Cardiologists in Toulouse, France, have used artificial intelligence to uncover a stark gap in survival rates among heart attack patients, finding that older women with atypical symptoms are far more likely to die within a year than other patients treated for the same condition.
The researchers analysed data from more than 5,000 patients hospitalised in cardiac intensive care after a myocardial infarction, drawing on a large epidemiological survey run by the French Society of Cardiology. Rather than relying on traditional statistical models, the team asked an AI tool to sort the patients into groups on its own, without any preset criteria or starting hypothesis, in order to assess their prognosis after a heart attack. This approach, known as unsupervised machine learning, allows an algorithm to find natural groupings within a dataset rather than testing a theory chosen in advance by researchers.
The findings, published in the scientific journal Frontiers, surprised the researchers themselves. Four distinct groups of patients emerged from the AI's classification, each with a markedly different outcome.
Four patient profiles, four different outcomes
The first group had what the researchers described as a classic clinical profile: patients aged 68 on average, 81% of whom underwent revascularisation, the procedure that restores blood flow through a blocked coronary artery. Their death rate within a year was 6%.
The second group consisted of older women, aged 82 on average, who presented atypical signs of a heart attack and had a history of coronary disease. Only 71% of them were revascularised. This group's one-year death rate was 15.7%, far higher than any of the other three groups, and the steepest gap the researchers found.
The third group was made up of younger patients, aged 44 on average, 73% of whom were smokers, who showed typical clinical signs of a heart attack. Of these, 85% underwent revascularisation, and their one-year death rate fell to just 1.5%.
The fourth group, with an average age of 57, brought together patients who were overweight and had high blood pressure and high cholesterol. Some 87% of them were revascularised, and their one-year death rate stood at 2.3%.
Gap persists six years on
Six years after the initial hospitalisation, the gap in prognosis between the group of older women and the other three groups remained just as wide, according to the researchers. They said this confirmed that a delay in treatment, or the absence of revascularisation altogether, leaves this group of patients more vulnerable in the long term as well as immediately after their heart attack.
'The risk is not written on the patient's forehead'
Professor Jean Ferrières, a cardiologist at Toulouse University Hospital, professor at the University of Toulouse and a researcher at the Centre for Epidemiology and Research in Population Health, a joint unit of Inserm and the University of Toulouse, said the results showed that risk is not written on a patient's forehead, and that some patients, particularly women, need to be treated earlier even when their symptoms are atypical.
He said doctors on the front line, including emergency physicians, need to be aware of these conclusions, and that patients themselves may need to pay closer attention to how they describe their symptoms when they first seek help. He said the disparities identified in the study need to be corrected.
Why women's heart attacks are often missed
A heart attack, or myocardial infarction, happens when blood flow to part of the heart muscle is suddenly blocked, usually by a clot in a coronary artery. The best-known warning sign is crushing chest pain that can spread down the arm, but women are more likely than men to experience atypical symptoms such as extreme fatigue, nausea, breathlessness, or pain in the jaw, neck or back. These symptoms are more easily mistaken for less serious conditions, which can delay both diagnosis and treatment.
The French Society of Cardiology, which supplied the epidemiological data used in the study, is the main professional body representing cardiologists in France. Frontiers, the journal in which the findings were published, is a Switzerland-based publisher of peer-reviewed scientific research.
