Share

We have been comparing man to machine since, well… man made machines. In the past several years, with the advancements in medical technologies, there have been numerous studies comparing a physician’s diagnostic skill versus a computer via health symptom checker apps. Most recently, in the Journal of the American Medical Association Internal Medicine, researchers compared diagnostic accuracy of physicians versus 23 different symptom checker websites or apps with standard clinical vignettes.

Not surprisingly to me, physicians fared much better at not only naming the correct diagnosis first more commonly, but also having the correct diagnosis in the first three of their differential more often than the computer system.

With the likely millions of pages of medical texts and journal articles written, it would inherently seem that a computer would be able to search through them instantly and have higher accuracy, but, even when incorporating likelihood ratios and instructing the computer to value them into their diagnostic/differential diagnoses, health symptom checker websites and apps are not as accurate as a doctor. When the symptom checker website or app takes into account patient history and symptoms the results are often full of “zebras” (rare diagnoses) rather than horses, common diagnoses based on the same set of symptoms & history.

Through my own experience as a clinical telemedicine specialist, I have tested numerous apps that claim to make a diagnosis and provide a treatment plan via a picture submitted through a smart phone camera. Very often, the diagnosis is incorrect. While not life threatening to come up with a misdiagnosis for simple acne, I’ve found I can take a picture of the palm of the hand, for example, and the computer still yields an acne diagnosis and treatment plan (obviously incorrect).  This scenario underscores the long way that health symptom checker websites, apps, and artificial intelligence still have to go in order to achieve human diagnostic accuracy.

Now, even with the shortcomings of a 100% automated medical diagnostic system, there are still issues with human error and no physician is always 100% accurate in his/her diagnostic work ups or treatment plans.  This is where medical intelligence by physicians (the human kind) and medical technology comes together. A combination of medical algorithms, patient history, symptoms, and laboratory findings can assist a healthcare provider with clinical decision-making. Having access to gold standard, evidence-based resources through The Medical Algorithms Company, can shorten the diagnostic work up time, improve accuracy, and potentially reduce costs while at the same time improving healthcare for millions of patients.

So while a rise of machines that take over doctors’ jobs and eliminate  the need for patient interaction with humans, is not likely to happen in the near future, physicians should realize that the future of medicine will rely on combining the elements of human touch and human intelligence with diagnostic and treatment assistance via medical algorithms and artificial intelligence software.


Share