Development of a predictive model for histological chorioamnionitis in patients with preterm premature rupture of membranes: a single-center retrospective cohort study

Objective The aim of this study is to explore the risk factors for histological chorioamnionitis (HC) and to develop a preliminary clinical prediction model among women with preterm premature rupture of membranes (PPROM) during inpatient expectant management. Methods This single-center retrospective cohort study consecutively enrolled women with singleton PPROM at 28 + 0 ∼ 36 + 6 weeks of gestation who delivered at Hangzhou Women’s Hospital between December 2022 and October 2025. Clinical and laboratory data recorded from admission to delivery were extracted. Univariate analysis was followed by multivariable logistic regression. Candidate predictors were selected on clinical plausibility and statistical significance. A binary logistic regression model was fitted, and apparent discriminative performance was summarized with a receiver operating characteristic (ROC) curve. Results A total of 306 women were included (HC group, n = 152; control group, n = 154). In the multivariable model, vaginal examinations (VE), antibiotic beginning time (ABT), reproductive tract pathogens (RTP), white blood cell count (WBC), and C-reactive protein (CRP) remained independently associated with HC. The fitted model was z = 1.874 × ln(VE) + 1.051 × ln(ABT) + 2.378 × ln(WBC) + 0.821 × ln(CRP) + 0.846 × RTP (0 or 1) −11.054. The apparent area under the ROC curve (AUC) was 0.901 (95% confidence interval (CI) [0.864–0.934]); the optimism-corrected AUC after bootstrap internal validation was 0.894, and calibration was adequate (Hosmer–Lemeshow P = 0.409; Brier score 0.128). Conclusions VE, ABT, RTP, WBC, and CRP were identified as independent risk factors for HC in patients with PPROM. The resulting model is a preliminary risk-stratification tool intended for use during inpatient management. It requires external validation in independent populations and an admission-only predictor version before clinical implementation. It is not intended for prediction at the time of admission.