mmrech/icskg-br-processed

ICSKG-BR: Index of Cities' Smartness & Knowledge for Global Surgery — Brazil

Processed data layer for the ICSKG-BR longitudinal ecological panel study. Cross-references the IESE Cities in Motion Index (CIMI) urban development framework against Lancet Commission on Global Surgery (LCoGS) indicators across all 5,570 Brazilian municipalities (2015-2023).

This dataset is currently private during the BMJ Global Health pre-submission window. It will become public (CC-BY-4.0) upon publication.

Provenance

• Source code: https://github.com/matheus-rech/ICSKG
• Dataset version: v0.1.0
• Schema version: 1.1
• Generated at: 2026-04-07T15:59:23Z
• Source DuckDB: icskg_br_export.duckdb (sha256 473d5ad523dab966)
• Compression: zstd level 6

Cookbook scope (v0.1.0)

This release implements the LCoGS-side of the ICSKG-BR Technical Cookbook v1.0 (March 2026). The canonical panel/municipal_health.parquet is derived at publish time from the source tables via the cookbook §5 merge pipeline (50,130 rows = 5,570 mun × 9 years, 49 columns).

Cookbook §3 sources INCLUDED in v0.1.0:

• IBGE population master (§3.3)
• DATASUS SIH surgical aggregates (§3.1)
• DATASUS CNES SAO + bellwether (§3.2)
• IBGE SIDRA GDP (§3.3)
• Atlas Brasil HDI (§3.4)
• FIRJAN IFGF (§3.5)
• ANS TABNET insurance (§3.9)
• Census 2022 sanitation (§3.8 partial)
• Mobility (vehicle fleet)

DEFERRED to v0.2.0 (extractors not yet built — see project phase 12):

• ANATEL broadband (§3.6) — needed for CUDS Technology dimension
• RAIS employment (§3.7) — needed for CUDS Economy/Workforce dimension
• SNIS sanitation proper (§3.8) — Census 2022 is a partial proxy
• SIOPS health spending (§3.10) — needed for CUDS Governance dimension
• International comparators (§3.11) — WHO/World Bank/UNDP
• CUDS composite + dimension scores (§6) — blocked by missing sources above

The v0.2.0 release will add the missing 5 extractors and the cookbook §6 CUDS composite (PCA weighting + geometric mean aggregation), and will move the municipal_health derivation into the build pipeline so it's stored as a base table in the source DuckDB instead of being computed at publish time.

Totals

Metric	Value
Tables	23
Total rows	20,037,428
Compressed size	232.80 MB

Tables by namespace

panel/

Table	Rows	Columns	Size
municipal_health	50,130	49	2601.98 KB

lcogs/

Table	Rows	Columns	Size
lcogs1_access	5,571	6	43.50 KB
v_lcogs_latest	5,570	30	367.28 KB
v_lcogs_panel	50,130	30	2442.88 KB
v_lcogs_summary	49,717	26	1394.48 KB

source_tables/

Table	Rows	Columns	Size
ans_cobertura	5,594	6	60.58 KB
bellwether_hospitals	765	7	23.84 KB
censo2022_saneamento	5,570	4	47.88 KB
cnes_professionals_raw	218,461	6	129.47 KB
frota_veiculos	5,571	3	44.96 KB
gdp	50,130	4	418.16 KB
idhm	5,564	7	53.21 KB
idhm_historical	16,692	7	113.26 KB
ifgf	50,112	7	581.19 KB
mobility	5,535	3	35.73 KB
municipalities	5,571	12	154.55 KB
national_indicators	51	5	3.08 KB
pipeline_log	490	9	9.07 KB
population	50,130	3	207.30 KB
sao_workforce	49,753	8	383.22 KB
sih_municipal	50,170	8	442.42 KB
sih_raw	19,355,506	25	228820.14 KB
v_sao_annual	645	6	5.82 KB

Loading

from huggingface_hub import snapshot_download
import duckdb

local = snapshot_download(
    repo_id='mmrech/icskg-br-processed',
    repo_type='dataset',
    revision='v0.1.0',
    allow_patterns=['**/*.parquet', 'manifest.json'],
)

# Read the main panel directly:
panel = duckdb.sql(f"SELECT * FROM '{local}/panel/municipal_health.parquet'").df()

Citation

If you use this dataset, please cite the project:

Rech, Matheus M. (2026). ICSKG-BR: Index of Cities' Smartness & Knowledge
for Global Surgery — Brazil. https://github.com/matheus-rech/ICSKG

Production Zenodo DOI will be minted at BMJ acceptance.

License

Source data is public under Brazilian Lei de Acesso à Informação (Law 12.527/2011). This processed dataset is released under CC-BY-4.0 after BMJ Global Health publication.