Overview
The goal of this clinical trial is to determine if a machine learning/artificial intelligence (AI)-based electrocardiogram (ECG) algorithm (rECHOmmend and ECG-AF) can identify undiagnosed cardiovascular disease in patients. It will also examine the safety and effectiveness of using this AI-based tool in a clinical setting. The main questions it aims to answer are:
- Can the AI-based ECG algorithm improve the detection of atrial fibrillation and structural heart disease?
- How does the use of this algorithm affect clinical decision-making and patient outcomes?
Researchers will compare the outcomes of healthcare providers who receive the AI-based ECG results to those who do not. Participants (healthcare providers) will:
Be randomized into two groups: one that receives AI-based ECG results and one that does not.
In the intervention group, receive an assessment of their patient's risk of atrial fibrillation or structural heart disease with each ordered ECG.
Decide whether to perform further clinical evaluation based on the AI-generated risk assessment as part of routine clinical care.
Description
There is a large burden of undiagnosed, treatable cardiovascular disease (CVD), encompassing various heart conditions such as arrhythmias (e.g., atrial fibrillation) and structural heart diseases (e.g., valvular disease). Early detection and accurate diagnosis can significantly improve patient outcomes by enabling timely, guideline-based interventions or therapies.
The goal of this study is to leverage machine learning approaches to enhance the detection and diagnosis of CVD. By identifying patients at risk of undiagnosed CVD and referring them for further clinical evaluation, the study aims to improve health outcomes.
Study Overview:
The NOTABLE study will compare the rates of new disease diagnoses, therapeutic interventions, and cardiovascular outcomes between two groups of patients managed by clinicians at Northwestern Medicine:
Patients whose clinicians use ECG predictive models. Patients whose clinicians do not use ECG predictive models.
Intervention Details:
This study utilizes Tempus AI algorithms (rECHOmmend and ECG-AF) to analyze 12-lead ECGs. Clinicians randomized to the intervention group will receive a "Risk-Based Assessment for Cardiac Dysfunction" when ordering a 12-lead ECG within EPIC. If a high-risk result is identified, clinicians receive an EHR inbox message suggesting a follow-up diagnostic test, such as echocardiography and/or ambulatory ECG monitoring.
Outcome Tracking:
Weekly summaries will inform clinicians in the intervention group of high-risk results identified by the AI algorithm. Clinicians in the usual care group will not receive any communication from the study investigators regarding AI predictions.
Eligibility
Inclusion Criteria:
- Atrial fibrillation algorithm
- Age 65 or over
- ECG obtained as part of routine clinical care
- Structural heart disease algorithm
- Age 40 or over
- ECG obtained as part of routine clinical care
Exclusion Criteria:
- Atrial fibrillation algorithm
- No history of AF
- No permanent pacemaker (PPM) or implantable cardioverter defibrillator (ICD)
- No recent cardiac surgery (within the preceding 30 days)
- Structural heart disease algorithm
- No history of SHD
- No echocardiogram within the past 1 year
