Make a ‘Choose One’ Table

library(tidyREDCap)


The Problem

It is often desirable to print variable labels above a summary table that shows the count of factor labels. The labels that are exported on choose all that apply questions include both the question and whichever response was chosen. This redundancy is often unwanted, and the results are not presented professionally.

For example, in the Nacho Craving Index data, the first ingredient is “Chips”. We see how R presents this information by simply printing the components of the ingredients___1 column.

redcap <- readRDS(file = "./redcap.rds")
redcap$ingredients___1
#  [1] Checked   Checked   Unchecked Unchecked Unchecked Unchecked Checked  
#  [8] Unchecked Unchecked Unchecked Unchecked Unchecked Checked   Unchecked
# [15] Unchecked Checked   Unchecked Unchecked Checked   Unchecked Unchecked
# [22] Unchecked Checked   Unchecked Checked   Unchecked Unchecked Unchecked
# [29] Unchecked Checked  
# attr(,"redcapLabels")
# [1] Unchecked Checked  
# attr(,"redcapLevels")
# [1] 0 1
# attr(,"label")
# [1] What ingredients do you currently crave?: Chips
# Levels: Unchecked Checked

As we can see, this information is quite ugly, so we want to tabulate the results instead. However, if we use the simple table() function to clean up this information, we lose the original question and the answer label for ingredients___1.

table(redcap$ingredients___1)
# 
# Unchecked   Checked 
#        21         9

We no longer know what the question was, or which “select all” option this information represents.

Aside: Loading REDCap Data into R

The redcapAPI package can be used to load data directly into R. To learn more about it, take a look here. Normally the code to automatically pull data with an API includes a person’s secret code “key”. Because I want to keep this hidden, I have hidden this API key in my user profile and the code below includes a call to Sys.getenv() to grab the key. To learn more about working with APIs, look here. Also notice that the data is saved using the saveRDS() function. REDCap data loaded with the API has the variable labels added as an extra attribute. To allow this vignette to run without sharing my secret key, I have saved the data to the package website.

rcon <- redcapAPI::redcapConnection(
  url = 'https://redcap.miami.edu/api/', 
  token = Sys.getenv("NCI_API_Key")
)

redcap <- redcapAPI::exportRecords(rcon)



The Solution

The make_choose_one_table() function can be used with a factor variable to tabulate the response while preserving the question and checked option context.

make_choose_one_table(redcap$ingredients___1) 
# What ingredients do you currently crave?: Chips
#   Response  n percent
#  Unchecked 21     70%
#    Checked  9     30%

Further, this output can be molded into a publication-ready table with a single additional function call.

make_choose_one_table(redcap$ingredients___1) %>% 
  knitr::kable()

What ingredients do you currently crave?: Chips

Response n percent
Unchecked 21 70%
Checked 9 30%

The subset option, if set to TRUE, will cause the function to remove the label’s text and only show the response option (i.e., not repeat the “What ingredients do you currently crave?” question).

make_choose_one_table(
  redcap$ingredients___2,
  subset = TRUE
) %>% 
  knitr::kable()

Yellow cheese

Response n percent
Unchecked 23 77%
Checked 7 23%

This function can also be used in an analysis pipeline with a data frame name and the name of the factor inside of that data frame. For example:

redcap %>% 
  make_choose_one_table(ingredients___3) %>% 
  knitr::kable()

What ingredients do you currently crave?: Orange cheese

Response n percent
Unchecked 27 90%
Checked 3 10%