Loading ...

An artificial intelligence (AI) system that sends text messages to alert hospital physicians about the high risk for mortality in their patients reduces the number of deaths, according to a study published in Nature Medicine.

Chin-Sheng Lin, PhD, associate professor of cardiology at the Tri-Service General Hospital of the National Defense Medical Center in Taipei, Taiwan, and his colleagues have developed an AI system that identifies patients with a high risk for mortality on the basis of a 12-lead ECG. The system is intended to identify patients who would benefit from intensified care.

"It is widely acknowledged that providing intensive care to critically ill patients reduces mortality. Delays in providing intensive care for critically ill patients result in catastrophic outcomes. Most in-hospital cardiac arrests are potentially preventable; however, the early signs of deterioration might be difficult to identify," wrote the researchers.

Detecting What Humans Miss

The 12-lead ECG is a common diagnostic tool in hospital emergency departments and other wards. The AI, which was trained on more than 450,000 ECGs, identifies deteriorating patients based on signs that are not readily accessible to human medical staff.

The randomized controlled study involved 39 doctors and 15,965 patients. The AI intervention included a more detailed report from the AI as well as warning messages sent to the caregivers' mobile phones. The affected patients were then monitored more closely or transferred to the intensive care unit.

In the control group, the AI also monitored the patients via ECG but did not send warning messages in cases of high risk. In the intervention group, the AI identified a high risk for mortality in 8.9% of patients compared with 8.6% in the control group.

The investigators reported that the AI warning messages led to a significant reduction in overall mortality over 90 days. While 3.6% of patients died in the intervention group, the mortality rate in the control group was 4.3%. A significant reduction was especially observed in cardiac deaths, with the rate dropping from 2.4% in the control arm of the study to 0.2% in the intervention arm.

High-Risk Patients

The reduction in overall mortality associated with AI warning messages was mainly observed in patients with high-risk ECGs. Patients in the intervention group with high-risk ECGs were significantly more likely to be transferred to the intensive care unit than those in the control group. They also received amiodarone more frequently, and further ECGs or tests for NT-proBNP, free calcium, or magnesium were more commonly performed.

The authors emphasized that exactly how the AI warning messages lead to a decrease in overall mortality must still be clarified. But the results suggest that they help in detecting high-risk patients, triggering timely clinical care, and reducing mortality, they wrote.

This story was translated from the Medscape German edition using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.