AI that determines risk of death helps save lives in hospital trial

An artificial intelligence system has proven it can save lives by warning physicians to check on patients whose heart test results indicate a high risk of dying. In a randomised clinical trial with almost 16,000 patients at two hospitals, the AI reduced overall deaths among high-risk patients by 31 per cent.

“This is actually quite extraordinary,” says Eric Topol at the Scripps Research Translational Institute in California, who was not involved in the research. “It’s very rare for any medication to [produce] a 31 per cent reduction in mortality, and then even more rare for a non-drug – this is just monitoring people with AI.”

Chin Lin at the National Defense Medical Center in Taiwan and his colleagues first trained their AI on more than 450,000 electrocardiogram (ECG) tests, which measure the heart’s electrical activity, along with the survival data of the ECG subjects. The AI learned to produce a percentile score representing each patient’s risk of death, with those in at least the 95th percentile considered high risk.

The researchers then tested the AI alert system with 39 physicians at two different hospitals. As nurses uploaded each new patient’s ECG tests to a computer server, the AI system would analyse the results and quickly alert physicians if it identified the patient as being high risk.

In addition to reducing the risk of death from all causes, the AI system reduced high-risk patients’ deaths from heart issues by more than 90 per cent.

Lin and his colleagues speculated that the AI predictions may help focus attention on the highest risk group – physicians usually followed up on alerts with more diagnostic tests and additional treatment.

The AI alert system is already being used across 14 military hospitals in Taiwan.

“This can be implemented in every hospital around the world – it shouldn’t be expensive,” says Topol. “It should be the standard of care when you see this extent of magnitude of benefit.”