Artificial intelligence-assisted diagnosis and prognostication in low ejection fraction using electrocardiograms in inpatient department: a pragmatic randomized controlled trial
Abstract Background Early diagnosis of low ejection fraction (EF) remains challenging despite being a treatable condition. This study aimed to evaluate the effectiveness of an electrocardiogram (ECG)-based artificial intelligence (AI)-assisted clinical decision support tool in improving the early di...
Saved in:
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2025-06-01
|
| Series: | BMC Medicine |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12916-025-04190-z |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Background Early diagnosis of low ejection fraction (EF) remains challenging despite being a treatable condition. This study aimed to evaluate the effectiveness of an electrocardiogram (ECG)-based artificial intelligence (AI)-assisted clinical decision support tool in improving the early diagnosis of low EF among inpatient patients under non-cardiologist care. Methods We conducted a pragmatic randomized controlled trial at an academic medical center in Taiwan. 13,631 inpatient patients were randomized to either the intervention group (n = 6,840) receiving AI-generated ECG results or the control group (n = 6,791) following standard care. The primary outcome was the incidence of newly diagnosed low EF (≤ 50%) within 30 days following the ECG. Secondary outcomes included echocardiogram utilization rates, positive predictive value for low EF detection, and cardiology consultation rates. Statistical analysis included hazard ratios (HR) with 95% confidence intervals (CI) for time-to-event outcomes and chi-square tests for categorical variables. Results The intervention significantly increased the detection of newly diagnosed low EF in the overall cohort (1.5% vs. 1.1%, HR 1.50, 95% CI: 1.11–2.03, P = 0.023), with a more pronounced effect among AI-identified high-risk patients (13.0% vs. 8.9%, HR 1.55, 95% CI: 1.08–2.21). While overall echocardiogram utilization remained similar between groups (17.1% vs. 17.3%, HR 1.00, 95% CI: 0.92–1.09), the intervention group demonstrated higher positive predictive value for identifying low EF among patients receiving echocardiogram (34.2% vs. 20.2%, p < 0.001). Post-hoc analysis revealed increased cardiology consultation rates among high-risk patients in the intervention group (29.3% vs. 23.5%, p = 0.027). Conclusions Implementation of an AI-ECG algorithm enhanced the early diagnosis of low EF in the inpatient setting, primarily by improving diagnostic efficiency rather than increasing overall healthcare utilization. The tool was particularly effective in identifying high-risk patients who benefited from increased specialist consultation and more targeted diagnostic testing. Trial registration ClinicalTrials.gov Identifier: NCT05117970. |
|---|---|
| ISSN: | 1741-7015 |