Prediction and Analysis of Heart Failure Decompensation Events Based on Telemonitored Data and Artificial...

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Informatics

Background:

• Cardiovascular diseases are a leading global cause of death.
• Heart failure (HF) significantly impacts patient quality of life due to worsening conditions.
• Predicting cardiac decompensation events (CDEs) is crucial for timely intervention in chronic HF patients.

Purpose of the Study:

• To develop an artificial intelligence (AI) model for accurate and timely prediction of CDEs in chronic HF patients.
• To identify key physiological and self-reported variables that predict CDEs.
• To enable proactive clinical interventions before decompensation occurs.

Main Methods:

• Monitored vital variables (n=488) of chronic HF patients from 2014-2022.
• Trained supervised classification models using patient monitoring data to predict CDEs.
• Applied feature extraction to identify significant predictive variables, using clinical annotations as the gold standard.

Main Results:

• The XGBoost classifier achieved an AUC of 0.72 (cross-validation) and 0.69 (testing set).
• Significant predictors for CDEs include weight gain, oxygen saturation, and heart rate.
• Patient-reported wellbeing, orthopnoea, and ankle symptoms are also strong predictors.

Conclusions:

• AI models can effectively predict CDEs in chronic heart failure patients.
• Early identification of high-risk patients is achievable through AI-driven analysis of monitoring data.
• This approach supports timely clinical interventions, potentially improving patient outcomes.