Machine Learning for the Prediction of Surgical Morbidity in Placenta Accreta Spectrum

Area of Science:

• Perinatal medicine
• Machine learning in healthcare
• Surgical risk assessment

Background:

• Placenta accreta spectrum (PAS) poses significant surgical risks.
• Predicting operative morbidity in PAS is crucial for patient management.
• Existing prediction models for PAS complications are limited.

Purpose of the Study:

• To develop a machine learning (ML) model for predicting surgical morbidity in placenta accreta spectrum (PAS) cases.
• To identify key variables contributing to surgical morbidity in PAS.
• To establish a foundation for risk stratification and optimized surgical planning in PAS.

Main Methods:

• Multicenter analysis of 401 PAS cases (2013-2022).
• Development of gradient boosted tree classifier models using 213 variables.
• Evaluation of model performance using AUC, PPV, NPV, and F1 score.

Main Results:

• The ML model achieved an AUC of 0.79 for predicting surgical morbidity.
• Key predictors included hysterectomy completion, prepregnancy BMI, socioeconomic status, and prenatal visit frequency.
• 77% of PAS cases experienced at least one surgical morbidity event.

Conclusions:

• ML models effectively predict PAS-related surgical morbidity.
• Social and obstetrical factors are significant predictors of surgical risk.
• ML can enhance risk stratification for improved surgical planning in PAS cases.