Machine Learning Compared With Conventional Statistical Models for Predicting Myocardial Infarction Readmission and
Sung Min Cho1, Peter C Austin2, Heather J Ross3
1Ted Rogers Centre for Heart Research, Toronto, Ontario, Canada; University of Toronto, Toronto, Ontario, Canada.
Machine learning (ML) models show slightly better prediction for myocardial infarction (MI) patient outcomes than conventional statistical models (CSM). However, many studies lacked validation and had bias concerns.
Area of Science:
- Cardiology
- Medical Informatics
- Biostatistics
Background:
- Machine learning (ML) and conventional statistical modeling (CSM) are used for predicting patient outcomes after myocardial infarction (MI).
- Systematic comparisons of ML and CSM for MI prognosis are lacking.
- This review addresses the need for a systematic comparison of ML and CSM in MI prognostication.
Purpose of the Study:
- To systematically review and compare the performance of ML and CSM in predicting mortality and readmission in patients with MI.
- To identify common ML and CSM techniques used in MI prognosis.
- To assess the quality and validity of studies comparing ML and CSM for MI prognosis.
Main Methods:
- A systematic literature review was conducted following PRISMA guidelines.
- Databases searched included Medline, Embase, Web of Science, and others.
- Studies published from January 2000 to March 2020 comparing ML and CSM for MI prognosis were included.
Main Results:
- 24 studies involving 374,365 patients were included.
- Common ML methods: artificial neural networks, random forests, decision trees, support vector machines, Bayesian techniques.
- Common CSM methods: logistic regression, risk scores, Cox regression.
- ML showed higher C-indexes for mortality prediction in 13/19 studies and for readmission in 1/1 study.
- Most comparisons (90%) showed small absolute differences (<0.05) in C-indexes between ML and CSM.
- Majority of studies had identifiable bias, and only 2 were externally validated.
Conclusions:
- ML algorithms demonstrated a tendency for higher predictive accuracy (C-indexes) compared to CSM for MI-related mortality and readmission.
- However, the included studies often suffered from threats to internal validity and lacked external validation.
- Further high-quality, validated comparative studies adhering to clinical quality standards are necessary for robust MI prognosis research.
More Related Videos
04:09Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
Published on: October 10, 2018
08:38Murine Myocardial Infarction Model using Permanent Ligation of Left Anterior Descending Coronary Artery
Published on: August 16, 2019
