In dosgillespie/hseclean: Health Survey Data Wrangling

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

hseclean

The package is usable but there are still bugs and further developments that are being worked through i.e. some code and documentation is still incomplete or in need of being refined. The code and documentation are still undergoing internal review by the analyst team.

Motivation

hseclean was created as part of a programme of work on the health economics of tobacco and alcohol at the School of Health and Related Research (ScHARR), The University of Sheffield. This programme is based around the construction of the Sheffield Tobacco and Alcohol Policy Model (STAPM), which aims to use comparable methodologies to evaluate the impacts of tobacco and alcohol policies, and investigate the consequences of clustering and interactions between tobacco and alcohol consumption behaviours.

The original motivation for hseclean was to standardised the way that the Health Survey for England (HSE) data were cleaned and prepared for our analyses and inputs to our decision-analytic models. The suite of functions within hseclean reads the data for each year since 2001, renames, organises and processes the variables that we use for our analyses. The package also includes functions to multiply impute missing data, and to summarise data considering survey design.

We have subsequently added functions to process the Scottish Health Survey (SHeS) into a form that matches our processing of the Health Survey for England.

Health Survey for England and Scottish Health Survey data are accessed via the UK Data Service. hseclean is designed to read the tab delimited files.

Usage

hseclean is a package for reading and cleaning the Health Survey for England and Scottish Health Survey data.

The inputs are the raw survey data files for each year.

The processes applied by the functions in hseclean give options to:

1. Read tobacco and alcohol related variables and the information on individual characteristics that we use in our analyses.
2. Clean alcohol consumption data, applying assumptions about beverage size and alcohol content.
3. Clean data on current smoking and smoking history.
4. Clean data on individual characteristics including age, sex, ethnicity, economic status, family, health and income.
5. Multiply impute missing data.
6. Summarise categorical variables using proportions, considering survey design.

The output of these processes is a cleaned dataset that is ready for further analysis. This dataset can be saved so that you don't need to run the cleaning processes in hseclean each time you want to use the cleaned data.

Installation

We would like to ask that since the code and documentation is still under development and is complex, that you consult with the authors before you use it.

Please cite the latest version of the package using:
"Duncan Gillespie, Laura Webster, Colin Angus and Alan Brennan (2020). hseclean: Health Survey Data Wrangling. R package version x.x.x. https://STAPM.github.io/hseclean/. DOI: 10.5281/zenodo.3748784"

---

Since you will be downloading and installing a source package, you might need to set your system up for building R packages:

It is a good idea to update R and all of your packages.

Mac OS: A convenient way to get the tools needed for compilation is to install Xcode Command Line Tools. Note that this is much smaller than full Xcode. In a shell, enter xcode-select --install. For installing almost anything else, consider using Homebrew.

Windows: Install Rtools. This is not an R package! It is “a collection of resources for building packages for R under Microsoft Windows, or for building R itself”. Go to https://cran.r-project.org/bin/windows/Rtools/ and install as instructed.

---

You can install the development version of hseclean from github with:

#install.packages("devtools")
devtools::install_github("STAPM/hseclean")

---

If there is an error with install_github(), one possible work-around is

1. Download the package "tarball" by copying this into your internet browser (making sure the numbers at the end indicate the latest version) e.g. https://github.com/STAPM/hseclean/tarball/0.3.4. When the window pops up, choose where to save the .tar.gz file.

2. Go to the Terminal window in R Studio (or a console window in Windows by searching for "cmd") and install the package from the downloaded file by typing R CMD INSTALL [file path].

---

Then load the package, and some other packages that are useful. Note that the code within hseclean uses the data.table::data.table() syntax.

# Load the package
library(hseclean)

# Other useful packages
library(dplyr) # for data manipulation and summary
library(magrittr) # for pipes
library(ggplot2) # for plotting

Getting started

To be able to download data from the UK Data Service, you will need to register with the UK Data Service website, which will enable you to request access to the datasets. Instructions on how to do this can be found here.

In ScHARR, all data is stored and processed according to the ScHARR Information Governance Policy. The HSE and SHeS data is stored in the university networked X-drive folder PR_Consumption_TA, which is accessible only to team members who are using data according to the purposes stated to the UK Data Service. No individual-level data is included within this package on Github.

Basic functionality

Reading the HSE data files

There are separate functions in hseclean to read each year of HSE data. You must specify the link to where the data is stored. The functions read in all variables related to tobacco and alcohol and selected socioeconomic and other descriptor variables.

test_2001 <- read_2001(
  root = "X:/",
  file = "ScHARR/PR_Consumption_TA/HSE/HSE 2001/UKDA-4628-tab/tab/hse01ai.tab"
)

---

hseclean contains separate functions for reading the survey data for each year, e.g. read_SHeS_2008().

