More Granular Risk Score May Improve Maternal Morbidity Prediction

A novel maternal Prenatal Risk Index (m-PRI) model that weighs together 28 maternal risk factors based on their severity on an ordinal scale improved upon previous models considering a narrower set as binary factors, a retrospective cohort study found.

The m-PRI model correctly predicted severe maternal morbidity with greater accuracy than did the Obstetric Comorbidity Scoring System (OCSS) model, which uses a binary composite value similar to other maternal risk scores. The precision-recall area under the curves were 0.223 and 0.173, respectively, for a modest but significant mean difference of 0.050 (95% CI 0.036-0.064, P<0.001), reported Deborah Kilday, MSN, RN, of the HHS Office on Women's Health in Rockville, Maryland, and colleagues in Lancet Regional Health Americas.

“Our work here uses this mortality-informed measure of severe maternal morbidity to ensure that we are defining severe maternal morbidity based on the most critical clinical outcomes, ” Kilday told MedPage Today. “The biggest takeaway is that we have identified the critical factors that impact the poorest outcomes for women during the time of delivery.”

Placenta accreta spectrum was weighted the most, followed by chronic renal disease, acquired cardiac disease, pulmonary hypertension, and infection of the amniotic sac and membranes. Most, but not all, of the risk factors included in m-PRI were also included in the OCSS composite.

David Hackney, MD, a maternal-fetal medicine physician at University Hospitals Cleveland Medical Center, who was not involved in the research, said “modeling obstetric morbidity is complicated by the predictors and outcomes both being quite heterogenous” and because any model will be “limited by adverse obstetric events that randomly arise in healthy patients with no risk factors at all.”

While this m-PRI model improved upon previous models, the authors said it “may provide greater utility for patient-level care in clinical practice,” such as triage transfers to a level IV maternal care facility.

Hackney, though, suggested that clinical utility will remain limited, especially on the individual patient level. Still, he said models like this “allow for risk-stratification of differing populations for comparative and public health purposes.”

In the last 5 years, maternal mortality rates in the U.S. haven’t improved to a meaningful degree overall and severe maternal morbidity has risen. Women with multiple comorbid conditions — such as infections, mental health disorders, cardiovascular conditions, and obesity — have elevated risk of maternal mortality, authors noted. Having multiple comorbid conditions “is recognized as a driver of adverse outcomes affecting both women and infants,” though it’s not well understood.

For this study, the researchers developed and validated the m-PRI from preexisting maternal conditions and patient characteristics known to be associated with severe maternal morbidity. They did so using data from the Premier Healthcare Database, a comprehensive, all-payer administrative database comprising a quarter of annual inpatient encounters in the U.S. The data included detailed information on patient demographics, disease states, and services provided.

“By incorporating a more expansive set of relevant maternal risk conditions, the m-PRI detects previously unmeasured risk, both overall and across clinical subgroups,” authors wrote.

Kilday’s group analyzed all inpatient deliveries between 2016 and 2023 across 864 hospitals in 49 states. Out of more than 7 million maternal delivery hospital encounters, 56,612 patients — about 1% — had at least one severe maternal morbidity indicator. Eligible patients delivered at 20 weeks’ gestation or later. Those who pursued abortions or had unknown gestational age were excluded.

The study population comprised 63% white, 15% Black, 5% Asian, and 1% American Indian or Alaska Native individuals. The mean age was 29.1 years. About half (52%) had commercial insurance and 42% had Medicaid. Black women and those on Medicaid disproportionately had higher levels of severe maternal morbidity.

The researchers noted some limitations, including reliance on ICD-10 codes, the impact of the COVID pandemic on patient care, and the possibility of overestimating risk. Also, the model is predictive and does not imply causal relationships.

“Further validation in diverse settings and exploration of integration into clinical workflows are warranted,” the group said.