Creating Custom Datasets in SDTM: A Simplified Guide

🌟 Hey, friends! 🌟

I am sure that you have only heard that datasets/domains in SDTM are fixed. But that is partially true, my friend.

We can create a custom dataset/domain in SDTM data tabulation, but there are certain steps or rules that need to be followed.

Today we are going to learn how we can create a custom dataset in SDTM, so let’s dive into the topic.

Identify the Need for a Custom Domain:

• Check if an existing domain fits your data. Create a custom domain only if the data are unique and don’t fit existing domains.

Follow These Guidelines:

• Group Data: Use variables like --CAT, --SCAT to categorize data within the domain.

• Avoid Time-Based Domains: Represent all observations in one domain (e.g., CM for non-study medication).

• Use Existing Domains: Data collected on separate forms may fit into existing domains (e.g., QS domain).

• Hierarchical Data: Use domain pairs (e.g., MB/MS) and describe relationships in RELREC if data have parent-child relationships.

Choose or Create a Domain Model:

1. Find a Prototype:

• Look for a similar existing domain model or select the closest general observation class.

2. Select Variables:

• Include required identifiers (STUDYID, DOMAIN, USUBJID, --SEQ).

• Add topic and qualifier variables from the chosen observation class.

• Include timing variables.

3. Domain Code:

• Use a unique 2-character code not in the CDISC CT SDTM Domain Abbreviation codelist.

• Avoid codes like AD, AX, AP, SQ, and SA.

Define the Custom Domain/Dataset:

• Variable Order: Follow the order from the general observation class.

• Variable Labels: Use title case and make labels clear.

• Standard Variables: Ensure they align with standard domains.

• Define-XML: Describe the dataset.

• Supplemental Qualifier Dataset: Place any non-standard variables here.

SDTMIG v3.4

Got opinions? Click the button and share them. We totally can't wait to hear all about it! 👇

THANKS FOR SURVIVING! SEE YOU NEXT WEEK!