SDTM Production: An AI-Driven Approach

SDTM datasets are essential to regulatory-ready clinical trial submissions. The traditional process of converting the vast range of data sources to the common SDTM format is time-intensive, inherently inefficient and costly.

While deterministic and rule-based mapping approaches using static raw data standards or similar machine-readable metadata represent the most precise and effective way to generate SDTM datasets, in many situations, standards are not as stable as needed, and machine-readable metadata information is unavailable.

In this session, we’ll present a novel approach for SDTM production to accelerate timelines and enhance quality that moves beyond rule-based mapping:

• AI-driven metadata extraction from "unstructured" protocol information and ancillary study documentation.
• Utilization of the machine-readable metadata extract to automate the generation of SDTM annotated CRFs using AI.
• Utilization of the machine-readable metadata extract to automate the generation of SDTM datasets using AI.