Classification tree for risk assessment in patients suffering from congestive heart failure via long-term heart rate...

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Congestive heart failure (CHF) risk stratification is crucial for patient management.
• Heart rate variability (HRV) is a recognized indicator of autonomic nervous system function and a potential marker for CHF severity.
• Existing risk assessment methods can be improved with objective, automated tools.

Purpose of the Study:

• To develop and validate an automated classifier for risk assessment in CHF patients.
• To utilize standard long-term heart rate variability (HRV) measures for classifying patients into lower and higher risk categories.
• To correlate classifier performance with New York Heart Association (NYHA) functional classification.

Main Methods:

• Retrospective analysis of two public Holter databases.
• Inclusion criteria: Patients with CHF (NYHA I-IV) and normal-to-normal (NN)/RR interval ratio > 80% for signal quality.
• Classification and Regression Trees (CART) algorithm employed for classifier development.

Main Results:

• Analysis included 30 higher-risk (NYHA III-IV) and 11 lower-risk (NYHA I-II) patients.
• The CART classifier achieved 93.3% sensitivity and 63.6% specificity in identifying higher-risk CHF patients.
• The derived classification rules were interpretable and aligned with existing literature on HRV and CHF.

Conclusions:

• Automated risk assessment using HRV is feasible for CHF patients.
• Depressed HRV, as identified by the classifier, is a valuable tool for CHF risk stratification.
• The developed CART-based classifier offers a promising approach for objective CHF risk assessment.