A study published in JAMA Internal Medicine found that physicians were twice as likely to correctly diagnose a patient, compared with computer algorithms or “symptom-checkers” used on apps or websites like WebMD.com.
Hannah L. Semigran, BA, of Harvard Medical School in Boston, Massachusetts, and authors asked 234 physicians to make a patient diagnosis in 45 different clinical scenarios. Each clinical scenario included a patient’s medical history, but no physical examination or test findings. Physicians had to identify the most likely diagnosis along with two additional possible diagnoses.
Physicians outperformed the symptom-checker apps, listing the correct diagnosis 72 percent of the time compared with 34 percent for the symptom-checkers (p<0.001). In addition, 84.3 percent of physicians versus 51.2 percent of the symptom-checkers listed the correct diagnosis as one of the top three possibilities (p<0.001). The difference between physician and computer performance was most notable in more severe and less common conditions.
However, physicians made errors in approximately 15 percent of cases. The authors noted that developing computer-based algorithms to be used in conjunction with human decision-making may help further reduce diagnostic errors.
Sources: Semigran HL, Levine DM, Nundy S, et al. Comparison of physician and computer diagnostic accuracy. JAMA Intern Med. 2016 October 10. [Epub ahead of print]; Harvard Medical School news release, October 11, 2016.