The patient had just undergone a caesarean section. A native Taiwanese speaker, she couldn’t put English words to her pain. Natasha Mehandru, the on-call obstetrician, was used to communicating with patients for whom English is a second language. But this time, the interpreter who assisted them by telephone was not enough to unblock the situation. “The interpreter was not very good”, she recalls. And she quickly realized that he did not speak the same dialect as the mother.
Natasha Mehandru therefore turned to a familiar tool: Google Translate. Using it to translate words and phrases from Taiwanese to English and vice versa, and with the help of the interpreter who stayed on the line, the doctor and his patient gradually came to understand each other. The pain was not from the C-section, but from an older problem. “I adapted the treatment”, says Natasha Mehandru, then a gynecology intern and now a surgeon at Kaiser Medical Center in San Jose, California. Two days later, the patient felt better.
Federal Rules [aux États-Unis] require hospitals to have on their staff interpreters for common languages, such as Spanish, and to consult interpreters by telephone for rarer languages. But the system leaves something to be desired: sometimes you have to wait a long time and some patients speak dialects for which it is difficult to find an interpreter. To overcome these difficulties, caregivers have found a solution with Google Translate.
Little by little, the tool has become ubiquitous in the hospital. “It is used on the sly”, acknowledges Elaine Khoong, internist and assistant professor of medicine at the University of California, San Francisco (UCSF). This practice is hidden in particular because it is officially discouraged by the health system, which considers it problematic. Several researchers, including Elaine Khoong, are increasingly demanding that we take a serious look at it: on the one hand, they want us to study the uses and risks of Google Translation in clinical medicine, On the other hand, better machine translation tools should be developed which would complement traditional language services.
Know your limits
“It’s the future”, said Breena Taira, researcher in emergency medicine at the University of California at Los Angeles, who evaluated, in a recent study, the translations by Google, in seven languages, of sheets given to patients when they left the hospital. Tech giants like Google and Microsoft, which are investing heavily in voice recognition software, are interested in medical translation.
“We need to be aware of the limitations of these tools,” explains the researcher. Notably because the accuracy rates are significantly lower for languages that are not common. Machine translation could meet a very important need [négligé pour le moment] by allowing physicians to provide personalized written instructions to non-English speakers. She insists :
“We need to make sure that we can safely use machine translation to give written information in other languages.”
According to research by Elaine Khoong, Breena Taira and other scientists, using Google Translate is particularly risky for emergency department discharge slips.