Routinely available clinical and laboratory variables demonstrated potential for early maternal risk stratification among women diagnosed with HELLP syndrome, according to a retrospective cohort study conducted at Fujian Maternity and Child Health Hospital in China. While HELLP syndrome is uncommon, it can lead to serious maternal adverse pregnancy outcomes, prompting researchers to evaluate whether routine medical data could support outcome-specific risk prediction models.
The investigation evaluated records from all women with HELLP syndrome who received treatment at Fujian Maternity and Child Health Hospital between January 2013 and June 2023. Across this ten-year observation window, the study cohort comprised a total of 272 patients whose routinely collected clinical indicators and laboratory measurements were analyzed to assess complication risks.
Researchers modeled five distinct maternal adverse outcomes across the cohort, documenting varying incidence rates among the 272 women. Placental abruption occurred in 33 women, representing 12.13 percent of the cohort, while posterior reversible encephalopathy syndrome was diagnosed in 23 women, or 8.46 percent. Postpartum hemorrhage occurred in 19 patients, accounting for 6.99 percent, disseminated intravascular coagulation developed in 9 women, representing 3.31 percent, and a severe maternal composite outcome occurred in 25 women, or 9.19 percent.
To evaluate predictive performance, the primary analytical framework utilized repeated nested cross-validation across the full patient cohort. Within each outer training fold, the investigators conducted missing-data filtering and imputation, predictor screening, model-family selection, hyperparameter tuning, and decision threshold estimation, ensuring that the held-out validation fold evaluated the complete modeling procedure without data leakage.
The modeling process considered three candidate algorithmic families: ridge logistic regression, elastic-net logistic regression, and random forest. Ridge logistic regression was selected across the majority of outer folds for several complications, specifically chosen in 62 percent of folds for placental abruption, 74 percent for posterior reversible encephalopathy syndrome, 56 percent for postpartum hemorrhage, and 62 percent for the severe composite outcome. For disseminated intravascular coagulation, a single-predictor logistic regression fallback was chosen in 85 percent of the outer folds.
Model discrimination yielded varying area under the receiver operating characteristic curve (AUC) values across the targeted conditions: 0.715 for placental abruption, 0.677 for posterior reversible encephalopathy syndrome, 0.833 for postpartum hemorrhage, 0.907 for disseminated intravascular coagulation, and 0.742 for the severe maternal composite outcome. In a paired sensitivity analysis assessing the replacement of mean and mode imputation with multiple imputation by chained equations (MICE), the AUC shifted by -0.002 to 0.009 and the area under the precision-recall curve (AUPRC) changed by -0.005 to 0.002.
When evaluated against prespecified analysis criteria, models for placental abruption, posterior reversible encephalopathy syndrome, and postpartum hemorrhage failed to meet the required benchmarks. In contrast, models for disseminated intravascular coagulation and the severe composite outcome successfully satisfied the prespecified criteria, though both estimates remained constrained by sparse event numbers and an absence of external validation. The authors concluded that although performance was not uniformly strong across all complications, routinely available data showed promise for early risk stratification, with the severe composite outcome offering a clinically meaningful approach to combine individually rare complications to identify women who warrant closer surveillance.
Subject of Research: Risk prediction models for maternal adverse pregnancy outcomes in HELLP syndrome
Article Title: Development and internal validation of prediction models for maternal adverse pregnancy outcomes in HELLP syndrome: a retrospective cohort study in China
Article References: Lu, Y., Huang, X., Lin, X., Zeng, J., & Su, Y. (2026). Development and internal validation of prediction models for maternal adverse pregnancy outcomes in HELLP syndrome: a retrospective cohort study in China. Frontiers in Endocrinology, 17, Article 1915554. https://doi.org/10.3389/fendo.2026.1915554
Image Credits: AI Generated
DOI: 10.3389/fendo.2026.1915554
Keywords: HELLP syndrome, placental abruption, posterior reversible encephalopathy syndrome, postpartum hemorrhage, disseminated intravascular coagulation, severe maternal composite outcome, adverse pregnancy outcomes, risk stratification, ridge logistic regression, elastic-net logistic regression, random forest, nested cross-validation, multiple imputation by chained equations, retrospective cohort study, Fujian Maternity and Child Health Hospital, receiver operating characteristic AUC, precision-recall curve
News Source: Harold Sullivan. (November 17, 2026). Routine Clinical Data Show Risk Stratification Potential in Chinese HELLP Cohort. Scienmag.



