Predicting need for heart failure advanced therapies using an interpretable tropical geometry-based fuzzy neural
Yufeng Zhang1, Keith D Aaronson2, Jonathan Gryak3
1Department of Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, Michigan, United States of America.
Plos One
|November 28, 2023
Summary
Predicting advanced heart failure therapies is crucial. This study developed an interpretable machine learning system using electronic health records to identify patients needing advanced heart failure therapies at their next hospitalization.
Area of Science:
- Cardiology
- Machine Learning
- Health Informatics
Background:
- Timely referral for advanced therapies (heart transplantation, left ventricular assist device) is critical for optimal heart failure patient outcomes.
- Electronic health records (EHRs) offer a valuable resource for developing predictive clinical decision-making systems.
- Predicting the need for advanced therapies based on a single hospitalization can improve patient management.
Purpose of the Study:
- To develop an interpretable clinical decision-making system using EHR data from a single hospitalization.
- To predict the need for advanced therapies (heart transplantation, left ventricular assist device) at a subsequent hospitalization for heart failure patients.
- To create a transparent and accessible system for identifying high-risk patients.
Main Methods:
- Trained an interpretable machine learning model using fuzzy logic and tropical geometry on EHR data from 300 heart failure patients (2013-2021).
- Utilized data from patients with left ventricular ejection fraction ≤ 35% and at least two heart failure hospitalizations within one year.
- Evaluated model performance using area under the receiver operating curve (AUC), area under the precision-recall curve (AUPRC), and F1 score, initializing with clinical knowledge.
Main Results:
- The model achieved a mean AUC of 0.747 (0.080), AUPRC of 0.642 (0.080), and F1 score of 0.569 (0.067).
- Demonstrated superior predictive performance and interpretability compared to other machine learning methods.
- Identified critical risk factors and generated transparent rule sets to justify predictions for advanced heart failure therapies.
Conclusions:
- Successfully predicted the need for advanced heart failure therapies using transparent and accessible clinical rules.
- The developed system can aid in identifying patients requiring advanced interventions.
- Further prospective research is needed to validate the identified risk factors.
Related Concept Videos
01:17Pathophysiology of Heart Failure
1.6K
Heart failure (HF) is a progressive syndrome involving ventricles that leads to inadequate cardiac output. It can be classified based on location and output or ejection fraction. Ejection fraction (EF) is an essential measurement in the diagnosis and surveillance of HF. Reduced EF corresponds to systolic heart failure (HFrEF). However, HF with preserved ejection fraction (HFpEF) is becoming increasingly prevalent. Also known as diastolic HF, this form of HF is related to aging. The...
1.6K
01:26Heart Failure Drugs: Inhibitors of Renin-Angiotensin System
436
The activation of the sympathetic nervous system and the renin-angiotensin-aldosterone system (RAAS) contributes to cardiac remodeling, and inhibiting the RAAS is a pharmacological target in heart failure management. As a result, neurohumoral modulation is a crucial treatment principle for managing heart failure. This approach involves using medications like ACE inhibitors (ACEIs), angiotensin receptor blockers (ARBs), β-blockers, mineralocorticoid receptor antagonists (MRAs), and neutral...
436