[Skip to Navigation]
Comment & Response
July 2019

Unsound Evaluations of Medical Machine Translation Risk Patient Health and Confidentiality—Reply

Author Affiliations
  • 1Division of General Internal Medicine, Department of Medicine at Zuckerberg San Francisco General Hospital, University of California, San Francisco
  • 2Center for Vulnerable Populations at University of California, San Francisco
JAMA Intern Med. 2019;179(7):1001-1002. doi:10.1001/jamainternmed.2019.1878

In Reply We would like to thank Downie and Dickson for their comments on our study.1 They raise important concerns about our methodology and advice for clinicians using Google Translate (GT).

We do not believe GT can replace certified medical interpreters/translators, who remain the gold standard and should be used for patient communication as required by the Affordable Care Act.2 We agree that certified medical translators are less likely to make the dangerous errors made by GT in our study. However, because systems are using telephones or video to provide interpretation at a distance,3,4 some clinicians are gravitating toward machine translation of patient instructions without knowledge of its risks or benefits. We did not examine benefits of GT to patients and instead wanted to assess its accuracy and safety when providing written instructions to patients with limited English proficiency in their preferred language. We remain convinced that our study adds important information about the risks associated with using GT.

Add or change institution