Science
AI tool could rule out heart attacks and reduce hospital stays
Researchers have developed a new artificial intelligence (AI) tool that can rapidly rule out heart attacks in people attending A&E, allowing them to be safely sent home.
The new research, funded by us and presented this week at the British Cardiovascular Society Conference in Manchester, suggests that by rapidly ruling out heart attacks in these patients, the new AI tool – Rapid-RO – could help tens of thousands avoid unnecessary hospital stays each year in the UK, freeing up valuable NHS time and resource.
When someone arrives in hospital with a suspected heart attack, blood tests are used to help confirm their diagnosis. The blood tests measure levels of a protein called troponin – which rises when there is damage to the heart muscle. As this rise may not be reliably seen until hours later, people are often kept in hospital for further troponin tests and ongoing evaluation. Some of these patients will eventually be discharged from hospital without needing hospital treatment after a heart attack is ruled out many hours later.
In the new study, Dario Sesia, BHF-supported PhD student at Imperial College London, developed a new AI tool to identify patients who are at very low risk, who could potentially be discharged early following their initial assessment, avoiding hospitalisation.
Rapid rule-out vs troponin testing
The team developed, trained and fine-tuned the Rapid-RO AI tool using data from over 60,000 patients across the UK, before retrospectively testing it on over 35,000 patients across four UK hospitals, over various time periods. They then compared Rapid-RO’s performance to existing methods of ruling out a heart attack based on troponin testing alone.The AI works by inputting the results of patients’ initial troponin blood test, which is then combined with other data collected upon hospital presentation, and analysed by the algorithm. Patients are then identified as either being in a very low risk group for having a heart attack, or not, supporting clinical decision-making about whether a patient is likely to be safe to discharge early.
Heart attacks ruled out in over a third of patients
The team found that of the 33,727 patients, the AI tool successfully ruled out heart attacks in over a third (36%) of patients, compared to 27% ruled out by troponin testing alone.In addition, among these patients, the current approach of troponin testing missed four times as many heart attacks (108 cases) compared to Rapid-RO (27 cases). The researchers found that the AI was effective regardless of ethnicity, sex and whether patients had Covid-19. With the Rapid-RO AI technology readily available, next the team hopes to develop it into an app that could in future be used by doctors to help support timely clinical decision-making to rule out heart attacks earlier and avoid unnecessary hospital stays.
Harnessing ever-evolving AI technology
Professor James Leiper, our Associate Medical Director, said: “This research demonstrates the important role AI could play in guiding treatment decision for heart patients. By quickly identifying patients who are safe to be discharged, this technology could help people avoid unnecessary hospital stays, allowing valuable NHS time and resource to be redirected to where it could have the greatest benefit.
“It is vital we harness the power and potential presented by ever-evolving AI technologies, and that the NHS is equipped to support their use. We look forward to more research to understand how Rapid-RO could in future be used to accelerate clinical decisions, improving patient treatment and care.”