I attended a brilliant seminar at Imperial College last week on the role of machine learning and artificial intelligence (AI) in infectious management, and to a lesser extent, infection prevention and control. There’s so much potential for this exciting technology to revolutionise the way we identify, treat, and prevent the spread of infectious diseases. But, there’s also some risks – some are already asking whether the robots are taking over, and whether that is an entirely good thing!
Dr Luke Moore kicked off by describing the development and impact of clinical decision support systems to improve antibiotic prescribing (reviewed here). There is evidence that overall prescribing is reduced, but a lack of good quality evidence around improved patient outcomes. Luke and Dr Pau Herrero Viñas described the use of machine learning to develop a tool to predict the risk of infection using pathology data, which performed impressively with a sensitivity of over 80% and a specificity of over 90% (more details on this approach here). Prof Anthony Gordon then described the development and impact of the ‘AI Clinician’: this clinical decision support tool has already demonstrated impressive impact in improving sepsis-related and other patient outcomes. Machine learning and AI (there is a difference by the way – although not sure that I understand it well enough to explain it to you!) is already changing the way we deliver medicine – and there’s much more is to come!
Another area where machine learning and AI is likely to revolutionise our practice is in laboratory diagnostics. For example, CPE are a diagnostic nightmare – with multiple species, multiple genetic mechanisms, and an indistinct phenotype. Dr Jesus Rodriguez Manzano described a clever approach by mining enhanced data from existing PCR outputs using a machine learning algorithm in order to improve diagnostic performance.
I was interested in Dr Raheelah Ahmad’s talk around public perceptions of AI and machine learning. I think it’s fair to say there’s a sense of disease among the public around the risks associated with machine learning and AI – these are captured in this article: “Scary robots – examining public responses to AI”. I think it’s fair to say some of these concerns would be shared by clinicians. Can we really trust AI to make or even support clinical decisions? Do these systems threaten to deskill? Will these systems be too technical to be adopted by some? (One speaker said “If a child can use my new point of care diagnostic kit, then a clinician can”. Well…remember children have been texting and playing video games since leaving the womb (note: never challenge a child to a thumb war)).
Overall, an excellent seminar. I learnt a lot and left thinking about how machine learning and AI could be applied to improve our infection prevention and control related decision making.