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

Recurrent spontaneous abortion (RSA) affects 1-5 % of couples, with over half of cases remaining unexplained. We developed an interpretable machine learning model using routine clinical, metabolic, and inflammatory indicators to predict RSA risk. A retrospective cohort of 523 women (285 RSA and 238 controls) was analyzed. Five algorithms were compared, and the random forest model showed the best performance (AUC = 0.85). SHAP analysis identified endometrial thickness, neutrophil percentage, and HOMA-β as the most influential predictors. This model enables individualized RSA risk estimation using readily available clinical data and suggests that metabolic dysregulation and subclinical inflammation are key contributors to RSA pathogenesis.