SidneyBissoli/sipni-covid

SI-PNI — COVID-19 Vaccination Microdata (Brazil, 2021–present)

Individual-level COVID-19 vaccination records from Brazil's National Immunization Program (SI-PNI), redistributed as partitioned Apache Parquet for efficient analytical access. Each row represents one administered dose.

Part of the healthbr-data project — open redistribution of Brazilian public health data.

Summary

Item	Detail
Official source	OpenDATASUS / Ministry of Health
Temporal coverage	January 2021 – present
Geographic coverage	All 5,570 Brazilian municipalities
Granularity	Individual record (one row per administered dose)
Volume	608,311,394 records (publication of 2026-07-23)
Format	Apache Parquet, partitioned by ano/mes/uf
Data types	All fields stored as string (preserves leading zeros)
Update frequency	Monthly
License	CC-BY 4.0

Resumo em português

SI-PNI — Microdados de Vacinação COVID-19 (Brasil, 2021–presente)

Microdados individuais de vacinação contra COVID-19 do Sistema de Informação do Programa Nacional de Imunizações (SI-PNI), redistribuídos em formato Apache Parquet particionado para acesso analítico eficiente. Cada linha representa uma dose aplicada.

Item	Detalhe
Fonte oficial	OpenDATASUS / Ministério da Saúde
Cobertura temporal	Janeiro/2021 – presente
Cobertura geográfica	Todos os 5.570 municípios brasileiros
Granularidade	Registro individual (uma linha por dose aplicada)
Volume	608M+ registros
Formato	Apache Parquet, particionado por ano/mes/uf
Atualização	Mensal

Para documentação completa em português, consulte o repositório do projeto.

Data access

Data is hosted on Cloudflare R2 and accessed via S3-compatible API. The credentials below are read-only and intended for public use.

R (Arrow)

library(arrow)
library(dplyr)

Sys.setenv(
  AWS_ENDPOINT_URL      = "https://5c499208eebced4e34bd98ffa204f2fb.r2.cloudflarestorage.com",
  AWS_ACCESS_KEY_ID     = "28c72d4b3e1140fa468e367ae472b522",
  AWS_SECRET_ACCESS_KEY = "2937b2106736e2ba64e24e92f2be4e6c312bba3355586e41ce634b14c1482951",
  AWS_DEFAULT_REGION    = "auto"
)

ds <- open_dataset("s3://healthbr-data/sipni/covid/microdados/", format = "parquet")

# Example: COVID vaccines in São Paulo, March 2021
ds |>
  filter(ano == "2021", mes == "03", uf == "SP") |>
  count(vacina_nome) |>
  collect()

Python (PyArrow)

import pyarrow.dataset as pds
import pyarrow.fs as fs

s3 = fs.S3FileSystem(
    endpoint_override="https://5c499208eebced4e34bd98ffa204f2fb.r2.cloudflarestorage.com",
    access_key="28c72d4b3e1140fa468e367ae472b522",
    secret_key="2937b2106736e2ba64e24e92f2be4e6c312bba3355586e41ce634b14c1482951",
    region="auto"
)

dataset = pds.dataset(
    "healthbr-data/sipni/covid/microdados/",
    filesystem=s3,
    format="parquet",
    partitioning="hive"
)

table = dataset.to_table(
    filter=(pds.field("ano") == "2021") & (pds.field("uf") == "SP")
)
print(table.to_pandas().head())

Note: These credentials are read-only and safe to use in scripts. The bucket does not allow anonymous S3 access — credentials are required.

File structure

s3://healthbr-data/sipni/covid/microdados/
  README.md
  ano=2021/
    mes=01/
      uf=AC/
        part-0.parquet
      ...
  ano=2022/
    ...
  ano=_invalid/
    mes=.../
      uf=.../

Records with dates outside the expected range (2021–present) are placed in ano=_invalid/ rather than discarded, preserving all original data without polluting the partition structure.

