Machine learning model for predicting improvement in left ventricular systolic function in patients with heart...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Approximately one-third of heart failure with reduced ejection fraction (HFrEF) patients experience left ventricular ejection fraction (LVEF) recovery with medical management.
• Predicting LVEF recovery is crucial for optimizing treatment strategies in HFrEF.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting LVEF improvement in HFrEF patients.
• To compare the predictive performance of the ML model against a traditional logistic regression (LR) model.

Main Methods:

• A cohort of 3124 HFrEF patients with at least two echocardiograms was divided into development and external testing sets.
• Both ML and LR models were trained using 49 features and internally validated using 5-fold cross-validation.
• Model performance was assessed using area under the curve (AUC) and Brier score for prediction and calibration.

Main Results:

• LVEF recovery (≥10% increase) was observed in 36.0% of the development cohort and 39.8% of the external testing cohort.
• The ML model showed a statistically significant advantage over LR in the development cohort (AUC 0.719 vs 0.700, p=0.045).
• No significant difference in predictive performance was found between ML and LR models in the external testing cohort (AUC 0.702 vs 0.696, p=0.498).

Conclusions:

• A machine learning model utilizing readily available clinical and echocardiographic data did not demonstrate superior performance to logistic regression in predicting LVEF improvement upon external validation.
• Factors such as lower baseline LVEF, smaller left ventricular dimensions, younger age, and non-ischemic etiology were key predictors of LVEF improvement.
• Larger studies incorporating additional variables or alternative predictive approaches may be necessary to enhance the accuracy of LVEF recovery prediction.