A machine learning model based on clinical, metabolic, and inflammatory indicators for predicting recurrent...

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Biostatistics

Background:

• Recurrent spontaneous abortion (RSA) impacts 1-5% of couples, often with unknown causes.
• Identifying predictive markers for RSA is crucial for effective management.
• Subclinical inflammation and metabolic factors are suspected contributors to RSA.

Purpose of the Study:

• To develop and validate an interpretable machine learning model for predicting RSA risk.
• To identify key clinical, metabolic, and inflammatory predictors of RSA.
• To enable individualized risk assessment for RSA using routine data.

Main Methods:

• Retrospective analysis of 523 women (285 RSA, 238 controls).
• Comparison of five machine learning algorithms, focusing on random forest.
• Utilized SHAP analysis to determine feature importance.

Main Results:

• The random forest model achieved high predictive performance (AUC = 0.85).
• Key predictors identified include endometrial thickness, neutrophil percentage, and HOMA-β.
• The model effectively uses readily available clinical data for risk estimation.

Conclusions:

• Machine learning offers a promising approach for predicting RSA risk.
• Metabolic dysregulation and subclinical inflammation are significant factors in RSA pathogenesis.
• This interpretable model facilitates personalized risk assessment and potential early intervention strategies.