Related Experiment Video
Updated: Jul 3, 2026
An Experimental Paradigm for the Prediction of Post-Operative Pain (PPOP)
Published on: January 27, 2010
Machine learning for prediction of histologic chorioamnionitis (stage ≥II) in parturients receiving labor analgesia:
Chunping Li1, Bo Liu2, Liang Ling3
1Department of Anesthesiology, Sichuan Jinxin Xinan Women and Children's Hospital, Chengdu, Sichuan, China.
Background:
Histological chorioamnionitis (HCA) is a serious pregnancy complication, but early diagnosis is challenging, especially in parturients receiving labor analgesia, where identification is even more difficult. Therefore, we developed and validated a machine learning model for early prediction of HCA (stage ≥II) intended for use at the time of epidural-related maternal fever (ERMF) onset, before delivery and confirmatory histopathology.
Methods:
This study utilized a multicenter retrospective cohort dataset, including parturients receiving labor analgesia and completed placental pathological examination. Candidate features were extracted from electronic health records (EHR), followed by data preprocessing procedures such as addressing multicollinearity, Z-score standardization, and minority class weighting. Subsequently, three machine learning models-logistic regression (LR), random forest (RF), and extreme gradient boosting (XGBoost)-were developed and compared, with their performance evaluated through internal and independent external validation. Finally, the SHapley Additive exPlanations (SHAP) method was used to interpret the best-performing model, aiming to clarify the predictive contribution of each feature.
Results:
A total of 2,715 parturients were included in this study, of which 676 cases (24.9%) were diagnosed with HCA (stage ≥II). After feature selection, the model retained 6 key features. The RF model exhibited the best performance, achieving an Area Under the Curve(AUC) of 0.945 in the internal validation set and an AUC of 0.849 in the independent external validation set, demonstrating balanced sensitivity (0.957) and specificity (0.867). SHAP analysis indicated that body mass index (BMI) was the most important predictive factor.
Conclusions:
The RF model performed the best in predicting the risk of HCA (stage ≥II) in parturients receiving labor analgesia. SHAP analysis further revealed that BMI was the most important predictive factor in the model.