Hospital Admissions from SIH with healthbR

Overview

The SIH (Sistema de Informacoes Hospitalares) records all hospital admissions financed by the Brazilian public health system (SUS) through the Autorizacao de Internacao Hospitalar (AIH). It is managed by the Ministry of Health through DATASUS.

Feature	Details
Coverage	Per state (UF), all 27 states
Years	1992–2024 (1992–2007 only on the R2 mirror; `RACA_COR` only from 2008)
Granularity	Monthly (one file per UF/month)
Unit	One row per hospital admission (AIH)
Format	Parquet on the healthbr-data R2 mirror (default), .dbc files from DATASUS FTP (fallback)

Getting started

library(healthbR)
library(dplyr)

Check available years

sih_years()
sih_years(status = "all")

Module information

sih_info()

Downloading data

Basic download

# All months of 2022 for Acre
admissions <- sih_data(year = 2022, uf = "AC")

Specific months

# First semester only
admissions <- sih_data(year = 2022, uf = "SP", month = 1:6)

# Single month
admissions <- sih_data(year = 2022, uf = "SP", month = 3)

Filter by diagnosis

Use CID-10 code prefixes to filter the principal diagnosis:

# Acute myocardial infarction (I21)
mi <- sih_data(year = 2022, uf = "SP", diagnosis = "I21")

# All respiratory diseases (Chapter X)
respiratory <- sih_data(year = 2022, uf = "SP", diagnosis = "J")

# Dengue (A90-A91)
dengue_hosp <- sih_data(year = 2022, uf = "SP", diagnosis = "A9")

Select variables

admissions <- sih_data(
  year = 2022,
  uf = "SP",
  month = 1,
  vars = c("DIAG_PRINC", "DT_INTER", "DT_SAIDA", "SEXO",
           "MORTE", "MUNIC_RES", "VAL_TOT")
)

Key variables

Variable	Description
DIAG_PRINC	Principal diagnosis (CID-10)
DT_INTER	Admission date
DT_SAIDA	Discharge date
SEXO	Sex (0=Unknown, 1=Male, 3=Female)
NASC	Date of birth
MORTE	Hospital death (0=No, 1=Yes)
MUNIC_RES	Municipality of residence (IBGE code)
MUNIC_MOV	Municipality of hospitalization
VAL_TOT	Total AIH value (R$)
DIAS_PERM	Length of stay (days)
PROC_REA	Procedure performed (SIGTAP code)
UTI_MES_TO	ICU days

Note: Sex codes in SIH differ from SIM/SINASC (0=Unknown, 1=Male, 3=Female).

Data dictionary

sih_dictionary()
sih_dictionary("SEXO")
sih_dictionary("MORTE")

Explore variables

sih_variables()
sih_variables(search = "diag")
sih_variables(search = "valor")

Example: Hospital mortality by diagnosis chapter

admissions <- sih_data(year = 2022, uf = "SP", month = 1:6)

mortality <- admissions |>
  mutate(chapter = substr(DIAG_PRINC, 1, 1)) |>
  group_by(chapter) |>
  summarise(
    total = n(),
    deaths = sum(MORTE == "1", na.rm = TRUE),
    mortality_rate = deaths / total * 100
  ) |>
  arrange(desc(mortality_rate))

Example: Hospitalization costs

admissions <- sih_data(year = 2022, uf = "SP", month = 1)

costs <- admissions |>
  mutate(
    chapter = substr(DIAG_PRINC, 1, 1),
    cost = as.numeric(VAL_TOT)
  ) |>
  group_by(chapter) |>
  summarise(
    admissions = n(),
    total_cost = sum(cost, na.rm = TRUE),
    mean_cost = mean(cost, na.rm = TRUE)
  ) |>
  arrange(desc(total_cost))

Example: Seasonal patterns

# respiratory admissions across all months
resp <- sih_data(year = 2022, uf = "SP", diagnosis = "J")

seasonal <- resp |>
  count(month, name = "admissions") |>
  arrange(month)

Smart type parsing

# parsed types (default)
admissions <- sih_data(year = 2022, uf = "AC", month = 1)
class(admissions$DT_INTER)  # Date
class(admissions$VAL_TOT)   # double

# all character
admissions_raw <- sih_data(year = 2022, uf = "AC", month = 1, parse = FALSE)

Cache and lazy evaluation

sih_cache_status()
sih_clear_cache()

# lazy query
lazy <- sih_data(year = 2022, uf = "SP", lazy = TRUE)
lazy |>
  filter(MORTE == "1") |>
  select(DIAG_PRINC, DT_INTER, SEXO, MUNIC_RES) |>
  collect()