Docs are also available in Portuguese (PT)

An opinionated ETL (Extract, Transform, Load) pipeline designed to process Brazil's public healthcare data (DataSUS) from raw CSV files into analysis-ready datasets.

Brazil's SUS (Sistema Único de Saúde) provides extensive public health data, but it requires domain-specific preprocessing before analysis. This includes removing unnecessary columns, handling missing values, and optimizing data types. Manually scripting these transformations for each dataset is time-consuming and error-prone.

etlSUS automates the entire process. Simply specify the dataset, and the library handles downloading, transforming, and loading the data into a database and/or merging all files.

poetry add git+https://github.com/GOPAD-Datasus/etlSUS.git

from etlsus import pipeline


if __name__ == '__main__':
    pipeline(
        dataset='SINASC',  # Choose between 'SINASC' or 'SIM'
        data_dir='path/to/data/dir',
    )

• Simple Interface: Select your dataset (SINASC and SIM) and specify the base directory
• Automated Processing: Handles download, transformation, and loading automatically
• Optimized Transformations: Removes irrelevant columns and values while preserving analytical value
  • SIM Dictionary (EN) (PT)
  • SINASC Dictionary (EN) (PT)
• Multiple Output Formats:
  • Direct export to relational databases
  • Merged single file for multi-year analysis
  • Multiple files

After running the pipeline, your data directory will be organized as follows:

# Using data_dir = "./data"

./data
├── raw/                  # Downloaded CSV files
├── processed/            # Cleaned and transformed files
└── dataset.parquet.gzip  # (Optional) Merged file

• Supports only PARQUET output files.

LGNU | © GOPAD 2025