The DECISION trial aims to evaluate the efficacy of an artificial intelligence (AI)-powered system, Willem™, for improving the detection of heart failure (HF) in primary care settings by interpreting electrocardiograms (ECGs). The study seeks to answer whether AI-assisted ECG interpretation enhances diagnostic accuracy and clinical outcomes compared to standard ECG evaluation in patients with suspected HF or those at high risk.

This multicenter, pragmatic, randomized clinical trial involves two groups: patients receiving AI-assisted ECG analysis and those undergoing standard ECG evaluation. The study's primary analysis will compare the diagnostic performance of AI-assisted ECG versus standard ECG using sensitivity, specificity, and predictive value metrics. Secondary analyses will evaluate healthcare resource utilization, clinical outcomes, and usability feedback from healthcare providers. Results will inform the potential integration of AI-assisted ECG in routine primary care workflows for earlier HF detection and better resource allocation.

Heart failure (HF) is a prevalent and underdiagnosed condition with high morbidity and mortality. Up to 50% of HF cases remain undetected, often due to subtle or absent symptoms in early stages. Early diagnosis is critical to improving outcomes, reducing hospitalizations, and alleviating healthcare costs. While ECGs are a cornerstone in HF diagnosis, their interpretation in primary care can be challenging, leading to diagnostic delays.

Artificial intelligence (AI) has emerged as a promising tool to support clinicians by enhancing ECG interpretation. In this regard, the DECISION trial evaluates the Willem™ platform, an AI-powered decision-support system, to improve HF detection. Willem™ uses a proprietary database to analyze ECGs, identifying over 80 cardiac patterns with high accuracy.

This study hypothesizes that AI-assisted ECG improves HF detection compared to standard ECG interpretation. Therefore, the main goal of the DECISION trial is to assess the diagnostic performance of AI-assisted ECG in detecting structural and functional cardiac abnormalities indicative of HF.

This multicenter, randomized trial includes primary care centers (PCCs) in Spain and Sweden, randomized into two groups: an intervention group using AI-assisted ECG and a control group using standard ECG. AI outputs will be available for physicians in the intervention group as supplementary information during decision making.

Primary outcomes focus on the accuracy of HF detection confirmed by transthoracic echocardiograms (TTE). Secondary outcomes include healthcare resource utilization, clinical outcomes, and physician satisfaction. The results will inform whether AI can be integrated into primary care workflows to optimize HF diagnosis and management.

Inclusion Criteria:

• Patients with Suspected HF (Group S):

  • Able to understand and accept the study constraints and to provide informed consent (either themselves or a legal representative).
  • Age over 65 years (i.e., 65 included).
  • Presence of symptoms and/or signs typical of Heart Failure (defined by the European Society of Cardiology, ESC), including breathlessness (during activity or at rest, lying down, waking up at night needing to catch their breath), fatigue, swollen ankles/legs, and/or palpitations.

• Patients at Risk of Heart Failure due to the presence of cardiovascular (Group R):

  • Able to understand and accept the study constraints and to provide informed consent (either themselves or a legal representative).
  • Age over 65 years (i.e., 65 included).
  • Absence of symptoms and/or signs typical of Heart Failure (defined by the ESC), including breathlessness (during activity or at rest, lying down, waking up at night needing to catch their breath), fatigue, swollen ankles/legs, and/or palpitations.
  • Presence of at least 1 ACC/AHA Heart Failure risk factor, including hypertension, cardiovascular disease (atrial fibrillation, coronary heart disease or stroke), diabetes, obesity, exposure to cardiotoxic agents, genetic variant for cardiomyopathy, or family history of cardiomyopathy that requires an ECG test for any reason in a primary care center or with an indication of a regular health examination where an ECG is included.

Exclusion Criteria:

• Unwillingness or inability to sign the written informed consent.
• Previous Heart Failure diagnosis.
• Unavailability or suboptimal quality ECG.

De-identified Individual Participant Data (IPD) that underlies the results reported in the publication will be shared after publications of primary results upon request of investigators whose proposed research has received EC/IRB approval and after data use agreement.

From the time of publication and for 5 years thereafter.

Upon request of investigators whose proposed research has received EC/IRB approval and after data use agreement.