Machine translation does not belong in healthcare

Our CMedO, Annie Backman, shares her opinion on machine learning tools in healthcare and why it is dangerous.

How many times haven’t we laughed at translation mistakes, so obviously generated by machine translation tools, or in some cases human translators who just haven’t had access or comprehended the context?

For example when a large Ecommerce site advertises "Drapes with rape seed fields"  and the word “rape” is translated to mean something else than the yellow flower.

Or literal translations such as “union” translated to the meaning of workers associations, instead of something that is combined or coupled.

A celebrity condoleance becoming “nobody liked you” instead of “(there’s) nobody like you” when translated. Honestly, these are usually hilarious! (For more crazy translation mistakes, visit language connection, the local, cultures connection or scrybs)

The danger of using machine translations in healthcare

But it isn’t nearly as fun in a healthcare context For example when

  • “Vomit” becomes “answer” (Tigrinya)
  • “Lay still” becomes “land in you field” (Finnish)
  • “Placenta” becomes “baby’s wife” (Turkish).


Healthcare professionals and patients rely on communication.

Communication is what lays the ground for decisions on sampling, further examinations and in the end diagnosis and treatment. Without it, or even worse, with faulty communication, patients do not only risk getting confused or doubting the quality and professionality of the care, but they also risk getting delayed treatment, being misdiagnosed or at least need a prolonged hospital stay, with all the risks that that means.

Even so, and even though healthcare professionals are highly skilled and intelligent and probably know all too well that there’s a risk of bad translations, usage of machine translation tools in healthcare is pretty common.

The staff simply feels they have no other option, when standing in front of a patient with no common language. (Interpreters are great, but they’re not there all the time during a patient’s contact with healthcare. Very far from all the time, actually.)

To sum up - using common machine translation tools in healthcare is definitely beyond a joke.  “The laugh catches in the throat”, as a big machine translation tool would translate the equivalent Swedish expression…

Well, no need to tell the readers of this blog that there IS a safe, practical, costworthy option literally at your fingertips twenty-four-seven. Please help us spread the word and the name of our (in my opinion) fantastic tool, so that it can save more time, money, misunderstandings and in the end, patients’ lives.