The scenario pitted 234 doctors against 23 online symptom checkers, websites, and apps that patients often use to come up with their own diagnosis. Who won? The doctors did, by a long-shot. Harvard-based researchers presented the doctors and computers with 45 patient cases that included symptoms and history but no physical exam, lab tests, or imaging results. Some of the cases were serious, others more benign 26 common and 19 uncommon conditions. When it came to making a diagnosis based on symptoms and history alone, physicians were found to be far more accurate than computers.
In their differential diagnosis, the doctors’ first diagnosis was correct 72.1% of the time, while the online symptom checkers were right only 34% of the time. Within the top three diagnoses on their list, physicians got the correct diagnosis 84.3% of the time, while the symptom checkers included the the correct diagnosis in the top three 51.2% of the time. When correctly diagnosing the more serious cases, doctors overwhelmingly beat the computers 79% to 24% (JAMA Internal Medicine online, October 10, 2016).
I wasn’t surprised that the physicians would perform better, said Harvard professor Dr. Ateev Mehrotra, the study’s lead investigator.
Most of the doctors were internists, though some were pediatricians and family practice physicians. The online symptom checkers included popular websites like WebMD, the Mayo Clinic’s Symptom Checker, AskMD, and drugs.com.
Does this mean computers will never be as good as doctors? Probably not, says Dr. Mehrotra (cbsnews.com). They’re likely to improve as we go forward generations and generations and they may even surpass doctors, but not right now. Dr. Keri Peterson, an internist at Lenox Hill Hospital in New York City, says, I don’t see that computers would ever replace doctors. We get so much by just walking into the room and seeing [the patient] and looking at how sick they look.
Although physicians clearly did better than online symptom checkers, they still failed to make the correct diagnosis 15% of the time. This statistic is consistent with other studies. Keep in mind, however, that the physicians in this particular study did not have the benefit of physical exam findings and test results.
No doubt, patients will continue to go to the symptom-checking websites to self-diagnose their ailments. Except for possibly causing undue stress and anxiety, in most cases no harm is done.
