(1176) Development and validation of a screening tool incorporating fetal indicators to predict severe maternal morbidity

Objective:
To evaluate whether fetal factors independently predict severe maternal morbidity (SMM) and to compare the predictive performance of fetal, maternal, and combined models for SMM risk stratification.

Study Design:
This retrospective cohort study included 110,319 patients delivering across a large health system from 2019–2023. To ensure independent observations, only each patient’s first delivery was analyzed. The dataset was randomly divided into training, validation, and test subsets (n=36,773 each). Fetal predictors included oligohydramnios, polyhydramnios, fetal growth restriction, large-for-gestational-age (LGA), abnormal umbilical artery Dopplers, and fetal anomaly. Maternal predictors were based on the obstetric comorbidity index (OB-CMI). SMM was defined using CDC criteria. Model performance was evaluated in the test dataset based on the area under the receiver operating characteristic curve (AUC-ROC) derived from logistic regression. Three models were evaluated: (1) maternal factors, (2) fetal factors alone, and (3) combined maternal + fetal model. ROC contrast tests compared AUC performance.

Results:
Fetal factors were independently associated with SMM, but predictive performance was limited (AUC 0.54; 95% CI, 0.53–0.55). Only polyhydramnios and LGA showed significant associations. The maternal model (OBCMI) performed better (AUC 0.65; 95% CI, 0.64–0.67), while the combined maternal + fetal model had the highest performance (AUC 0.69; 95% CI, 0.68–0.70). ROC contrast testing confirmed that the combined model significantly outperformed both fetal-only and maternal-only models (p< 0.001 for both).

Conclusion:
Fetal factors contribute independently to SMM risk but offer limited predictive value on their own. Integrating fetal and maternal data improves model performance, supporting combined risk stratification.