Related Experiment Video
Updated: Apr 21, 2026
06:55
Inverse Probability of Treatment Weighting Propensity Score using the Military Health System Data Repository and National Death Index
Published on: January 8, 2020
14.3K
Statistical Relational Learning to Predict Primary Myocardial Infarction from Electronic Health Records
Jeremy C Weiss1, David Page1, Peggy L Peissig2
1University of Wisconsin-Madison 1300 University Ave, Madison,WI {jcweiss,page}@biostat.wisc.edu.
Proceedings of the ... Innovative Applications of Artificial Intelligence Conference. Innovative Applications of Artificial Intelligence Conference
|November 1, 2014
Summary
Statistical relational learning algorithms applied to electronic health records (EHRs) show promise for predicting myocardial infarction. One algorithm, relational functional gradient boosting, outperformed traditional methods, especially for identifying high-risk patients.
Area of Science:
- * Computational epidemiology
- * Machine learning in healthcare
- * Health informatics
Background:
- * Electronic health records (EHRs) represent a valuable, yet underutilized, relational data source.
- * Predicting clinical outcomes like myocardial infarction is crucial for preventative medicine.
- * Existing predictive models often lack the sophistication to leverage complex relational data within EHRs.
Purpose of the Study:
- * To evaluate the effectiveness of statistical relational learning (SRL) algorithms for predicting primary myocardial infarction using EHR data.
- * To compare the performance of SRL methods against traditional propositional learning approaches.
- * To demonstrate the potential of SRL to enhance current epidemiological practices.
Main Methods:
- * Application of two SRL algorithms: relational functional gradient boosting and another SRL algorithm.
- * Comparison of SRL algorithms against their propositional counterparts.
- * Analysis focused on predictive performance, particularly in the high recall region.
Main Results:
- * Both SRL algorithms demonstrated superior predictive performance compared to their propositional analogs.
- * Relational functional gradient boosting showed particular strength in the high recall region, crucial for clinical relevance.
- * The study validates the utility of SRL for analyzing complex EHR data.
Conclusions:
- * SRL algorithms offer a significant advancement over propositional methods for clinical outcome prediction using EHRs.
- * Relational functional gradient boosting is a promising technique for improving myocardial infarction prediction.
- * These findings suggest a pathway for integrating advanced machine learning into epidemiological surveillance and clinical practice.
