Skip to contents

Provides a reproducible workflow for DataSUS microdata: fetch_datasus() discovers, downloads, and combines published DBC files; read_dbc() reads a local DBC file directly, with structural and checksum validation; and the process_*() functions interpret official TabWin DEF/CNV/DBF dictionaries to produce period-correct values, labels, and column types.

Details

Start with fetch_datasus() and then use the processor corresponding to the selected system, such as process_sim(), process_sinasc(), process_sih(), process_sia(), process_cnes(), or process_sinan(). fetch_cadger() and fetch_sigtab() also retrieve current auxiliary tables for standalone use.

Use datasus_variables() to inspect parsed dictionary definitions and relations, validate_datasus_schema() to compare raw DBC fields with the selected historical definitions and processed types, and processing_diagnostics() to retrieve unknown codes, coercion failures, and dictionary provenance from an individual processing call.

Processing automatically reuses repeated date values and vectorizes code padding and CNV threshold lookup. For large requests, fetch_datasus() can process and save one file at a time with process = TRUE, collect = FALSE, and destination. Configure microdatasus.cache_dir for persistent dictionary reuse by processors called directly; see datasus_cache_dir() and the performance section of process_sim().

Network access and local data

fetch_datasus() needs access to DataSUS directory listings even when DBC files are cached. By default it warns about connection failures and retains successful downloads; with collect = TRUE, check for NULL before passing the result to a processor. Use stop_on_error = TRUE to abort on failures. A direct call to a processor or dictionary helper can fail if a required dictionary cannot be downloaded and no valid cached copy is available.

read_dbc() reads local files without Internet access. Packaged reference tables, including tabMun, are also available offline. Saved processing results can be reopened with readRDS(). Persistent dictionaries are reused when valid and refresh = FALSE; see fetch_tabwin_dictionary().

References

For concepts, coverage, data flows, and caveats of each Brazilian health information system, see Saldanha (2026), Sistemas de Informação em Saúde no Brasil.

Author

Maintainer: Raphael Saldanha raphael.saldanha@fiocruz.br (ORCID)

Authors:

Other contributors:

  • Sidney da Silva Pereira Bissoli (DBC decompression code) [contributor, copyright holder]

  • Mark Adler (blast decompressor) [contributor, copyright holder]