While research on software and artificial intelligence gains more momentum than ever in all the fields, a new study suggests that physicians are still more competent than their software counterparts when it comes to diagnosis.
The research, which was published on Oct. 10 in the journal Jama Internal Medicine, suggests that physicians' knowledge when it comes to correct diagnoses is more than twice as often better than the symptoms-cracker's apps. The insight is the first to directly compare humans and devices when it comes to diagnoses.
During the last two decades, computer checklists and connected digital applications have gained popularity — becoming widely used in order to prevent medication errors and to respect treatment protocols. Scientists' also wanted to determine if computers can improve clinical diagnoses and also reduce errors. Every day, hundreds of millions of people use the internet or different software to check for symptoms and diagnose themselves based on the information they find online. However, these checkers are not half as efficient as physicians' diagnoses, according to the study.
More than 200 physicians were asked to diagnose 45 medical cases as part of the study — both in common and uncommon conditions — varying in the degree of severity of the illnesses. For each case, the practitioners were supposed to give a diagnosis based on the symptoms they were provided with, along with two other possible conditions. Each of the cases was diagnosed by at least 20 physicians.
Thy managed to obtain better results than the apps, giving correct diagnoses on 72 percent of the hypotheses, while the digital software only managed to score the right diagnosis 34 percent of the time. Of all the physicians, 84 percent gave correct diagnosis on the three possibilities, while only 51 percent of the digital checkers managed to do so.
The less common the conditions, the better the physicians' performance in diagnosing compared to the digital machines. However, according to the comparative study, common illnesses and symptoms could be easily identified by the devices as well.
"While the computer programs were clearly inferior to physicians in terms of diagnostic accuracy, it will be critical to study future generations of computer programs that may be more accurate," explained senior investigator Ateev Mehrotra, an associate professor of health care policy at HMS.
Although the doctors managed to perform better than the machines, 15 percent of their diagnoses also contained errors. According to the study, developing algorithms to handle this type of activity can be used as a complementary means to help doctors make their decisions better in order to reduce the chances of diagnostic errors.
As it seems that more and more people use the internet to check their symptoms and self-medicate accordingly, investment in this area would be necessary in order to shrink the differences between smart diagnosis and the real ones.
