Early detection of vulnerable patients can improve outcomes but poses a substantial challenge in clinical practice. In a multisite randomized controlled trial involving 39 physicians and 15,965 patients, Chin Lin and colleagues at the National Defense Medical Center in Taiwan assessed the efficacy of artificial intelligence (AI)- enabled electrocardiogram (ECG) in identifying hospitalized patients with a high risk of mortality. 

The AI-ECG alert intervention 

The AI-ECG alert intervention involved an AI report and warning messages alerting doctors about patients predicted to be at high risk of death. The trial’s main goal was achieved when it was discovered that sending the AI-ECG alert was linked to a notable drop in all-cause deaths after 90 days. The AI decreased overall mortality in patients at high risk by 31%. 

Eric Topol of the Scripps Research Translational Institute in California, who was not engaged in the research, said, “This is quite extraordinary.” “It is quite uncommon for any pharmaceutical to result in a 31% decrease in death, and even more uncommon for a non-drug — this is merely AI monitoring people.” 

The phenomenal trial 

The AI developed by Chin Lin and his colleagues at the National Defense Medical Center in Taiwan was initially trained on over 450,000 electrocardiogram (ECG) tests, which record the electrical activity of the heart and the survival statistics of the ECG subjects. With each patient’s risk of mortality represented by a percentile score, the AI learned to identify individuals who were at least in the 95th percentile as high-risk. 

The researchers then tested the AI alarm system with 39 doctors at two different hospitals. Every time a nurse uploaded a new patient’s ECG test results to a computer server, an AI system analyzed the data and promptly notified doctors if it determined that the patient was at high risk. 

Focusing on the highest-risk groups 

The AI system not only decreased the probability of mortality from all causes but also decreased the death rate from heart problems in high-risk individuals by almost 90%. According to Lin and his colleagues, the AI predictions might help focus attention on the highest-risk group since doctors typically follow up on alerts with more diagnostic testing and therapy. As per the reports, 14 military hospitals in Taiwan have already deployed the AI alert system. “This can be implemented in every hospital in the world—it shouldn’t be expensive.” “When you see this level of benefit, it should be the standard of care,” Topol said.

• Nature
• Unsplash