Analyzing Health Data from POF with healthbR

Overview

The POF (Pesquisa de Orçamentos Familiares) is a household budget survey conducted by IBGE that investigates household expenditures, living conditions, and nutritional profiles of the Brazilian population. It is conducted in partnership with the Ministry of Health.

The healthbR package provides access to POF microdata with a focus on health-related data:

Module Description Available editions
Food Security (EBIA) Brazilian Food Insecurity Scale 2017-2018
Food Consumption Detailed personal food intake 2008-2009, 2017-2018
Anthropometry Weight, height, BMI 2008-2009
Health Expenses Medications, insurance, consultations All editions

Getting started

library(healthbR)
library(dplyr)

Check available editions

pof_years()
#> [1] "2002-2003" "2008-2009" "2017-2018"

Survey information

Use pof_info() to see which health modules are available for each edition:

pof_info("2017-2018")

List available registers

Each POF edition contains multiple data registers. Use pof_registers() to see them:

# all registers
pof_registers("2017-2018")

# only health-related registers
pof_registers("2017-2018", health_only = TRUE)

Explore variables

Before downloading data, you can browse available variables:

# list all variables in the domicilio register
pof_variables("2017-2018", "domicilio")

# search for food security variables
pof_variables("2017-2018", search = "ebia")

# search for weight-related variables
pof_variables("2017-2018", "morador", search = "peso")

Food Security Analysis (EBIA)

The EBIA (Escala Brasileira de Insegurança Alimentar) is available in the 2017-2018 edition through the domicilio register. The variable V6199 contains the food security classification.

Download domicilio data

domicilio <- pof_data("2017-2018", "domicilio")

EBIA classification

The EBIA classifies households into four levels:

Code Classification
1 Food security
2 Mild food insecurity
3 Moderate food insecurity
4 Severe food insecurity

Create EBIA categories

domicilio <- domicilio |>
  mutate(
    ebia = factor(
      V6199,
      levels = 1:4,
      labels = c(
        "Food security",
        "Mild insecurity",
        "Moderate insecurity",
        "Severe insecurity"
      )
    )
  )

# frequency table
domicilio |>
  count(ebia) |>
  mutate(pct = n / sum(n) * 100)

Weighted estimates with survey design

For proper population estimates, use the survey design:

library(srvyr)

domicilio_svy <- pof_data("2017-2018", "domicilio", as_survey = TRUE)

# add EBIA categories
domicilio_svy <- domicilio_svy |>
  mutate(
    ebia = factor(
      V6199,
      levels = 1:4,
      labels = c(
        "Food security",
        "Mild insecurity",
        "Moderate insecurity",
        "Severe insecurity"
      )
    )
  )

# weighted prevalence
domicilio_svy |>
  group_by(ebia) |>
  summarize(
    prevalence = survey_mean(na.rm = TRUE, vartype = "ci"),
    n = unweighted(n())
  )

EBIA by region (UF)

# food insecurity by state
domicilio_svy |>
  group_by(UF, ebia) |>
  summarize(
    prevalence = survey_mean(na.rm = TRUE, vartype = "ci"),
    n = unweighted(n())
  ) |>
  filter(ebia == "Severe insecurity") |>
  arrange(desc(prevalence))

Food Consumption Analysis

The consumo_alimentar register contains detailed personal food intake data from a subsample. This data is available for the 2008-2009 and 2017-2018 editions.

Download food consumption data

consumo <- pof_data("2017-2018", "consumo_alimentar")

Key variables

Variable Description
V9001 Food item code
V9005 Amount consumed
V9007 Unit of measure
ENERGIA_KCAL Energy (kcal)
PROTEINA Protein (g)
CARBOIDRATO Carbohydrate (g)
LIPIDIO Total lipids (g)

Average caloric intake

# total daily caloric intake per person
consumo |>
  group_by(COD_UPA, NUM_DOM, NUM_UC, COD_INFORMANTE) |>
  summarize(
    total_kcal = sum(ENERGIA_KCAL, na.rm = TRUE),
    total_protein = sum(PROTEINA, na.rm = TRUE),
    total_carb = sum(CARBOIDRATO, na.rm = TRUE),
    total_fat = sum(LIPIDIO, na.rm = TRUE),
    .groups = "drop"
  ) |>
  summarize(
    mean_kcal = mean(total_kcal, na.rm = TRUE),
    mean_protein = mean(total_protein, na.rm = TRUE),
    mean_carb = mean(total_carb, na.rm = TRUE),
    mean_fat = mean(total_fat, na.rm = TRUE)
  )

Health Expenses

The despesa_individual register contains individual expenses, including health-related spending such as medications, health insurance, and medical consultations.

Download expense data

despesas <- pof_data("2017-2018", "despesa_individual")

Filter health expenses

Health-related expenses can be identified by product group codes:

# explore expense categories
despesas |>
  count(QUADRO) |>
  arrange(desc(n))

Combining registers

For many analyses you need to combine data from multiple registers. Use the household identifier variables (COD_UPA, NUM_DOM, NUM_UC) to merge:

# download morador (demographic data) and domicilio (household data)
morador <- pof_data("2017-2018", "morador")
domicilio <- pof_data("2017-2018", "domicilio")

# merge: add household-level EBIA to individual-level data
morador_ebia <- morador |>
  left_join(
    domicilio |> select(COD_UPA, NUM_DOM, NUM_UC, V6199),
    by = c("COD_UPA", "NUM_DOM", "NUM_UC")
  ) |>
  mutate(
    ebia = factor(
      V6199,
      levels = 1:4,
      labels = c(
        "Food security",
        "Mild insecurity",
        "Moderate insecurity",
        "Severe insecurity"
      )
    )
  )

# food insecurity by age group
morador_ebia |>
  mutate(age_group = cut(V0403, breaks = c(0, 5, 12, 18, 30, 60, Inf))) |>
  count(age_group, ebia) |>
  group_by(age_group) |>
  mutate(pct = n / sum(n) * 100)

Comparing editions

The POF has been conducted in different years, and data structure may vary. Use pof_info() to check what is available in each edition:

# check health modules by edition
pof_info("2017-2018")  # EBIA + food consumption
pof_info("2008-2009")  # anthropometry + food consumption
pof_info("2002-2003")  # expenses only

Cache management

POF data files are large. healthbR caches downloaded files locally so you only download once:

# check cached files
pof_cache_status()

# clear cache if needed
pof_clear_cache()

If the arrow package is installed, data is cached in Parquet format for faster loading:

# install arrow for optimized caching (recommended)
install.packages("arrow")

Additional resources