Augmented Intelligence to Identify Patients With Advanced Heart Failure in an Integrated Health System.

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Health Informatics

Background:

• Timely specialist referral is crucial for advanced heart failure (HF) patients.
• The transition to stage D HF is often missed in routine care, leading to delayed referrals and increased mortality.

Purpose of the Study:

• To develop an augmented intelligence workflow using machine learning to identify patients with stage D HF.
• To streamline the referral process for advanced heart failure patients.

Main Methods:

• An ensemble machine learning model was created to predict stage C or D HF using data from a HF registry (2007-2020).
• The model's predictions were integrated into the electronic health record for clinical use.

Main Results:

• The model demonstrated a positive predictive value of 60% and sensitivity of 25% for stage D HF in a test set.
• Prospective implementation showed 50.3% agreement between the model and clinical coordinators for stage D HF predictions.
• The workflow led to 24 patients scheduled for HF clinic evaluation, 4 starting advanced therapy evaluations, and 1 receiving a left ventricular assist device.

Conclusions:

• An augmented intelligence workflow was successfully integrated into clinical operations for advanced HF patient identification.
• Multidisciplinary collaboration and dedicated resources are essential for developing and maintaining such systems.