMM composite score, ranging from 0 to 1. Out-of-sample performance of the proposed w-SMM and established binary SMM (b-SMM) indices was evaluated using an 80:20 derivation-validation split framework. Results: An SMM indicator was identified in 96% of 433 inpatient maternal mortality cases in the study population. The incidence and crude mortality percentages varied by SMM indicator, with cardiac arrest/ventricular fibrillation and conversion of cardiac rhythm having the highest rates of mortality percentages (33% and 29%, respectively), and therefore, the highest severity weights. Approximately 21% of patients identified with an SMM indicator had multiple SMM indicators. The w-SMM composite score, scaled between 0 and 1, accounts for the specific additive risk of multiple indicators, with higher scores indicating higher mortality risk. The w-SMM demonstrated stronger association with in-hospital mortality than the b-SMM measure, with a precision-recall area under the curve (PR-AUC) of 0.269, compared to the current binary composite PR-AUC of 0.007. Conclusion: The mortality-weighted w-SMM index improves upon the binary SMM composite measure by accounting for the varying severity of individual and combined SMM indicators and can further serve as a more granular measure of in-hospital mortality risk. The w-SMM index can be used retrospectively to stratify high-risk patients, enhance risk adjustment models, and support comparative outcomes research that informs public health policy development, future research, policy decisions, and clinical practice.