A new artificial intelligence (AI) tool could help GPs spot patients at ‘highest risk of conditions leading to heart-related death’, researchers have claimed.
Research findings presented at the European Society of Cardiology Congress in London revealed that the AI algorithm could identify over 400,000 high-risk patients across the UK.
From a testing sample of GP records, the researchers found that those at high risk made up three quarters of the participants who later died of a heart-related condition at a 10-year follow-up.
The tool, named Optimise, has been trained by researchers at the University of Leeds using the GP records of over two million patients.
The researchers hope that it could be rolled out to GPs within two years, telling Pulse that they need to secure ‘buy-in’ from the suppliers who would implement their AI algorithm into GP record systems.
During this initial test, data was collected for 10 ‘cardio-renal-metabolic events’ which included heart failure, diabetes, chronic kidney disease (CKD) and COPD.
The researchers ran a pilot with the AI tool on a group of 82 patients, among whom the prevalence of such diseases was ‘was high’.
They found that, of these 82 patients, one in five received a CKD diagnosis that would ‘not have been picked up otherwise’.
According to these initial findings, many of the high-risk patients had undiagnosed conditions, or had not received the appropriate medications that ‘could help reduce their risk’.
The researchers will now carry out a larger clinical trial which will provide ‘doctor led care to patients’, with hopes that the findings of this will secure more leverage with GP records suppliers.
GPs would then be able to identify high-risk patients ‘at the simple push of a button’, they claimed.
Study author Dr Ramesh Nadarajah, a health data research UK fellow at the University of Leeds, said the AI ‘uses readily available data’ which could help GPs ‘ensure that they are providing timely care for their patients’.
He continued: ‘We hope our research will ultimately benefit patients living with heart and circulatory diseases, as well as helping relieve pressure off our NHS systems, as prevention is often a cheaper solution than treatment.
‘Next, we plan to perform a clinical trial where we are providing doctor led care to patients.’
The research, funded by the British Heart Foundation, used GP records from two million patients aged 30 and over.
Of the 416,000 patients used for the ‘testing dataset’, there were 11,676 ‘cardiovascular deaths’ during the 10-year follow-up.
Those patients who were identified by the AI as being at higher risk – which accounted for around 10% of the total testing cohort – made up 74% of those cardiovascular deaths.
Imperial College London’s head of primary care and public health Professor Azeem Majeed told Pulse that the study had some ‘encouraging findings’ but that that consideration is needed before implementing the AI tool in GP practices.
He said: ‘Risk prediction models often work better in research studies where they have used retrospective data. When used prospectively in real-world settings, they often work less well. Hence, we would need a real-world evaluation of the tool to better understand its utility.’
Professor Majeed also pointed out that there are already other tools for measuring the risk of cardiovascular disease, such as QRISK.
‘We would need to know what newer AI tools add to these more established tools,’ he added.
Dr Nadarajah told Pulse that the Optimise tool is not intended to replace software such as QRISK, but instead aims to look at CVD risks more broadly, rather than only at which patients require statins.
The full peer-reviewed paper is expected in the coming months, while findings from the larger clinical trial are expected to be submitted in the next 12 to 18 months.
