Introduction to sdtmchecks

The purpose of the sdtmchecks package is to help detect and investigate potential analysis relevant issues in SDTM data. This is done using a set of data check functions. These check functions are intended to be generalizable, actionable, and meaningful for analysis.

Setting Up

To start using sdtmchecks first install it via

# install.packages("devtools")
devtools::install_github("pharmaverse/sdtmchecks", ref="main")

Then just load the package

library(sdtmchecks) 

Documentation

Here’s how to access the help page for the package

# type ??sdtmchecks into the console
??sdtmchecks 

Metadata

The package comes with the sdtmchecksmeta dataset which contains metadata on each check function. It contains details like function name, category, priority, and descriptions. Each function is given a Category (Cross Therapeutic Area, Oncology, Covid-19, Patient Reported Outcomes, Ophthalmology) and a Priority (High, Medium, Low).

#Just type this in
sdtmchecksmeta

## # A tibble: 10 × 5
##    check                              title            category priority domains
##    <chr>                              <chr>            <chr>    <chr>    <chr>  
##  1 check_ae_aeacn_ds_disctx_covid     COVID AE trt di… COVID    Low      ae, ds 
##  2 check_ae_aeacnoth                  AE AEACNOTH mul… ALL      Low      ae     
##  3 check_ae_aeacnoth_ds_disctx        AE AEACNOTx Dis… ALL      Low      ae, ds 
##  4 check_ae_aeacnoth_ds_stddisc_covid COVID AE study … COVID    Low      ae, ds 
##  5 check_ae_aedecod                   AE Missing PT    ALL      High     ae     
##  6 check_ae_aedthdtc_aesdth           AE Death Date v… ALL      High     ae     
##  7 check_ae_aedthdtc_ds_death         DS Death Dates … ALL      High     ae, ds 
##  8 check_ae_aelat                     AE AELAT Missing OPHTH    High     ae     
##  9 check_ae_aeout                     AE Death Outcome ALL      High     ae     
## 10 check_ae_aeout_aeendtc_aedthdtc    Fatal AE Resolu… ALL      High     ae

Running a Check

Let’s do an example using check_ae_ds_partial_death_dates(AE,DS)

This check flags records with partial death dates (i.e. length <10) in AE and DS. If any are found, then data check returns FALSE with attributes containing a list of flagged records as well as a brief message explaining the result. If no issues are detected the check returns TRUE.

# Use sample data frames.
AE

##   USUBJID AEDECOD   AEDTHDTC
## 1       1     AE1 2017-01-01
## 2       2     AE2       2017
## 3       3     AE3       <NA>

DS

##   USUBJID       DSSCAT DSDECOD    DSSTDTC
## 1       4 STUDY DISCON   DEATH 2018-01-01
## 2       5 STUDY DISCON   DEATH 2017-03-03
## 3       6 STUDY DISCON   DEATH 2018-01-02
## 4       7 STUDY DISCON   DEATH    2016-10

# Run the data check.
check_ae_ds_partial_death_dates(AE,DS)

## [1] FALSE
## attr(,"msg")
## [1] "There are 2 patients with partial death dates. "
## attr(,"data")
##   USUBJID       DSSCAT DSDECOD DSSTDTC AEDECOD AEDTHDTC
## 1       2         <NA>    <NA>    <NA>     AE2     2017
## 2       7 STUDY DISCON   DEATH 2016-10    <NA>     <NA>

Running Many Checks

Running all the checks on your data is super easy. Just use the run_all_checks function. This function assumes you have all of your sdtm datasets as objects in your global environment, e.g. ae,dm,ex,etc.

# Read data to your global environment
ae = haven::read_sas("path/to/ae.sas7bdat")
ds = haven::read_sas("path/to/ds.sas7bdat")

# Run the checks and save as an object called "myreport"
myreport=run_all_checks(metads = sdtmchecksmeta,
               priority = c("High", "Medium", "Low"), #subset checks based on priority
               type = c("ALL", "ONC", "COVID", "PRO", "OPHTH"), #subset checks based category
               verbose = TRUE)

class(myreport) #results in a list object
names(myreport) #each check result is saved in a slot of the list
myreport[["check_ae_aedecod"]] #investigate the results of a check

The run_all_checks function also lets you easily subset on category or priority

myreport=run_all_checks(metads = sdtmchecksmeta,
               priority = c("High"),
               type = c("ONC"),
               verbose = TRUE)

You can also choose specific checks to run. Here’s a way to get started with some checks that should work fairly well for most datasets

# Read data to your global environment
ae = haven::read_sas("path/to/ae.sas7bdat")
cm = haven::read_sas("path/to/cm.sas7bdat")
dm = haven::read_sas("path/to/dm.sas7bdat")

# Subset to checks that should work OK for most datasets
metads = sdtmchecksmeta %>%
  filter(check %in% c("check_ae_aedecod",
                      "check_ae_aetoxgr",
                      "check_ae_dup",
                      "check_cm_cmdecod",
                      "check_cm_missing_month",
                      "check_dm_age_missing",
                      "check_dm_usubjid_dup",
                      "check_dm_armcd"
                      ))

myreport=run_all_checks(metads = metads,
               verbose = TRUE)

Writing Out Results

You can then write results out to an xlsx for easy sharing.

report_to_xlsx(res=myreport,outfile="check_report.xlsx")

Making a Customizable Script

There’s also a convenient helper function to write out a user friendly R script with all the check function calls.

create_R_script(file = "run_the_checks.R")