Brazil can lead digital health. The path starts with data.
The United States standardized how systems exchange a minimum set of data. Brazil has the chance to go further: record every clinical observation in a structured way, with the rule it came from, and build the data foundation that health AI needs.
Talk to our team See protocols in the recordWhy Brazil, and why now
A public system for more than 200 million people
SUS is the largest universal public health system in the world, with primary care, specialists, hospitals, vaccines and surveillance under the same national rules.
FHIR to exchange, a Brazilian standard to record
The RNDS already uses the open HL7 FHIR standard for data exchange, with profiles defined by Brazil. What is still missing is a Brazilian standard for clinical content: what each visit records. In the United States, the US government decides this through the USCDI. Brazil can define its own, with SUS protocols and more depth.
There is funding and a strategy
The Brazilian Artificial Intelligence Plan foresees up to R$ 23 billion through 2028, with actions for SUS. What is missing is structured, reliable clinical data so these projects work outside the lab.
What the United States built
Since the 21st Century Cures Act of 2016, the United States has required certified record systems to open their data through a standardized API. The result is the largest health data exchange effort in the world, and it teaches a lot.
ONC criterion (g)(10)
To be certified, a record system must offer a FHIR R4 API with the US Core profile, app sign-in through SMART on FHIR and bulk export of data for groups of patients.
USCDI: the minimum set
The USCDI defines which data classes the API must expose, such as allergies, problems, medications, vital signs and lab results. Version 3 took effect on January 1, 2026.
Inferno and the CHPL
Inferno is the official tool for testing the API. Products that pass go on the public list of certified systems (CHPL), which hospitals and governments check before buying.
TEFCA: the national network
A network of networks connects certified systems. In 2026, the US government reported more than 1 billion health records exchanged through it.
The lesson: the United States decided what is recorded across the country, and that decision stopped at the minimum set. Brazil should decide its own, and can decide with more depth: which questions the doctor answers, why, and under which rule.
The minimum to exchange data, and the complete record to provide care
AENTE delivers data in FHIR, as the standard requires. The difference is in what exists before the exchange: how each data point is born inside the visit.
| Topic | (g)(10) and USCDI standard | AENTE |
|---|---|---|
| Exchange format | FHIR R4 API, US Core, SMART on FHIR and bulk export. | A FHIR R4 API certified in the ONC program through Drummond Group, and CMD documents generated in the RNDS Brazilian profiles. |
| Management model | Outside the standard: each system organizes facilities, professionals and patients its own way. | One account per organization, facilities created from CNES, one identity for each patient and professional at any facility, and rules that follow the facility and the country. Patent-pending technology. |
| Depth | A minimum set of data classes, the same for everyone. | Every question in the record is a structured item, with a defined answer type and options, for each procedure. |
| Where the question comes from | Not part of the standard. | Every question names the rule, protocol or guideline it comes from, and doctors review the form before it goes live. |
| Context of the visit | Depends on how each system records it. | Data stays linked to the procedure, the diagnosis (ICD-10), the professional, the facility (CNES) and the payer. |
| Completeness | The standard requires no clinical fields beyond the minimum. | Required questions must be answered before the visit can be finished. |
| National rules | US terminologies. | ICD-10, SIGTAP, TUSS, CNES and the Ministry of Health protocols (PCDT) built into the care workflow. |
Forms that become a national dictionary
In the AENTE record, clinical documentation is built from two pieces. Form items are the questions and measurements, such as blood pressure, anticoagulant use or bowel prep scale. Form categories group those items into tabs and sections and link the set to procedures and diagnoses.
- Each item is created once and reused in every form that needs it. The same question means the same thing at any facility.
- Each item has a defined answer type, such as number, date, single or multiple choice, and closed options when the protocol requires them.
- Categories are linked to the procedure and the diagnosis. When the ICD-10 code is recorded, the Ministry protocol form appears ready.
- Everything is configured without code, by the clinical team, and changes are recorded.
Why this matters for artificial intelligence
Health AI models depend less on new algorithms and more on well-defined data. Free text changes from doctor to doctor and hospital to hospital. Structured items create a common vocabulary.
A dictionary of observations
Each form item is a variable with a name, a type and possible values. Added up at national scale, they form a dictionary of clinical observations that models can read without guessing what the doctor meant.
Observation linked to treatment
Because the record stays linked to the diagnosis, the procedure and what was done next, it becomes possible to study which treatments follow which findings, and with what results over time.
A trail back to the source
Knowing which protocol each question came from helps explain what a model learned, and correct it when a protocol changes.
Complete data from the start
Required questions reduce the gaps that today make much of the record data useless for research and training.
Training models with health data requires a legal basis, anonymization and governance, as Brazil's data protection law (LGPD) demands. The proposal is that this happens under public rules, with anonymized data and with the state coordinating. AENTE does not sell patient data.
What is already built
- A FHIR R4 API certified in the US ONC Health IT Certification Program, tested by Drummond Group: US Core profile, SMART on FHIR, granular permissions and bulk export.
- Minimum Data Set (CMD) documents generated in the RNDS FHIR profiles for every finished visit. Sending them to the RNDS depends on accreditation.
- Protocol forms (PCDT) that appear from the ICD-10 code, with the patient consent term ready.
- A national catalog with more than 300 service templates, with SIGTAP and TUSS codes, and facility setup from CNES.
- SUS and private health plan billing from the same clinical record.
- Patent-pending technology for health management with configurable workflows and rules per country.
How the government can take part
The proposal is to treat the forms dictionary as public infrastructure, and to test the model at real scale before expanding.
1. A pilot in a state or municipal network
A health department uses AENTE in a set of facilities, with targets for record completeness and CMD generation agreed before the start.
2. A Brazilian clinical content standard
The Ministry, medical societies and universities review the form items of priority protocols. The result is a Brazilian clinical content standard, built from the AENTE model and open to any system.
3. Data for SUS AI
Anonymized data from the pilot feeds SUS's own AI projects, under LGPD and the governance defined by the government.
The content of public protocols stays open as a public good. The platform that turns that content into care, billing and structured data belongs to AENTE, and the data always leaves in the RNDS FHIR standard, with no vendor lock-in. For the pilot, the instruments the law provides for innovative solutions are the Public Contract for Innovative Solutions (Complementary Law 182/2021) and technology procurement (Law 10.973/2004), plus funding calls linked to the Brazilian Artificial Intelligence Plan. Each agency makes the choice.
Frequently asked questions
Does AENTE replace e-SUS or the RNDS?
No. AENTE generates documents in the RNDS standard and feeds the national network. The goal is for more quality data to reach the Ministry, not to create a parallel system.
Is AENTE certified in the United States? And with the RNDS?
Yes, in the United States: AENTE's FHIR API received certification in the ONC program, with testing by Drummond Group. For the RNDS, AENTE already generates CMD documents in the official profiles, and sending them depends on accreditation and a company digital certificate, steps that are part of the pilot.
Who owns the forms dictionary?
The proposal is that the items of public protocols become a public good, reviewed by the Ministry and specialists and available to any system.
Does AENTE train AI on patient data?
No. Any use of data to train models must follow the LGPD, with anonymization and rules set by the public authorities.
Where is the data kept?
On AENTE servers, with role-based access and an audit log of changes. In a public pilot, hosting is decided with the agency.
Let's build this standard together.
We present the platform to your technical team, with protocol forms, the FHIR API and CMD documents working, and design a pilot with clear targets.