Welcome! This is a proof of concept that looks to leverage AI to sort ICD-10 codes into either Problems or Allergies. This example comes from a real need we identified in one of our projects and we decided to make it public so the whole community can benefit from the experiment.
- Clone the repository.
- Set up a
.envfile with your OpenAI API key. - Run the project with
npm run dev. - Access it at
localhost:3000.
Different OpenAI models were tested during the development of this proof of concept, here are the ones that were considered:
- GPT-4 Turbo
- GPT-4.1
- GPT-4.5 Preview
- o3-mini
- o1
After manual testing, we found that o1 model produced the most consistent results.
This is not a production-ready solution. EHR development demands caution, as mistakes can have serious consequences for patients and providers. However, this can be used as an experimental feature provided that:
- You implement code to verify that the AI’s filtered results are correct and present in the original list.
- Users can always access the full, unfiltered list from your chosen ICD-10 source.
Enjoy!
This is an open-source project maintained by Vinta Software. We are always looking for exciting work! If you need any commercial support, feel free to get in touch: contact@vinta.com.br