[Development and evaluation of a machine learning prediction model for large for gestational age]

Area of Science:

• Perinatal Medicine
• Computational Biology
• Public Health

Background:

• Large gestational age (LGA) is associated with adverse perinatal outcomes.
• Accurate prediction of LGA is crucial for effective prenatal care and management.
• Existing predictive models often lack sufficient accuracy.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting LGA.
• To compare the performance of various ML algorithms against traditional logistic regression for LGA prediction.
• To identify the optimal ML model for LGA risk assessment in pregnancy.

Main Methods:

• Utilized data from China's National Free Preconception Health Examination Project (2010-2012).
• Included 104,936 newborns and their mothers, addressing data imbalance with under-sampling.
• Developed and evaluated ten ML algorithms, including CatBoost, and logistic regression for LGA prediction.

Main Results:

• The incidence of LGA was 11.7% in the study cohort.
• Machine learning models significantly outperformed logistic regression (AUC 0.555).
• The CatBoost model achieved the highest predictive accuracy with an AUC of 0.932.

Conclusions:

• Machine learning algorithms, particularly CatBoost, offer superior performance in predicting LGA compared to logistic regression.
• The CatBoost model shows potential for enhancing prenatal risk assessment for LGA.
• Further investigation into CatBoost for clinical application in predicting LGA is warranted.