In face-off, human doctors more accurate than symptom checker apps

A pediatrician and his patient

Human doctors made a correct diagnosis more than twice as often as 23 commonly used symptom-checker apps in a head-to-head comparison.

The research, published in JAMA Internal Medicine, builds on a previous evaluation of symptom-checker apps and websites that found many had deficits in both diagnosis and triage accuracy.

The researchers asked 234 physicians to make diagnoses for 45 “clinical vignettes” that included patient medical history, but no results from a physical exam, blood test or other kind of lab work. They were asked to name the most likely diagnosis, along with two other possibilities.

Webinar

Curating a Higher Level of Personalized Care: Digital Health + Mom

A long-term digital health strategy is needed to respond to the technology demands of the modern patient while thriving as an independent hospital in a fiercely competitive market. In this webinar, Overlake and one of its digital health partners, Wildflower Health, will discuss how Overlake has approached digital health and why it chose to focus early efforts on expectant moms within its patient population.

The physicians were in internal medicine, family practice or pediatrics. About half were in residency or fellowship, which meant they had not completed their training, a Los Angeles Times article points out.

Still, they named the correct diagnosis first 72 percent of the time, while the online tools listed the correct diagnosis first just 34 percent of the time. Doctors made the correct call among their top three picks 84 percent of the time, while the symptom checkers did so just 51 percent of the time.

For common conditions, doctors’ accuracy was 70 percent, compared to 38 percent for apps. The difference grew to 76 percent vs. 28 percent for uncommon conditions. Accuracy registered at 65 percent to 41 percent for cases with low acuity and 79 percent to 24 percent for cases with high acuity.

However, physicians still made errors in about 15 percent of cases.

"Clinical diagnosis is currently as much art as it is science, but there is great promise for technology to help augment clinical diagnoses," co-author Ateev Mehrotra, an associate professor of healthcare policy at Harvard Medical School, said in an announcement of the findings.

Suggested Articles

Average list prices for the most common specialties associated with surprise medical bills outpaced growth for all other specialties, analysis finds.

Medicare Advantage rebates to plans are expected to reach historic highs next year, MedPAC analysis finds.

The VA launched the National Artificial Intelligence Institute to prioritize AI R&D to improve veterans' health and public health initiatives.