Processing socioeconomic, demographic and health variables

There are separate functions that focus on processing a different theme of socioeconomic, demographic and health variables. See vignette("covariate_data"). Note that the order in which the cleaning functions are applied can matter - as some functions use variables that are cleaned by others.

library(magrittr)

temp <- read_2017(root = root_dir) %>%
  clean_age %>%
  clean_family %>%
  clean_demographic %>% 
  clean_education %>%
  clean_economic_status %>%
  clean_income %>%
  clean_health_and_bio

Alcohol data

Detailed description of how to clean the alcohol data are given in vignette("alcohol_data"). As an example, here is the workflow to plot the frquency of drinking among people who drank in 2017.

library(magrittr)
library(dplyr)
library(ggplot2)

# Frequency of drinking in 2017 among drinkers
root_dir <- "/Volumes/Shared/"
#root_dir <- "X:/"

read_2017(root = root_dir) %>%
  clean_age %>%
  clean_demographic %>%
  alc_drink_now_allages %>%
  filter(age < 90, age >= 8, drinks_now == "drinker") %>%
  group_by(imd_quintile, age_cat) %>% 
  summarise(av_freq = mean(drink_freq_7d, na.rm = T)) %>% 
  ggplot(aes(x = imd_quintile, y = av_freq, fill = age_cat)) +
  geom_bar(stat = "identity", position = "dodge") +
  theme_minimal() +
  ylab("average number of days drink in a week")

Clean all years of smoking and alcohol data

See vignette("smoking_data").

library(magrittr)

# Wrap the individual cleaning functions in another function for applying to each year

cleandata <- function(data) {

  data %<>%
    clean_age %>%
    clean_demographic %>% 
    clean_education %>%
    clean_economic_status %>%
    clean_family %>%
    clean_income %>%
    clean_health_and_bio %>%
    smk_status %>%
    smk_former %>%
    smk_life_history %>%
    smk_amount %>%
    alc_drink_now_allages %>%
    alc_weekmean_adult %>%
    alc_sevenday_adult %>%
    alc_sevenday_child %>%

    select_data(
      ages = 12:89,
      years = 2001:2017,

      # variables to retain
      keep_vars = c("wt_int", "psu", "cluster", "year", "quarter",
                    "age", "age_cat", "sex", "imd_quintile",
                    "ethnicity_4cat", "ethnicity_2cat",
                    "degree", "relationship_status", "employ2cat", "social_grade", "kids", "income5cat",
                    "nssec3_lab", "man_nonman", "activity_lstweek", "eduend4cat",

                    "hse_cancer", "hse_endocrine", "hse_heart", "hse_mental", "hse_nervous", "hse_eye", "hse_ear", "hse_respir", 
                    "hse_disgest", "hse_urinary", "hse_skin", "hse_muscskel", "hse_infect", "hse_blood",

                    "weight", "height", "bmi",

                    "cig_smoker_status", "years_since_quit", "years_reg_smoker", "cig_ever",
                    "cigs_per_day", "smoker_cat", "banded_consumption", "cig_type", "time_to_first_cig",
                    "smk_start_age", "smk_stop_age", "censor_age", "giveup_smk",

                    "drinks_now", 
                    "drink_freq_7d", "n_days_drink", "peakday", "binge_cat",
                    "beer_units", "wine_units", "spirit_units", "rtd_units", 
                    "weekmean", 
                    "perc_spirit_units", "perc_wine_units", "perc_rtd_units", "perc_beer_units", 
                    "drinker_cat", 
                    "spirits_pref_cat", "wine_pref_cat", "rtd_pref_cat", "beer_pref_cat", 
                    "total_units7_ch"
      ),

      # The variables that must have complete cases
      complete_vars = c("age", "sex", "year", "quarter", "psu", "cluster")
    )

  return(data)
}

# Read and clean each year of data and bind them together in one big dataset
data <- combine_years(list(
  cleandata(read_2001(root = root_dir)),
  cleandata(read_2002(root = root_dir)),
  cleandata(read_2003(root = root_dir)),
  cleandata(read_2004(root = root_dir)),
  cleandata(read_2005(root = root_dir)),
  cleandata(read_2006(root = root_dir)),
  cleandata(read_2007(root = root_dir)),
  cleandata(read_2008(root = root_dir)),
  cleandata(read_2009(root = root_dir)),
  cleandata(read_2010(root = root_dir)),
  cleandata(read_2011(root = root_dir)),
  cleandata(read_2012(root = root_dir)),
  cleandata(read_2013(root = root_dir)),
  cleandata(read_2014(root = root_dir)),
  cleandata(read_2015(root = root_dir)),
  cleandata(read_2016(root = root_dir)),
  cleandata(read_2017(root = root_dir))
))

# clean the survey weights
data <- clean_surveyweights(data)

Summarise data

The function survey::svyby() in the survey R package is used by the function prop_summary() in hseclean to estimate the uncertainty around proportions calculated from a binary variable - prop_summary() was designed to simplify the process of estimating smoking prevalence from the HSE data, stratified by a specified set of variables.

prop_smokers <- prop_summary(
  data = hse_data,
  var_name = "smk.state",
  levels_1 = "current",
  levels_0 = c("former", "never"),
  strat_vars = c("year", "sex", "imd_quintile")
)

Missing data imputation

hseclean uses the function mice::mice() in the mice R package, implemented in a basic way by the impute_data_mice() function. See vignette("missing_data").

# Run the imputation (takes a long time)
imp <- impute_data_mice(data = hse_data,
                        var_names = c("smk.state", "agegroup", "sex", 
                                      "imd_quintile", "degree", "kids", "income5cat",
                                      "relationship_status", "employ2cat", "social_grade"),
                        var_methods = c("", "", "", 
                                        "polr", "logreg", "polr", "polr",
                                        "polyreg", "logreg", "logreg"),
                        n_imputations = 5)

# imp$data is a single data.table containing all 5 imputed versions of the data
hse_data_imputed <- copy(imp$data)

dosgillespie/hseclean documentation built on May 2, 2020, 1:15 a.m.