The Predictive Value of the Pan-Immune-Inflammation Value for Atrial Fibrillation Risk in Patients with Coronary...

Area of Science:

• Cardiology
• Inflammation Research
• Medical Informatics

Background:

• Atrial fibrillation (AF) is a common arrhythmia in patients with coronary heart disease (CHD).
• Inflammatory response is a key factor in AF pathogenesis.
• The predictive value of pan-immune-inflammation value (PIV) for AF in CHD patients requires further investigation.

Purpose of the Study:

• To investigate the predictive value of PIV for AF in patients with CHD.
• To compare the performance of machine learning models (XGBoost and MLP) in predicting AF risk.
• To identify key predictors of AF in this patient cohort.

Main Methods:

• A multicenter retrospective study involving patients diagnosed with CHD.
• Collected clinical characteristics and laboratory data, including PIV.
• Utilized logistic regression for feature selection and developed XGBoost and MLP models for AF prediction.
• Evaluated model performance using AUC and calibration, with SHAP values for interpretability.

Main Results:

• Patients with AF exhibited significantly higher PIV, age, AST, WBC, and TBIL levels compared to the non-AF group.
• PIV, age, and diabetes were identified as independent predictors of AF.
• The XGBoost model demonstrated superior predictive performance (AUC = 0.73 in testing) over the MLP model (AUC = 0.69).
• SHAP analysis highlighted PIV as the most influential predictor of AF risk.

Conclusions:

• PIV is a valuable predictor of AF in patients with CHD.
• The XGBoost model offers a robust tool for individualized AF risk assessment in this population.
• Machine learning approaches can enhance the prediction of AF in CHD patients.