Globally, doctors are in short supply and are also overworked. But now, machine learning could soon help this community to bring down errors in primary care settings.
Artificial Intelligence (AI) symptom checkers are extremely useful in offering medical data and providing safe triaging advice to all users. But none of these symptom checkers performs diagnosis similar to a physician.
Prevalent symptom checkers are different from doctors and give advice only on the basis of correlations—and correlation is not a causative factor.
For the first time, a team of researchers from Babylon Health has applied the principles of causal reasoning to help AI diagnose the written test cases.
The team used a novel method called causal machine learning—which is attracting a great deal of interest in the AI community—to serve as an “imagination.” This would allow the AI to consider the kind of symptoms it is likely to see if the patient develops an illness that is different from the one it was considering.
Published in the Nature Communications journal, the peer-reviewed study demonstrates that the AI is made considerably more precise by disentangling correlation from causation.
We took an AI with a powerful algorithm, and gave it the ability to imagine alternate realities and consider ‘would this symptom be present if it was a different disease’? This allows the AI to tease apart the potential causes of a patient’s illness and score more highly than over 70% of the doctors on these written test cases.
Dr Jonathan Richens, Study Lead Author and Scientist, Babylon Health
Dr Ali Parsa, the Founder and CEO of Babylon Health, added, “Half the world has almost no access to healthcare. We need to do better. So it’s exciting to see these promising results in test cases. This should not be sensationalized as machines replacing doctors, because what is truly encouraging here is for us to finally get tools that allow us to increase the reach and productivity of our existing healthcare systems.”
Dr Parsa continued, “AI will be an important tool to help us all end the injustice in the uneven distribution of healthcare, and to make it more accessible and affordable for every person on Earth.”
A pool of more than 20 general practitioners from Babylon Health developed as many as 1,671 realistic written medical cases; these comprised both atypical and typical cases of symptoms for over 350 illnesses. Every medical case was penned by an individual physician and subsequently validated by numerous other physicians to make sure that it denoted a true diagnostic case.
Yet another separate group, comprising 44 general practitioners from Babylon Health, was individually given at least 50 written cases (the mean being 159) to assess them. The physicians subsequently listed the types of illnesses that, according to them, were most likely (on average yielding 2.58 possible diseases for every diagnosis).
The physicians were evaluated for precision by the number of cases in which they included the actual disease in their diagnosis. The same tests were also taken by Babylon Health’s AI. This AI used the newer causal algorithm as well as an older one based on correlations (particularly created for this study and not taken from Babylon Health’s product).
For every test, the team observed that the AI could only report as many answers as those provided by the physician.
The physicians had a mean score of 71.40% (± 3.01%) and spanned from 50% to 90%. While the older correlative algorithm scored 72.52% (± 2.97%), which was equivalent to the average physicians, the newer causal algorithm achieved 77.26% (± 2.79%), which was higher than 32 of the physicians, equal to 1, but less than 11.
I’m excited that one day soon this AI could help support me and other doctors reduce misdiagnosis, free up our time and help us focus on the patients who need care the most. I look forward to when this type of tool is standard, helping us enhance what we do.
Dr Tejal Patel, General Practitioner and Associate Medical Director, Babylon Health
Dr Saurabh Johri, the study’s author and Chief Scientist from Babylon Health, stated, “Interestingly, we found that the AI and doctors complemented each other—the AI scored more highly than the doctors on the harder cases, and vice versa.”
Dr Johri continued, “Also, the algorithm performed particularly well for rare diseases which are more commonly misdiagnosed, and more often serious. Switching from using correlations improved accuracy for around 30% of both rare and very-rare conditions.”
The fundamental disease models used by an AI to get improved precision do not have to be modified. It is an advantage that would be applicable to prevalent correlative algorithms, such as those beyond the medical environment.
Causal machine learning allows us to ask richer, more natural questions about medicine. This method has huge potential to improve every other current symptom checker, but it can also be applied to many other problems in healthcare and beyond—that’s why causal AI is so impressive, it’s universal.
Dr Ciaran Lee, Study Author and Honorary Lecturer at University College London
Dr Lee was previously working at Babylon Health.
The technology presents new opportunities for an upcoming association between AI and clinicians that will expedite a physician’s diagnosis, free up time for clinicians, enhance precision, and improve patient experiences and patient outcomes. The technology could also boost the clinicians’ work and continue to fuel a more improved healthcare system for patients.
The latest causal algorithm is yet to be presented in the publicly available app of Babylon Health. Moreover, the algorithm has to be subjected to additional development and testing before it can be released. Once it has fulfilled all the required regulatory approvals in the United Kingdom and other markets, it will be launched in these markets.
Richens, J. G., et al. (2020) Improving the accuracy of medical diagnosis with causal machine learning. Nature Communications. doi.org/10.1038/s41467-020-17419-7.