Machine learning prediction and interpretability analysis of high-risk chest pain: a study from the MIMIC-IV database

Area of Science:

• Cardiology
• Emergency Medicine
• Machine Learning

Background:

• High-risk chest pain is a critical emergency department presentation.
• Accurate diagnosis of chest pain is vital for patient survival.
• Life-threatening cardiopulmonary conditions often present as chest pain.

Purpose of the Study:

• To develop and optimize a machine learning model for predicting high-risk chest pain.
• To enhance diagnostic accuracy in emergency departments.
• To provide a decision-support tool for emergency physicians.

Main Methods:

• Utilized the MIMIC-IV database (n=14,716 patients).
• Implemented feature engineering with SMOTE and under-sampling to address class imbalance.
• Optimized models using Bayesian hyperparameter tuning and evaluated seven algorithms (Logistic Regression, Random Forest, SVM, XGBoost, LightGBM, TabTransformer, TabNet).

Main Results:

• The LightGBM model achieved superior performance: accuracy=0.95, precision=0.95, recall=0.95, F1-score=0.94.
• SHAP analysis identified maximum troponin and creatine kinase-MB levels as key predictive features.
• The model demonstrated significant predictive capability for high-risk chest pain.

Conclusions:

• The optimized LightGBM model offers clinically significant predictive value for high-risk chest pain.
• This tool can enhance diagnostic accuracy and improve patient outcomes in emergency settings.
• Machine learning models show promise in supporting clinical decision-making for critical presentations.