Predicting pregnancy-related pelvic girdle pain using machine learning.

Area of Science:

• Medical research
• Data science in healthcare
• Obstetrics and Gynecology

Background:

• Pregnancy-related pelvic girdle pain (PPGP) significantly impacts pregnant women's quality of life.
• Limited understanding exists regarding the predictability of PPGP diagnosis.
• Identifying risk factors is crucial for timely intervention.

Purpose of the Study:

• To compare the predictive performance of machine learning (ML) models against traditional methods for PPGP.
• To evaluate the efficacy of various ML algorithms in predicting PPGP diagnosis.

Main Methods:

• Reanalysis of data from 780 pregnant women at a tertiary hospital.
• Application of ML algorithms: Logistic Regression (LR), Random Forest, XGBoost, K-Nearest Neighbors.
• Optimization using feature selection and cross-validation; AUROC as the primary metric.

Main Results:

• ML models, especially XGBoost and LR, showed high predictive accuracy (AUROC = 0.70).
• Significant predictors included prior LBP/PGP history, family history, gestational age, and prolonged standing.
• Previous LBP/PGP history was the most influential predictive factor.

Conclusions:

• ML models demonstrate significant potential for predicting PPGP, improving risk identification.
• Integration of ML into clinical practice can enhance early detection and preventative strategies.
• This approach may reduce the adverse effects of PPGP on pregnant individuals.