An interpretable machine learning model for predicting adverse perinatal outcomes in pregnant women with preterm premature rupture of membranes
Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background To establish and validate a machine learning (ML) model for predicting adverse perinatal outcomes (APOs) in pregnant women with preterm premature rupture of membranes (PPROM). Methods From March 2021 to December 2024, a total of 528 pregnant women with PPROM were retrospectively recruited from two centres. The primary endpoints were APOs, including acute respiratory distress syndrome, sepsis, grade III–IV intraventricular haemorrhage, necrotising enterocolitis and neonatal mortality. Five ML methods were used to construct models to predict the APOs. Model performance was evaluated using the area under the curve (AUC). Restricted cubic spline (RCS) analyses were performed to evaluate the association between the interval from PPROM to delivery and the APOs. SHapley Additive exPlanations (SHAP) analysis was performed to interpret the best-performing model. Results Among the five ML models, the multilayer perceptron (MLP) model exhibited robust performance for predicting survival in both the test cohort (AUC = 0.802) and the validation cohort (AUC = 0.847). SHAP analysis indicated that the five most important variables are gestational age at PPROM, Apgar score at 5 min, PT, interval from PPROM to delivery and Apgar score at 1 min. In the subgroup analysis of gestational age at PPROM at 34–36 + 6 weeks, the RCS revealed that a nonlinear relationship was found between the interval from PPROM to delivery and the APOs, with a threshold of approximately 26 hours. Below the threshold value, the effect size was 1.07 (95% CI, 1.01–1.14; P = 0.034). Conclusions We developed and validated an explainable ML model for predicting APOs among pregnant women with PPROM. For pregnant women who experience PPROM at 34–36 + 6 weeks of gestation, delivery should be pursued within the first 26 hours.
Source: Research Square
Authors: Hui Chen, Qiu Mo, Qiu-Xia Feng, Jun-Ping Zhang, Qing-Lan Lin, Yan-Ping Yu
Institution: Shantou University Medical College
Language: English
Subfield: Obstetrics and Gynecology
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7214024075