[Skip to Navigation]
Comment & Response
May 2017

Limitations of Study on Symptom Checkers

Author Affiliations
  • 1Yorkshire Centre for Health Informatics, University of Leeds, Leeds, England
JAMA Intern Med. 2017;177(5):740-741. doi:10.1001/jamainternmed.2017.0333

To the Editor In their Research Letter in a recent issue of JAMA Internal Medicine, Semigran et al1 provide some interesting insights on the diagnostic ability of physicians vs “symptom checkers” on clinical vignettes. However, the study has 2 important limitations that were not mentioned.

First, an earlier article2 by the same authors shows a wide range of performance of symptom checkers. Mean for including the correct diagnosis in the top 3 was 51.2%, compared with 84.5% for the physicians, but results for the best systems were much better: 67% for DocResponse; 68%, AskMD; and 69%, Isabel. A realistic comparison of symptom checkers and physician accuracy would exclude poorly performing systems.

Add or change institution