Computer-Interpreted ECGs Sometimes Miss Acute Coronary Occlusion
By Stacey Kusterbeck, Special for ReliasMedia.com
Was the patient’s electrocardiogram (ECG) called “normal” by the computer algorithm? It still is possible the patient has acute coronary occlusion.
Stephen W. Smith, MD, has published nearly 40 cases in which the conventional computer algorithm called the ECG normal, but the patient had an acute coronary occlusion that was evident on that ECG.
“Acute coronary occlusion results in a large area of myocardial infarction, and possibly death of the patient, or heart failure. Most do not manifest the ST elevation that physicians are taught to look for,” explains Smith, a faculty emergency physician at Hennepin Healthcare and professor of emergency medicine at the University of Minnesota School of Medicine.
Despite this, the results of several small studies have suggested ECGs labeled “normal” by the computer are unlikely to be clinically significant, and that any delay in physician review would not compromise patient care. Jesse McLaren, MD, audits code ST-elevation myocardial infarctions (STEMIs), and conducted a quality improvement project to reduce diagnostic time for acute coronary occlusion.
As part of those projects, McLaren found several falsely “normal” ECGs. This contradicted the previous studies and echoed the dozens of cases Smith had identified. McLaren, Smith, and colleagues went on to formally analyze seven years of Code STEMI patients requiring coronary intervention. The researchers sought to determine the incidence of computer-interpreted “normal” ECGs, and to see how these patients were managed.
The researchers analyzed all Code STEMI patients from 2016-2022 at two urban academic EDs that were labelled normal by the computer interpretation. Of 536 ED cath lab activations with angiography, there were 18 patients with an initial ECG that had been labeled “normal” but were referred for primary angioplasty for STEMI. In those cases, the algorithm did not reveal any abnormality, much less acute coronary occlusion. Among 394 Code STEMI patients with acute culprit lesions requiring coronary intervention, 16 produced an initial ECG labeled “normal” or “otherwise normal.”
More than one-third of the “normal” ECGs were identified as acute ischemia in real time by the treating emergency physician (EP), despite the false reassurance of the computer interpretation. “These were associated with faster reperfusion,” reports McLaren, an EP at Toronto General Hospital and an assistant professor in the department of family and community medicine at the University of Toronto.
Nearly two-thirds of the cases with initial “normal” ECGs activated the cath lab without the existence of a computer ECG interpretation of STEMI, based on other signs of occlusion MI (including ECG changes, refractory ischemia, and regional wall motion abnormalities).
EPs can shield against risk by viewing ECGs of chest pain patients immediately to identify subtle signs of acute coronary occlusion, according to Smith. “Just as important are the ECGs that the computer labels as non-specific, or as not having STEMI. Approximately 30% of patients with non-ST elevation myocardial infarction have acute coronary occlusion, which is missed on the ECGs, and yet can be detected by experts, or people who have invested a great deal of time and effort into learning how to do it,” Smith says.
Smith advocates for replacing the “STEMI/non-STEMI” paradigm with the “OMI (occlusion myocardial infarction)/Non-OMI” paradigm. “Acute coronary occlusion is not something that should be missed,” Smith stresses.
Ideally, EPs should review every ECG, since conventional computer interpretations can falsely apply the label of “normal” to ECGs that are diagnostic of occlusion MI, according to McLaren. “Moreover, they should learn the subtle signs of acute coronary occlusion. These are high-risk patients, often sitting in the waiting room, with an occluded coronary artery and a falsely reassuring computer interpretation,” McLaren says.
McLaren stresses that EPs should not just sign off ECGs as "STEMI negative," but, instead, should look for signs of occlusion MI. According to 2022 guidelines from the American College of Cardiology, “the application of STEMI ECG criteria on a standard 12-lead ECG alone will miss a significant minority of patients who have acute coronary occlusion. Therefore, the ECG should be closely examined for subtle changes that may represent initial ECG signs of vessel occlusion."
Moreover, for many years, the guidelines have recommended angiography within two hours for non-STEMI patients that are “very high-risk” — with persistent uncontrolled chest pain, new pulmonary edema, shock or hypotension, or electrical instability, Smith notes.
“These ECG findings are evidence-based, can be learned and taught, and can help us advocate for the high-risk group of patients whose ECG doesn't meet STEMI criteria but whose coronary artery is occluded,” McLaren says.
Smith says that in the bigger picture, the solution is artificial intelligence (AI), and has tested a deep convolutional neural network solution for the diagnosis of acute coronary occlusion on the ECG. “The ultimate answer is to use the AI tools for ECG diagnosis of acute OMI,” says Smith. “AI for the ECG diagnosis of OMI is here, and works, and will save lives.”
Considering the risk of acute coronary occlusion in the absence of diagnostic ST elevation, and American College of Cardiology guidelines establishing that ST elevation is not reliable for the ECG diagnosis, EPs face potential legal exposure. “There may be significant liability risk in the future in missing acute coronary occlusion among non-STEMI patients, with the attendant short- and long-term mortality risk and risk of developing heart failure,” Smith warns.
For more on this and related subjects, be sure to read the latest issues of Clinical Cardiology Alert and ED Management, along with the “ECG Review” feature that runs in the 15th edition of each issue of Internal Medicine Alert. For more education on STEMI specifically, check out Relias Media’s STEMI Watch 2023.