Key result
A 12-lead ECG-based logistic regression model predicted reduced ejection fraction with an area under the curve of 0.684.
Why the study?
Early identification of heart failure with reduced ejection fraction by emergency medical service providers is challenging, especially in rural settings.
Does a logistic regression model using 12-lead ECG and demographic features predict reduced ejection fraction?
Observational (n=2,289)
No
Does a logistic regression model using 12-lead ECG and demographic features predict reduced ejection fraction?
A computationally efficient logistic regression model using standard 12-lead ECG data and demographics can predict reduced ejection fraction with an AUC of 0.684, offering a potential decision support tool for EMS providers.
Accelerates early GDMT initiation for rEF in rural EMS care; confirms efficacy of pragmatic decision aid from standard 12-lead data.
Objectives/Goals: Early identification of heart failure with reduced ejection fraction (rEF) by emergency medical service (EMS) providers is challenging, particularly in rural settings. We developed a rEF decision aid that addresses barriers to adoption using demographics, measurements, and computerized findings from ubiquitous EMS 12-lead defibrillators. Methods/Study Population: We identified all patients at Tufts Medical Center between Jan 1, 2012 and Apr 16, 2024 who had a 12-lead ECG taken within 30 days of an echocardiogram. Using the echocardiogram left ventricular ejection fraction (LVEF), each ECG-echo pair was classified as reduced EF (rEF – LVEF < 40) or non-rEF (LVEF >= 40%). The rEF is the outcome in a step-wise logistic regression model (LRM) to predict rEF. Variables of interest in the model were selected based on literature review in clinical assessment for heart failure with reduced ejection fraction (HFrEF) and what can be provided by a 12-lead ECG. We also included left ventricular damage score (e.g., Selvester Score) due to previous myocardial infarction (MI). Prior MI is a known risk factor in HFrEF. We used R 4.5 to perform the analysis and logistic regression modeling. Results/Anticipated Results: 2,289 patients were identified with a 12-lead ECG taken within 30-days of an echocardiogram. Our final regression model included demographic and ECG features: age, sex, myocardial scarring score, computerized interpretation of conduction defects (bundle branch block, early repolarization, and intraventricular conduction defect), and acute myocardial infarction findings. The rEF prediction model had an area under the curve of 0.684 and good calibration along the identity line up to predicted probability of 0.3. Discussion/Significance of Impact: A LRM to predict rEF using an ECG is computationally efficient and address important barriers of adoption in the EMS setting. It does not require additional EMS training or equipment. LRM transparency facilitates communication between paramedics and medical control clinicians to make shared decisions regarding treatment and transport decisions.
No takes yet. Share an insight, caveat, or question.
Manlik Kwong (2026) conducted an observational in Heart failure with reduced ejection fraction (HFrEF) (n=2,289). 12-lead ECG-based logistic regression model was evaluated on Prediction of reduced ejection fraction (LVEF < 40%) (AUC 0.684). A 12-lead ECG-based logistic regression model predicted reduced ejection fraction with an area under the curve of 0.684.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: