AI-enabled ECGs reduce time to cath lab in patients with suspected STEMI
Click Here to Manage Email Alerts
Key takeaways:
- An AI-enabled ECG identified patients with STEMI, the worst kind of heart attack, and alerted the cath lab.
- Use of the system reduced ECG-to-cath-lab time by 9 minutes compared with the usual care.
PHILADELPHIA — An artificial intelligence-enabled ECG was associated with shorter time from ECG to catheterization laboratory entry in patients with suspected STEMI, according the results of the ARISE study.
Chin-Sheng Lin, MD, cardiologist at Tri-Service General Hospital, Neihu District, Taipei, Taiwan, and colleagues developed an AI ECG algorithm to detect STEMI and improve STEMI detection performance.
Image: Adobe Stock
“It is frequently a challenge for emergency physicians to recognize STEMI, with an initial misdiagnosis rate of 20.5%,” Lin said during a press conference at the American Heart Association Scientific Sessions. “Developing a decision support system to diagnose STEMI by ECG early provides an unmet need for clinical physicians.” At Tri-Service General Hospital, before implementation of the AI-based ECG, door-to-balloon time was 68.5 minutes, but after implementation, it fell to 60.5 minutes, he said.
For ARISE, the researchers randomly assigned more than 43,000 patients with suspected acute MI presenting to Tri-Service General Hospital from May 2022 to April 2023 to the AI-supported ECG or the usual care.
After exclusions, the intervention group included 21,555 patients and the control group included 21,622 patients. In both groups, the mean age was 60 years, 49% were men and approximately 37% came from the inpatient department as opposed to the ED.
For the intervention group, the AI algorithm analyzed the ECG and sent an alert for the highest-risk patients to the on-duty cardiologists and the ED doctors to activate the primary PCI team, Lin said during the press conference.
During the process, patients were classified as STEMI confirmed by coronary angiography, non-STEMI ACS confirmed by coronary angiography, probably not STEMI (not confirmed by coronary angiography) and non-coronary angiography-validated STEMI.
The primary outcome was time between ECG and arrival to cath lab door. Secondary outcomes included event analysis, ejection fraction, peak troponin-I level, peak creatinine kinase level, length of hospitalization and diagnostic accuracy.
The primary outcome was calculated in patients with confirmed STEMI.
Among patients with STEMI confirmed by coronary angiography (77 in the intervention group and 68 in the control group), the mean age was 65 years, 80% were men, and 9.1% of those in the intervention group came from the inpatient department compared with 1.5% in the control group, Lin said.
The time from ECG to arrival at cath lab door was 43.3 minutes (95% CI, 29-58.3) in the intervention group and 52.3 minutes (95% CI, 44.1-68.6) in the control group, according to the researchers.
In patients who came from the ED, the difference was more pronounced overall (P < .001) and when the patient arrived during regular hours compared with off-hours (P for interaction = .012), Lin said.
The intervention group had a lower rate of non-coronary-angiography-validated STEMI than the control group (6.5% vs. 15.8%; OR = 0.37; 95% CI, 0.14-0.94), according to the researchers.
There were no differences between the groups in ejection fraction, peak troponin-I level, peak creatinine kinase level and length of hospitalization, the researchers found.
In the diagnostic accuracy analysis of the AI-enabled ECG, the positive predictive value was 88%, the negative predictive value was 99.9%, the sensitivity was 88.8% and the specificity was 99.9%, Lin said at the press conference.
“We showed that application of this AI ECG system significantly reduced [time from ECG to arrival at cath lab door] from 52 to 43 minutes,” Lin said at the press conference.
In a discussion during the press conference, Brahmajee K. Nallamothu, MD, MPH, FAHA, interventional cardiologist and professor of internal medicine at the University of Michigan, said the study demonstrated “an incredible use of AI technology in a real-world problem. This is a clinical problem that has large implications, particularly for under-resourced areas.
“While the time-to-treatment improvement may appear to be small ... at a population level, this is an enormous improvement, particularly at an institution where they already had a high-functioning system,” he said.
Collapse
Lin CS, et al. LBS.08. Artificial intelligence at the bedside. Presented at: American Heart Association Scientific Sessions; Nov. 11-13, 2023; Philadelphia.
Click Here to Manage Email Alerts