Enhancing Clinical Decision Making by Predicting Readmission Risk in Patients With Heart Failure Using Machine...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Heart failure patients face high readmission rates, straining healthcare systems.
• Existing predictive models lack effectiveness for the Chinese population.
• Accurate prediction is vital for clinical decision-making and patient care optimization.

Purpose of the Study:

• To develop a predictive model for heart failure readmission.
• To assess the likelihood of rehospitalization in heart failure patients.

Main Methods:

• Analyzed data from 1948 heart failure patients (Sichuan Province, 2016-2019).
• Identified 29 relevant variables using 3 selection strategies.
• Constructed 6 predictive models, including logistic regression, SVM, GBM, XGBoost, MLP, and GCN.

Main Results:

• The Graph Convolutional Network (GCN) model achieved the highest prediction accuracy.
• GCN model performance: AUC 0.831, accuracy 75%, sensitivity 52.12%, specificity 90.25%.

Conclusions:

• The developed GCN model effectively predicts heart failure readmission risk.
• This model serves as a valuable clinical decision-making reference.
• Improved prediction can optimize patient management and reduce healthcare costs.