- Dr. CliniData
- Posts
- Splitting Domains in SDTMIG: The Secret That Experts Rarely Talk About!
Splitting Domains in SDTMIG: The Secret That Experts Rarely Talk About!
Discover the SDTMIG trick that experts rarely talk about – Simple rules for splitting domains in SDTMIG
Dr. CliniData
September 11, 2024
🌟 Hey, friends! 🌟
Welcome to this week’s episode of Dr. Clinidata! Hope your week was epic! 🚀
In clinical trials, when dealing with large datasets, it's sometimes necessary to split a domain into smaller, more manageable datasets. Here’s a quick overview of how and when you can split a domain, following the rules set by SDTM Implementation Guide (SDTMIG).
When Can a Domain Be Split?
Based on --CAT: A domain can be split according to the values in the category (--CAT) variable. However, remember, the --CAT value should never be left blank.
Findings About (FA) Domain: This domain can be split based on the values of the --OBJ variable. For example, findings about medical history (MH) records can be stored in the FAMH domain.
Key Rules for Splitting Domains:
Consistency in Domain Name: The value of DOMAIN should remain consistent across all split datasets, just as it would have if the domain wasn't split.
Variable Prefix: All variables requiring a domain prefix (e.g., --CAT, --TESTCD) must use the domain name as a prefix (e.g., QSCAT, VSTESTCD).
Unique --SEQ Variable: The --SEQ value should be unique within each USUBJID across all split datasets. Even if there are 100 records for a subject, each must have a unique --SEQ value.
Linking with Relationship Datasets: When linking relationship datasets like SUPPxx, FAxx, or RELREC to the parent domain, ensure that the IDVAR used (e.g., --SEQ, --SPID, --GRPID) allows smooth joining with the parent domain.
Variable Distribution Across Splits: Some permissible variables included in one split dataset don't need to be included in every split.
Naming the Split Dataset: You can use up to 4 characters to name a split dataset (e.g., QSPI for 'Questionnaire on Pain Intensity'). This naming approach helps in maintaining compatibility with supplemental qualifier datasets.
Supplemental Qualifiers for Split Domains: When splitting domains, even the supplemental qualifiers (SUPPxx) must be split. The naming follows the same 1 or 2 extra character approach (e.g., SUPPQSPI).
RELREC Relationships for Split Datasets: If a dataset-level relationship exists for a split domain, the RDOMAIN variable in RELREC may contain the full 4-character split dataset name instead of the usual 2-character domain code.
Reply