Machine Learning-Based Prediction of Digoxin Toxicity in Heart Failure: A Multicenter Retrospective Study

Area of Science:

• Pharmacology
• Medical Informatics
• Machine Learning

Background:

• Digoxin toxicity is linked to worsening heart failure (HF).
• Predicting adverse drug reactions is crucial for patient safety.
• Decision tree (DT) analysis offers a flowchart-like approach for risk prediction.

Purpose of the Study:

• To develop a predictive flowchart for digoxin toxicity using DT analysis.
• To aid medical staff in identifying patients at high risk of digoxin toxicity.

Main Methods:

• Multicenter retrospective study of 333 adult HF patients on oral digoxin.
• Chi-squared automatic interaction detection algorithm for DT model construction.
• Plasma digoxin concentration (≥0.9 ng/mL) as the dependent variable.

Main Results:

• DT analysis identified patients with creatinine clearance <32 mL/min, daily digoxin dose ≥1.6 µg/kg, and LVEF ≥50% as high-risk (91.8% incidence).
• Multivariate logistic regression confirmed creatinine clearance <32 mL/min and daily digoxin dose ≥1.6 µg/kg as independent risk factors.
• The DT model achieved 88.2% accuracy with a 46.2% misclassification rate.

Conclusions:

• A straightforward DT-based flowchart can assist medical staff in predicting digoxin toxicity.
• The model requires further validation but shows potential for guiding initial digoxin dosing in HF patients.