Why healthcare systems don't understand how Colombians speak

Imagine you walk into a doctor’s visit and describe your symptoms the way you always have: with the words you use at home, the ones you learned growing up, the expressions from your neighborhood. The doctor understands you. But when they record the visit in the system, something shifts. The language you used and the language the system understands aren’t the same.

That gap exists. And it has real consequences.

In Colombia, this is especially pronounced. It’s not just the difference between everyday and clinical language; there are differences between regions too. What you say in Barranquilla isn’t the same as what you say in Pasto, in Medellín, in the Chocó. The expressions people use to describe symptoms, discomfort, and illness vary more than health information systems account for.

The result is threefold: clinical information lost at the moment of registration, data that can’t be compared across institutions, and epidemiological analyses working with incomplete information.

The world’s most advanced health systems use terminology standards like SNOMED CT (an international library of clinical concepts) to ensure the same symptom is called the same thing everywhere. The problem is that standard was built on formal medical language, not how real people talk.

That’s what the Colombian Medical Semantic Translator is trying to solve: building a bridge between patient language and standardized clinical terminology, using AI trained specifically for the Colombian context. Not to replace the doctor or automate diagnosis. To make sure the information a patient shares doesn’t evaporate in the process of recording it.

There’s a principle I keep coming back to: health data is only as good as the language that captures it. If that language doesn’t faithfully reflect what the person said, the system is already starting off wrong.

For me, working on this isn’t just a technical problem. It’s a problem of equity. Everyday language is, in many cases, the language of the communities with the least access to quality healthcare. If systems don’t understand them, they exclude those communities twice.

If you’re curious about the technical details of how the system works, you can check out the full Colombian Medical Semantic Translator write-up.