Schema

The dataset contains 32 variables, all stored as string. Key variables include:

Variable	Description
vacina_dataaplicacao	Vaccination date
vacina_codigo	Vaccine code
vacina_nome	Vaccine name
vacina_descricao_dose	Dose description (1st, 2nd, booster, etc.)
estabelecimento_uf	State of the health facility
estabelecimento_municipio_codigo	IBGE municipality code
paciente_datanascimento	Patient date of birth
paciente_enumsexobiologico	Biological sex
paciente_racacor_codigo	Race/ethnicity code
vacina_fabricante_nome	Vaccine manufacturer
vacina_lote	Vaccine lot number
sistema_origem	Source system (PNI, e-SUS APS, VACIVIDA, etc.)

Source and processing

Original source: CSV files by state published by the Ministry of Health on the open-data portal (dadosabertos.saude.gov.br, formerly OpenDATASUS) and served from an S3 bucket: 27 states × 5 parts = 135 files, ~292 GB uncompressed. Each republication gets a new "publication hash" in the file names and the previous files disappear from S3; the current copy on R2 comes from the publication of 2026-07-23 (0656f017-…), fully reprocessed on 2026-08-18.

Processing: CSV → Parquet (via polars) → upload to R2 (via rclone). No transformations are applied — values are published exactly as provided by the Ministry of Health.

Note on CSV artifacts: Unlike the routine vaccination data (which has JSON available), COVID data exists only as CSV. The source CSVs quote every field, so leading zeros and codes are intact; the pipeline reads all columns as text (infer_schema_length=0) and applies no correction — IBGE/CNES codes, ZIP codes and dictionary-coded fields are published exactly as serialized by the Ministry.

Reproducibility & provenance

Every Parquet file written by pipeline version ≥ 1.1.0 carries a JSON provenance record in its schema metadata (key healthbr): source_url, source_file, source_etag (the OpenDATASUS S3 object hash — the source publishes no separate MD5), source_size_bytes, download_date, pipeline_script, pipeline_version and (≥ 1.2.0) git_commit. The same facts are recorded per partition in manifest.json (plus SHA-256 and record count of each output file) and per source file in data/controle_versao_covid.csv in the GitHub repository. Together they let anyone re-derive a partition from the Ministry's original file and the exact code commit, and verify that the copy they hold is intact. Every file carries a native record: the whole dataset was reprocessed from the Ministry's 2026-07-23 publication on 2026-08-18 with pipeline 1.2.1 (nothing from the 2026-03 bootstrap remains). Data are not versioned: the R2 copy is the latest publication; revisions by the Ministry replace files, and the history of what was published when lives in the version-control CSV's git log. Full policy and audit recipe: docs/policy-reproducibility-pt.md.

arrow::read_parquet("part-0.parquet", as_data_frame = FALSE)$metadata$healthbr

Known limitations

1. Government data, not ours. Errors in the original data are intentionally preserved. No cleaning or correction is applied.
2. Variable completeness. Many fields have optional reporting.
3. All fields are strings. Type casting must be done by the user.
4. Invalid dates. Records with vaccination dates outside 2021–present exist in the source data and are preserved in ano=_invalid/.
5. Does not include routine vaccination. Routine (non-COVID) vaccination data is in a separate dataset: sipni-microdados.

Citation

@misc{healthbrdata,
  author = {Sidney da Silva Bissoli},
  title  = {healthbr-data: Redistribution of Brazilian Public Health Data},
  year   = {2026},
  url    = {https://huggingface.co/datasets/SidneyBissoli/sipni-covid},
  note   = {Original source: Ministry of Health / OpenDATASUS}
}

Contact

• GitHub: https://github.com/SidneyBissoli
• Hugging Face: https://huggingface.co/SidneyBissoli
• E-mail: sbissoli76@gmail.com

---

Last updated: 2026-08-18 (full reprocessing from the 2026-07-23 publication)