Predicting Risk of 30-Day Readmissions Using Two Emerging Machine Learning Methods.

Area of Science:

• Cardiology
• Health Informatics
• Machine Learning

Background:

• Heart Failure (HF) is a leading cause of expensive hospitalizations and frequent 30-day readmissions.
• Effective risk stratification at discharge could enable targeted interventions to reduce readmission rates.

Purpose of the Study:

• To compare the predictive performance of two machine learning methods for HF readmission risk.
• To identify optimal machine learning approaches for HF patient risk stratification.

Main Methods:

• Utilized electronic health records data from 1778 unique Heart Failure patients across 31 US hospitals.
• Employed 56 predictors for risk assessment.
• Compared boosted trees and spike-and-slab regression models for predictive accuracy.

Main Results:

• Boosted trees achieved a higher predictive accuracy (AUC: 0.719) compared to spike-and-slab regression (AUC: 0.621).
• The study demonstrates the potential of machine learning in predicting HF readmissions.

Conclusions:

• Boosted trees represent a more effective machine learning approach for predicting Heart Failure readmission risk.
• Accurate risk stratification using machine learning can guide post-discharge interventions, potentially lowering healthcare costs.