Chapter 10 Common use cases
10.1 Calculated versus manual vars
Many numeric variables can be entered in two ways:
calculated var: The user provides dates (e.g., date of birth, date of index cardiotoxicity), and the variable is automatically calculated from these dates (e.g. age as the difference between index cardiotoxicity date and date of birth).
manual var: The user could not provide dates, but had a free text field to input the value.
As a result, 2 variables contain the same data. They are not mutually exclusive, which means an observation can have a value for the 2 variables (e.g. the user provided both dates, which allows for automated computation in the calculated var, and also entered age in the manual var free text variable). There has to be a rule of thumb to choose which variable is to be used in the analysis. Here is ours:
Calculated vars are preferred over manual vars
This means that, for a single case:
If the calculated var is available, it will be retained.
Else, if the manual var is available, it will be retained.
Else, if none are available, the value is missing.
Here is an implementation of this simple logic into an R function
numvar_uni <- # Numeric variables unifier
function( # used to organize data entered from 2 variables, currently its a prioritization
var1, # a quasi quoted name of column from data. Usually, one is the calculated var, one is the manual var, var1 will be prioritize over var2
var2 # also a quasi quoted name
# underlying data.frame data argument is omitted
){
var1 <- rlang::enexpr(var1)
var2 <- rlang::enexpr(var2)
ex <- rlang::expr(dplyr::case_when(
!is.na(!!var1) ~ as.numeric(!!var1),
!is.na(!!var2) ~ as.numeric(!!var2),
TRUE ~ NA_real_
))
ex
}10.2 Calculated and manual vars identifiers
For a manual var, the associated calculated var has the
__csuffix.
Example:
| manual var | calculated var |
|---|---|
p_age (patient age, from instrument admin) |
p_age__c |
10.3 Free text variables
Users are often provided additional fields in the case their patient
falls out of the checkboxes. For example, a patient may have experienced
an auto-immune disease that is not listed in the p_ai_ vars.
In this case, the user can check the p_ai_other box. When this box is
checked, the p_ai_other__ft variable is displayed. It is a free text
field where the user can input additional data (e.g. Bullous
pemphigoid).
Free text vars have the same name as the branching logic displayer, with a
__ftsuffix.
At the moment, data from free text variables is not used to compute additional variables for 2 main reasons:
There are few data in these vars
Data quality checking requires additional time consuming ressources that cannot be applied to a general framework
We recommend you use free text variables only if their data closely match your research question.