Health Facility Registry with CNES

Overview

The CNES (Cadastro Nacional de Estabelecimentos de Saude) is Brazil’s national health facility registry, maintained by the Ministry of Health through DATASUS. It contains monthly snapshots of all registered health establishments, professionals, beds, equipment, and more.

The healthbR package provides access to CNES data from the DATASUS FTP:

Feature Details
File types 13 (establishments, beds, professionals, etc.)
Frequency Monthly (one file per type/UF/month)
Years 2005–2024 (final + preliminary)
Coverage Per state (27 UFs)
Format .dbc files, decompressed internally

File types

CNES data is organized into 13 file types:

Code Name Description
ST Estabelecimentos Health facility registry (default)
LT Leitos Hospital beds
PF Profissional Health professionals
DC Dados Complementares Supplementary facility data
EQ Equipamentos Health equipment
SR Servico Especializado Specialized services
HB Habilitacao Facility accreditations
EP Equipes Health teams
RC Regra Contratual Contractual rules
IN Incentivos Financial incentives
EE Estab. de Ensino Teaching facilities
EF Estab. Filantropico Philanthropic facilities
GM Gestao e Metas Management and targets

Getting started

library(healthbR)
library(dplyr)

Check available years

cnes_years()
#> [1] 2005 2006 ... 2023

cnes_years(status = "all")
#> [1] 2005 2006 ... 2023 2024

Module information

cnes_info()

Downloading data

Basic download (establishments)

# all establishments in Acre, January 2023
ac_jan <- cnes_data(year = 2023, month = 1, uf = "AC")
ac_jan

Hospital beds

leitos <- cnes_data(year = 2023, month = 1, uf = "AC", type = "LT")
leitos

Health professionals

prof <- cnes_data(year = 2023, month = 1, uf = "AC", type = "PF")
prof

Month parameter

The month parameter controls which monthly snapshots to download:

# single month
jan <- cnes_data(year = 2023, month = 1, uf = "AC")

# first semester
sem1 <- cnes_data(year = 2023, month = 1:6, uf = "AC")

# specific months
q1_q3 <- cnes_data(year = 2023, month = c(3, 6, 9), uf = "AC")

# all 12 months (default when month = NULL)
full_year <- cnes_data(year = 2023, uf = "AC")

Selecting variables

# only key variables (faster)
cnes_data(
  year = 2023, month = 1, uf = "AC",
  vars = c("CNES", "CODUFMUN", "TP_UNID", "VINC_SUS")
)

Key variables (ST type)

Variable Description
CNES Facility CNES code
CODUFMUN Municipality (UF + IBGE 6 digits)
TP_UNID Facility type (22 categories)
VINC_SUS SUS-linked (0=No, 1=Yes)
TP_GESTAO Management type (M=Municipal, E=State, D=Dual)
ESFERA_A Administrative sphere (1=Federal, 2=State, 3=Municipal, 4=Private)
TURNO_AT Operating hours
NIV_HIER Hierarchy level
ATV_AMBUL Outpatient care (0/1)
ATV_HOSP Hospital care (0/1)
ATV_URG Emergency care (0/1)
COMPETEN Reference period (YYYYMM)

Using the dictionary

# all coded variables
cnes_dictionary()

# facility types (22 categories)
cnes_dictionary("TP_UNID")

# administrative sphere
cnes_dictionary("ESFERA_A")

Joining dictionary labels

# get facility type labels
tp_unid_labels <- cnes_dictionary("TP_UNID") |>
  select(code, label)

# join to data
ac_facilities <- cnes_data(year = 2023, month = 1, uf = "AC") |>
  left_join(tp_unid_labels, by = c("TP_UNID" = "code")) |>
  rename(facility_type = label)

ac_facilities |>
  count(facility_type, sort = TRUE)

Example: SUS-linked facilities by type

ac <- cnes_data(year = 2023, month = 1, uf = "AC")

sus_by_type <- ac |>
  filter(VINC_SUS == "1") |>
  count(TP_UNID, sort = TRUE)

# add labels
tp_labels <- cnes_dictionary("TP_UNID") |>
  select(code, label)

sus_by_type |>
  left_join(tp_labels, by = c("TP_UNID" = "code"))

Example: hospital beds per capita

Combine CNES bed data with Census population:

# step 1: count beds by UF (December snapshot)
beds <- cnes_data(year = 2023, month = 12, type = "LT") |>
  group_by(uf_source) |>
  summarize(total_beds = n(), .groups = "drop")

# step 2: population from Census 2022
pop <- censo_populacao(year = 2022, territorial_level = "state")

# step 3: calculate beds per 1,000 inhabitants
# beds_rate <- beds |>
#   left_join(pop, by = ...) |>
#   mutate(beds_per_1000 = (total_beds / population) * 1000) |>
#   arrange(desc(beds_per_1000))

Example: tracking facility counts over time

# quarterly snapshots for Sao Paulo
sp_quarterly <- cnes_data(
  year = 2020:2023,
  month = c(3, 6, 9, 12),
  uf = "SP"
)

facility_trend <- sp_quarterly |>
  group_by(year, month) |>
  summarize(
    total = n(),
    sus_linked = sum(VINC_SUS == "1", na.rm = TRUE),
    .groups = "drop"
  ) |>
  arrange(year, month)

facility_trend

Download tips

CNES data is monthly and per-state, so full downloads can involve many files:

Use uf and month to limit downloads. Start with a single UF and month to explore the data before scaling up.

Smart type parsing

# parsed types (default)
ac <- cnes_data(year = 2023, month = 1, uf = "AC")
class(ac$COMPETEN)  # Date

# raw character columns
ac_raw <- cnes_data(year = 2023, month = 1, uf = "AC", parse = FALSE)

Cache management

Downloaded data is cached locally for faster future access:

# check cache status
cnes_cache_status()

# clear cache if needed
cnes_clear_cache()

If the arrow package is installed, data is cached in Parquet format. You can also use lazy evaluation:

# lazy query (requires arrow)
cnes_lazy <- cnes_data(year = 2023, uf = "AC", lazy = TRUE)
cnes_lazy |>
  filter(VINC_SUS == "1", month == 1L) |>
  select(CNES, CODUFMUN, TP_UNID) |>
  collect()

Additional resources