A recent study conducted by Western Sydney Local Health District (WSLHD) researchers has revealed that artificial intelligence (AI)-driven algorithms exhibit greater accuracy in predicting the risk of future heart attacks or strokes compared to the models currently used worldwide.
However, these findings suggest that the excitement behind AI algorithms potentially revolutionising the way we assess cardiovascular risk may be overhyped.
The team led by the esteemed Professor Clara Chow AM, Cardiologist at Westmead Hospital, assessed current research on the place of AI machine learnings in clinical practice, and identified several significant gaps in this area.
Startlingly, of the 16 research papers analysed in the review, only one provided the programming code necessary to replicate their AI model—a critical requirement in artificial intelligence research.
Moreover, none of the papers included a link to a functional example of their AI-driven prediction algorithm.
Another suggestion of the WSLHD research, published in European Health Journal Quality of Care and Clinical Outcomes, was that AI-driven algorithms can potentially be implemented into hospital and GP electronic medical record systems as ‘living’ algorithms, which self-correct with each patient encounter.
This could allow for significant public health benefits and automatic identification of at-risk patients as they attend hospitals or GP clinics, and provide more targeted and timely advice for heart disease prevention to our local populations.
Machine learning and artificial intelligence applications in medicine lie in the intersection of medical research and computer science, and as such greater collaboration between these fields are needed.
Future research needs to satisfy the requirements of both areas of research for robust models which make sense and can be applied.