When—and where—can ECG algorithms help humans?

GE HealthCare

Many medical professionals, not just cardiologists, conduct ECGs. And of course, accurate interpretation of those ECGs not only drives accurate referrals, but it also accelerates diagnostic decision-making and saves lives.

Fortunately, there are multiple resources available to help improve ECG interpretation for both cardiologists and non-cardiologists. Physicians are increasingly relying on sophisticated ECG algorithms, or computerized ECG interpretations (CEI), to enable efficient workflows and augment human capabilities regarding interpretation.

Research on CEI support for ECG interpretation

In a series of recent studies, Dr. Anthony Kashou (known as "The EKG Guy") set out to test if human ECG interpretations were better with the support of rules-based CEI—and if so, how much and in which scenarios.1

"Our work centered on evaluating how computerized ECG interpretation affects the interpretation skills of a diverse group of medical professionals spanning different disciplines and training levels," Dr. Kashou said. "We sought to understand whether computerized ECG interpretation can genuinely enhance diagnostic performance, bridge the gaps in ECG interpretation proficiency, and adapt to the growing role of diagnostic and predictive analytic models in contemporary medicine."

He found that there are indeed many use cases for CEI that have shown significant improvement—up to 19.9%—in interpretation accuracy, including for STEMI and primary rhythm analysis.

While Dr. Kashou's study observed non-cardiologists because he wanted to study those with the most to learn and who might derive the most benefit from CEI, he says his findings apply to everyone reading an ECG, cardiologists included. And perhaps with the findings of improved confidence and accuracy in interpretation among non-cardiologists most pronounced, there is now an opportunity to test the assumption that cardiologists indeed have less to learn in future iterations of this research.

"Everyone becomes an expert at something, and that something may not always be ECG," he said. "If you're not interpreting ECG every day, you might benefit from computerized interpretation assistance to refamiliarize yourself with those skills over time."

Improvements across diagnostic accuracy, efficiency, and confidence

Dr. Kashou's work has stratified the impacts of ECG algorithms across medical professions—including Primary Care Practitioners, residents, medical students, nurses, advanced practice providers, and allied health professionals.

Of that group, it's potentially unsurprising that the providers who received the most cardiology-specific training (fellows), also received the least benefit from CEI: Fellows' interpretation accuracy improved by 10%, compared to medical students at 19.9%, advanced practice providers at 17.1%, and nurses at 16.1%.

Even though it's not 19.9%, an improvement of 10% is quite significant. Fellows—like cardiologists—may start from a higher baseline understanding of ECG, but they may still derive benefit from the computer.

"The use of computerized ECG interpretation led to significant improvements in ECG interpretation accuracy, efficiency, and confidence among the diverse group of healthcare professionals," Dr. Kashou said. "Computerized ECG interpretation greatly aided in the primary rhythm analysis and the identification of emergent findings, including STEMI, which alter patient management."

While we don't know to what extent the computer could have assisted cardiologists because they were not included in the study, it's reasonable to assume there could have been at least some marginal impact from CEI. That said, since the CEI is based on a certain set of pre-ordained rules, expert cardiologists with a vast array of domain knowledge of these rules may not necessarily benefit. However, a trusted CEI may support and guide a cardiologist's clinical workflow.

Similarly important is the potential for fellowship training, Dr. Kashou adds. For cardiologists who supervise fellow programs, consider how CEI might transform the learning experience without asking too much of time-strapped specialists: If you're not using CEI during fellows' ECG training, you may be missing an efficient, effective, and underutilized opportunity for fellows under your wing to sharpen their ECG skills.

"CEI is not often used in medical training and that's something we hope to see change," Dr. Kashou said. "If we demonstrate at the learning stage during a clinician's training where CEI is most valuable, we can potentially change their perspective on these tools as they go into practice. If you know, 'I can trust this computer because I remember it's good at sinus rhythms,' for example, you're much more likely to optimize CEI's benefits."

Balancing the human factor

Dr. Kashou points out that his study was a positive one in that it highlighted many of CEI's core capabilities. CEI can and does support human ECG readers, and the combination of human reader and rules-based, algorithm support is better than humans alone for the clinical situations identified.

"We saw an accuracy of efficiency and confidence, and the computer also provides immediate feedback on interpretation. This feedback can help learners identify and rectify mistakes. One way that computers are effective is that they're consistent; if they're consistently shown a pattern, they can very quickly identify that. Plus, it's consistent from case to case, and always learning," said Dr. Kashou. "The other important aspect is how computers evolve with these deep learning models, and they're able to continuously learn and pick up on more nuanced findings."

Although Dr. Kashou says he was surprised by how much CEI supported human interpreters, he emphasized that this is not an opportunity for the machine to take over. The human factor will always be necessary when it comes to ECG interpretation.

"While computerized ECG interpretation offers numerous advantages, it should be considered a complementary tool," he said. "It aids and enhances human expertise but should not replace the critical thinking and clinical correlation inherent to human judgment."

Ultimately, there's a time and place for everything—even CEI. Education is critical, including how to read ECGs as well as how to read them with algorithm support. This is a key learning worth consideration for any future implementation of rules-based support.

"Computerized ECG tools present an opportunity to bolster traditional training methods," Dr. Kashou said. "They provide immediate feedback, enabling trainees to cross-check their interpretations, hone their skills, and gain confidence in real time. As learners acquaint themselves with the nuances of these software tools, they'll grasp how invaluable they can be in clinical settings. Acting as a safety net, these tools can detect potential oversights, enhancing clinical efficiency and patient safety."


1. Kashou AH, Noseworthy PA, Beckman TJ, et al. Impact of computer-interpreted ECGs on the accuracy of healthcare professionals. Current Problems in Cardiology. 2023;48(11):101989. https://doi.org/10.1016/j.cpcardiol.2023.101989.