Machine learning approaches to predict 30-day mortality following percutaneous coronary intervention in an Australian population

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Background

PCI is an effective treatment for coronary artery disease. Pre-procedural 30-day mortality post-PCI risk prediction aids in clinical decision-making and benchmarking hospital performance. This study aimed to identify pre-procedural factors to predict the risk of 30-day mortality following Percutaneous Coronary Intervention (PCI) using machine learning (ML) approaches.

Methods

The study analysed 93,055 consecutive PCI procedures from the Victorian Cardiac Outcomes Registry (VCOR) in Australia to develop a pre-procedural 30-day mortality prediction model. Five ML approaches—Adaptive Booster (AdB), Decision Tree (DT), Gradient Booster (GB), Random Forest (RF), and Extreme Gradient Booster (XGB) were employed, utilizing Logistic Regression (LR) for comparison. Model performance was evaluated using k-fold cross-validation, with metrics including sensitivity, specificity, accuracy, ROC curve, Brier score, and calibration curve.

Results

The study showed that the RF model outperformed other ML models in predicting 30-day mortality, achieving accuracy of 98.4% and a ROC of 94.3%. Utilizing the SHapley Additive exPlanations method, the RF model identified cardiogenic shock, ejection fraction, acute coronary syndrome, estimated GFR, cardiac arrest, age, mechanical ventricular support, complex lesion, lesion location, BMI, sex, and diabetes as the variables that were associated with 30-day mortality post-PCI. In comparison, the traditional LR model exhibited an accuracy of 98.2% and a ROC of 92.9%.

Conclusion

A 30-day mortality post-PCI risk prediction model was developed with high accuracy using a ML method. It’s essential to underscore the need for further validation with external data to ensure the applicability of the model to other populations.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • A risk-adjustment model for an Australian PCI patient population was previously developed to predict 30-day mortality using traditional regression model.

  • Medical knowledge, patient characteristics, and clinical practices evolve over time, requiring frequent model updates to reflect new evidence, guidelines, and interventions

WHAT THIS STUDY ADDS

  • A machine learning (ML)-based preprocedural risk prediction model for 30-day mortality following percutaneous coronary intervention (PCI) was developed.

  • The ML-based model was compared with the traditional regression model. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • Risk prediction models aid clinical decision-making, enhance patient counselling, improve care quality, inform healthcare policies, and advance research.

Article activity feed