Artificial intelligence (AI) tools may help to overcome sexism that put women having heart attacks at higher risk of death and undertreatment, research presented at the European Cardiology Congress has shown.
In one study, also published in The Lancet, Swiss and British researchers found that established risk models used to guide management of heart attacks are less accurate in female patients.
They analysed data from 420,781 patients across Europe with non-ST-segment elevation acute coronary syndromes and found that the current GRACE (Global Registry of Acute Coronary Events) risk score was good at determining risk in male patients but ‘notably lower’ in females.
It underestimated the risk of death in female patients and meant they were less likely to receive early invasive treatment.
But using machine learning to account for sex-related differences, they developed an AI-based risk score – GRACE 3.0 – that improved the prediction of who was most at risk to better guide treatment decisions.
Presenting the results, study leader Professor Thomas Luscher from the University of Zurich said: ‘I hope the implementation of this novel score in treatment algorithms will refine current treatment strategies, reduce sex inequalities, and eventually improve the survival of patients with heart attacks – both male and female.’
UK researchers behind a second study also presented at the Congress in Barcelona reported that an algorithm developed using AI could help doctors diagnose heart attacks in women more quickly and more accurately.
Researchers at the University of Edinburgh used data from 10,038 people, 48% of whom were women, who went to hospital with a suspected heart attack to develop the CoDE-ACS tool.
They then validated it on a further 3,035 patients outside of the UK, a third of whom were female.
The tool looks at factors including sex, age, observations, ECG findings and medical history and troponin test results.
It was able to rule out a heart attack with 99.5% accuracy, and identified those who needed further tests, in whom the final diagnosis was a heart attack, with an accuracy of 83.7%.
This compares to an accuracy of just 49.4 per cent with current tests, the researchers said.
Dimitrios Doudesis, data scientist at the BHF Centre for Cardiovascular Science, University of Edinburgh, said they have now embedded the algorithm into a mobile app to support doctors in making treatment decisions.
‘Whilst the troponin test takes 30 minutes to process, we take an array of other health information and add it into the app alongside the troponin measurement.
‘This provides doctors with a precise and instantaneous score to confirm if they can reassure their patient that they haven’t had a heart attack and send them home, or if they require further tests.’