library(EDCimport)
db1 = edc_example()
load_database(db1)
enrol$subjid %>% class()
#> [1] "integer"
enrol$subjid %>% head()
#> [1] 1 2 3 4 5 6
db2 = edc_example() %>%
edc_unify_subjid(preprocess=~paste0("#", .x))
load_database(db2)
enrol$subjid %>% class()
#> [1] "factor"
enrol$subjid %>% head()
#> [1] #1 #2 #3 #4 #5 #6
#> 50 Levels: #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 #14 #15 #16 #17 ... #50
#missing patients in table `ae`
ae$subjid %>%
forcats::fct_count() %>%
dplyr::filter(n==0)
#> # A tibble: 2 × 2
#> f n
#> <fct> <int>
#> 1 #35 0
#> 2 #37 0Introduction
Caution
Before reading this, you should read
vignette("reading")to learn how to import your database.
After importing your database with EDCimport, you end up with an edc_database object, which can be loaded to the global environment using load_database().
However, EDCimport provides a few functions to improve the database before loading it.
Harmonize Subject ID across the database
The Subject ID column, usually SUBJID for CDISC data, is the primary key, shared by almost all your datasets.
Using edc_unify_subjid(), you can harmonize this column across the whole database, so that it becomes a factor, consistant for all datasets. With preprocess, you can even customize it.
This is especially convenient for joining your data and checking for missing patients.
Note
If your SUBJID column is numeric and
preprocessis empty, SUBJID will be casted to numeric.
Clean dataset names
Is your database from a messy EDC software, filled with special characters or camelCase column names?
Fear not! With edc_clean_names() you can clean all dataset names at once.
By default, it converts names to lowercase letters, numbers, and underscores only. For this example, since edc_example() already provides clean column names, let’s convert all columns to uppercase:
library(EDCimport)
db = edc_example() %>%
edc_clean_names(toupper)
load_database(db)
names(enrol)
#> [1] "SUBJID" "AGE" "ENROL_DATE" "ARM" "CRFNAME"
#> [6] "CRFSTAT"Split some dataset to short+long
Note
This one is a bit more complex, but bear with me, I’ll try to make is understandable.
When a CRF form contains both repeated and non-repeated measures, the export usually duplicates the non-repeated measure.
This results in a “mixed” data format, combining both “long” and “short” structures. (You usually call the latter “wide”, but in this case it is not really.)
For example, in the dataset long_mixed from edc_example(), you have two long-format variables (one value per observation) and one wide-format variable (one value per subject).
head(long_mixed)
#> # A tibble: 6 × 6
#> SUBJID CRFNAME LONG1 LONG2 SHORT CRFSTAT
#> <int> <chr> <dbl> <dbl> <chr> <chr>
#> 1 1 both short and long data 1.33 11.0 B Complete
#> 2 1 both short and long data -0.869 10.9 B Complete
#> 3 2 both short and long data 0.0555 10.0 C Complete
#> 4 2 both short and long data 0.0491 10.1 C Incomplete
#> 5 3 both short and long data -0.578 9.28 D Complete
#> 6 3 both short and long data -0.999 9.80 D CompleteWith complex CRFs and lengthy forms, this mixed structure can complicate analysis, as repeated and non-repeated data may be unrelated.
With edc_split_mixed(), you can split this dataset into two, one short and one long:
db = edc_example() %>%
edc_split_mixed(long_mixed)
load_database(db)
head(long_mixed_short) #one row per subject
#> # A tibble: 6 × 3
#> subjid crfname short
#> <int> <chr> <chr>
#> 1 1 both short and long data B
#> 2 2 both short and long data C
#> 3 3 both short and long data D
#> 4 4 both short and long data E
#> 5 5 both short and long data F
#> 6 6 both short and long data G
head(long_mixed_long) #one row per observation
#> # A tibble: 6 × 4
#> subjid long1 long2 crfstat
#> <int> <dbl> <dbl> <chr>
#> 1 1 1.33 11.0 Complete
#> 2 1 -0.869 10.9 Complete
#> 3 2 0.0555 10.0 Complete
#> 4 2 0.0491 10.1 Incomplete
#> 5 3 -0.578 9.28 Complete
#> 6 3 -0.999 9.80 CompleteYou can combine!
Obviously, these functions can be piped to one another:
db = edc_example() %>%
edc_split_mixed(long_mixed) %>%
edc_unify_subjid(preprocess=~paste0("#", .x))%>%
edc_clean_names(toupper)
load_database(db)Don’t hesitate to submit a feature request if you think another function can be useful to others!