MedicamentosTransparentes: Using AI to compare medicine prices in public procurement

The problem

Over 150 million people depend on Brazil’s universal healthcare system, which promises free access to essential medicines. But medicine procurement is fragmented, with thousands of local governments buying products across scattered systems, resulting in large price variations and inconsistent documentation. Without reliable information, public buyers can’t make smart decisions and citizens pay more.

At the technical level, the challenge was recognising which procurement records were for medicines. The second was comparing medicine prices. Brazil’s National Public Procurement Portal contains millions of purchasing records, but medicine purchases are often hidden in short, inconsistent free-text descriptions. Around 12 million purchase items were published on the portal in 2024 alone, with no reliable automated way to identify which items were medicines. Brazil's new procurement law also requires procuring entities to consult a “Health Price Bank”.

For a country where more than 15 million people lack access to essential medicines, this data gap matters.

The solution

We matched procurement item descriptions from the public procurement system to entries in the goods and services catalogue, using the open contracting data published in the National Public Procurement Portal (PNCP) and the Sistema de Catálogo de Materiais e Serviços do Governo Federal (CATMAT/CATSER) – Brazil’s federal catalogue.

In the PNCP, descriptions of goods and services purchased appear as free text in the item fields associated with tenders and awards. We used an LLM-based embedding model to associate each medicine-related PNCP item with a CATMAT entry (identified by its catalogue code, codeBr), based on the similarity between the descriptions, and then applied a similarity threshold to decide whether the item should be classified as a medicine.

The project is live at https://medicamentos.transparencia.org.br/

The impact

Public officials have started using the new platform Medicamentos Transparentes to create price comparisons that help them make better and fairer decisions. It also serves as the “price bank” reference for medicine purchases, as required by public procurement law.

At the technical level, in tests using 1,000 manually labelled medicines, the model achieved 98% accuracy in identifying whether an item was a medicine and 86% accuracy in predicting the national catalogue code.

Lessons learned

Better price comparison starts with knowing what is being bought. AI can help governments turn messy procurement text into decision-ready data, but it depends on good source data, a usable catalogue and validation against manually labelled examples.

Brazil’s Public Procurement Law 14.133/2021 mandated the creation of the National Public Procurement Portal. Procuring entities are also required to consult a Health Price Bank for reference prices when buying medicines and health products for the universal health system.