A dynamic risk score to identify increased risk for heart failure decompensation

Area of Science:

• Biomedical Engineering
• Cardiology
• Artificial Intelligence in Medicine

Background:

• Heart failure hospitalization (HFH) poses a significant clinical challenge.
• Current diagnostic methods for HFH risk may lack comprehensive integration of real-time patient data.
• Implantable devices offer continuous monitoring of physiological parameters relevant to heart failure.

Purpose of the Study:

• To develop and validate a Bayesian belief network (BBN) for integrating diverse diagnostic data to predict HFH risk.
• To enhance the early identification of patients at high risk for HF hospitalization.
• To enable proactive and personalized patient management strategies.

Main Methods:

• Utilized daily data from implantable devices, including intrathoracic impedance, atrial fibrillation (AF) burden, heart rates, and activity.
• Extracted features capturing out-of-range values and temporal trends at weekly and monthly scales.
• Developed a BBN to combine these features into a composite risk score for HFH.

Main Results:

• The BBN-derived risk score significantly improved the ability to identify patients at risk for HFH compared to individual parameters.
• Patients with very high risk scores were 15 times more likely to experience HFH within 30 days than those with low scores.
• The integrated risk score facilitates 'management by exception', prioritizing high-risk patients.

Conclusions:

• A BBN framework effectively combines multi-parameter diagnostics from implantable devices for improved HFH risk prediction.
• This approach offers a more sensitive and specific method for identifying patients requiring closer clinical attention.
• The developed risk score supports timely interventions, potentially reducing HFH events.