A fitting machine learning prediction model for short-term mortality following percutaneous catheterization...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Predicting mortality after percutaneous coronary intervention (PCI) requires a robust multivariate model.
• Current predictors for post-PCI mortality are not definitively established.
• A nationwide database was utilized to develop and evaluate mortality prediction models.

Purpose of the Study:

• To construct and identify the most suitable multivariate prediction model for mortality following PCI.
• To compare the performance of various machine learning models in predicting post-PCI mortality.
• To validate the chosen model's effectiveness using a dedicated test dataset.

Main Methods:

• Data from 3,421 acute myocardial infarction (AMI) patients undergoing PCI (2004-2013) from Taiwan's National Health Insurance Research Database (NHIRD).
• Development of multivariate prediction models using 22 input features and 2 mortality output features.
• Implementation and comparison of Artificial Neural Network (ANN), Decision Tree (DT), Linear Discriminant Analysis (LDA), Logistic Regression (LR), Naïve Bayes (NB), and Support Vector Machine (SVM) models.

Main Results:

• The Decision Tree (DT) model demonstrated the highest suitability based on performance and applicability.
• DT model achieved an Area Under the ROC Curve (AUC) of 0.895.
• The DT model yielded high performance metrics: F1 score of 0.969, precision of 0.971, and recall of 0.974.

Conclusions:

• The Decision Tree (DT) model effectively predicts 30-day mortality in patients with AMI undergoing PCI.
• The DT model, built using NHIRD data, shows significant clinical utility.
• This study identifies DT as a valuable tool for risk stratification in post-PCI patients.