Apps or appointments? Physicians beat algorithms in diagnostic accuracy

Everyone has checked symptoms online, using websites such as WebMD, to help diagnose themselves, but are these kinds of computer diagnostics accurate—or at least better than an actual physician? Harvard Medical School student Hannah L. Semigran with researchers from Brigham and Women’s Hospital and the Human Diagnosis Project have good news for doctors.

Patients relying on symptom checker apps and websites aren’t getting as accurate of a diagnosis as they would by visiting a real physician, according to a study published in JAMA Internal Medicine. Researchers compared the diagnostic accuracy of physicians and a symptom checker app in evaluating 23 symptoms using 45 clinical articles that included patient medical history but no physical exam.

Results included

• Physicians were able to list the correct diagnosis one the first try 72.1 percent of the time, symptom checkers were only able to correctly diagnose 34 percent.
• Physicians could diagnosis correctly for high-acuity articles and uncommon cases, symptom checkers could do the opposite and were more likely to give the correct diagnosis for low-acuity and common articles.
• Physicians were also able to list the top 3 diagnoses at a rate of 84.3 percent compared to the symptom checkers 51.2 percent.

“Despite physicians’ superior performance, they provided the incorrect diagnosis in about 15 percent of cases, similar to prior estimates (10 to 15 percent),” wrote Semigran et al. “While in this project we compared diagnostic performance, future work should test whether computer algorithms can augment physician diagnostic accuracy.”