AI-Derived Score to Predict Effort Intolerance and Adverse Outcome across the Heart Failure Spectrum

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• Effort intolerance is a key symptom in heart failure (HF).
• Accurate prediction of reduced peak oxygen uptake (VO₂) is crucial for HF management.
• Current methods for assessing VO₂ may be limited in clinical practice.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI)-driven model for predicting effort intolerance (peak oxygen uptake [VO₂] <16 mL/kg/min).
• To identify patients at risk or with established heart failure (HF).
• To assess the prognostic value of the developed AI score for adverse cardiovascular events.

Main Methods:

• A derivation cohort (1,333 participants) and an external validation cohort (1,101 participants) were used.
• Participants included individuals with heart failure with reduced ejection fraction (HFrEF), heart failure with preserved ejection fraction (HFpEF), and cardiovascular risk factors.
• An AI model was developed using clinical, laboratory, echocardiographic, and cardiopulmonary exercise testing data.

Main Results:

• A neural network model incorporating age, sex, BMI, hemoglobin, LV S', sPAP, and beta-blocker therapy showed high discrimination (AUC 0.86 in derivation, 0.76 in validation).
• A simplified AI-VO₂ score (BMI, hemoglobin, LV S', sPAP) demonstrated good performance (AUC 0.79) and independently predicted HF hospitalization or death.
• External validation confirmed the score's predictive (AUC 0.73) and prognostic capabilities, with significant risk stratification.

Conclusions:

• The AI-VO₂ score is a clinically interpretable and externally validated tool.
• It effectively identifies patients with effort intolerance across the HF spectrum.
• The score supports personalized risk stratification and management of patients with heart failure.