Capitalizing on AI to CapTure Undiagnosed Structural Heart Disease (CACTUS)

Pierre Elias

Status: Enrolling

Conditions: Valve Heart Disease; Cardiovascular Diseases; Heart Diseases; Structural Heart Abnormality

Treatments: Device: EHR Alert

Study type: Interventional

Funder types: Other; Industry; NIH

Identifiers

NCT07843004

AAAU9699

1R01HL177055-01 (U.S. NIH Grant/Contract)

Details and patient eligibility

About

Doctors are testing a new tool called EchoNext to see if it can help find heart problems earlier. EchoNext looks at the heart's electrical test, called an ECG, and uses artificial intelligence (AI) to check for signs of structural heart disease (SHD). SHD includes conditions like weak heart pumping, heart valve problems, or extra thickening of the heart muscle. These problems are common but often go undiagnosed until they cause serious issues like heart failure or stroke.

The main questions this study will answer are:

Can EchoNext alerts help emergency doctors find hidden heart problems sooner?

Does this lead to more follow-up heart tests, like an echocardiogram (heart ultrasound)?

What this means for patients: If a person has an ECG in the Emergency Department, the EchoNext tool may be used to check their heart. If the tool finds something unusual, their doctor may receive an alert and may then recommend further heart tests or follow-up care.

This study will help researchers learn if using EchoNext improves early diagnosis and treatment of heart disease.

Full description

Structural heart disease (SHD) is a major cause of illness and death, especially in older adults. SHD includes conditions such as valvular heart disease (e.g., aortic stenosis, mitral regurgitation), left or right ventricular dysfunction, left ventricular hypertrophy, pulmonary artery hypertension, and pericardial effusions. These conditions are often underdiagnosed, and delayed recognition can lead to complications such as heart failure, stroke, and cardiomyopathy. Early detection and treatment can improve outcomes, but current approaches often miss patients in the early stages of disease.

EchoNext is an artificial intelligence model trained on over 400,000 ECG-echocardiogram pairs. It analyzes the standard 12-lead ECG to predict the presence of structural heart disease, as defined by echocardiographic findings. The model has demonstrated strong diagnostic performance, with an area under the receiver operating characteristic (AUROC) curve of 0.86 across diverse populations.

The purpose of this trial is to evaluate whether deploying EchoNext into the Emergency Department (ED) electronic health record (EHR) as a clinical decision support alert can increase the detection of undiagnosed SHD. The ED setting is ideal because: (1) ECGs are widely obtained for many presenting complaints, (2) abnormal ECG findings related to SHD are often overlooked in the acute setting, and (3) EDs provide access to disadvantaged populations and a unique opportunity to deliver population-level interventions.

In this study, ECGs obtained in the ED will be analyzed in real time by the EchoNext model. If the model predicts moderate or severe SHD, an EHR alert will be displayed to the treating provider (attending physician, fellow, resident, physician assistant, or nurse practitioner). The alert will inform providers of the potential for underlying SHD and recommend consideration of follow-up echocardiography or specialty referral. Outcomes will include rates of new SHD diagnoses confirmed by echocardiography, follow-up testing and referrals, and downstream patient outcomes.

The Community Tele-Paramedicine (CTP) program, an established post-discharge care initiative at NewYork-Presbyterian, will also serve as a follow-up resource for patients identified in the ED. CTP combines home visits by community paramedics, telehealth visits by emergency physicians, and nurse care management to coordinate outpatient testing and follow-up care. Integration with CTP will facilitate timely echocardiography and referral for patients flagged by EchoNext, particularly for high-risk or underserved populations.

This trial will determine whether AI-driven ECG interpretation can improve early detection of SHD and lead to better patient outcomes compared to standard care.

Enrollment: 4,000 estimated patients

Sex: All

Ages: 40+ years old

Volunteers: No Healthy Volunteers

Inclusion criteria

• 40 years of age or older
• Received an ECG upon presentation to the Emergency Department

Exclusion criteria

• Had an echocardiogram within the prior 2 years.
• Has a history of structural heart disease.
• Not able to receive follow-up based on Community Tele-Paramedicine care manager review.
• Quality of life will not improve over the next five years.

Trial design

Primary purpose: Screening

Allocation: Randomized

Interventional model: Crossover Assignment

Masking: None (Open label)

---

4,000 participants in 2 patient groups

Standard Treatment - No EHR Alert

No Intervention group

Description:

Emergency department providers will not be alerted in the electronic health record (EHR) if a patient's electrocardiogram (ECG) ordered in the ED is identified by the EchoNext algorithm as suggesting structural heart disease (SHD) and potentially benefit from an echocardiogram.

Intervention - EHR Alert & Community Tele-paramedicine

Experimental group

Description:

Emergency department providers will be alerted in the electronic health record (EHR) if a patient's electrocardiogram (ECG) ordered in the ED is identified by the EchoNext algorithm as suggesting structural heart disease (SHD) and potentially benefit from an echocardiogram. These patients will also be referred to Community Tele-paramedicine for follow-up for potential echocardiogram.

Treatment:

Device: EHR Alert