Standardised and Reproducible Phenotyping Using Distributed Analytics and Tools in the Data Analysis and Real World...

Area of Science:

• Health Informatics
• Observational Studies
• Real-World Data Analysis

Background:

• Reliable disease phenotypes are crucial for accurate observational study findings.
• The Data Analysis and Real-World Interrogation Network (DARWIN EU) requires robust phenotyping methods.
• Standardized approaches are needed for generating phenotypes from real-world data.

Purpose of the Study:

• To present a reproducible framework for generating reliable and traceable disease phenotypes using real-world data.
• To facilitate phenotype generation within the DARWIN EU network.
• To demonstrate the framework's application for pancreatic cancer and systemic lupus erythematosus (SLE).

Main Methods:

• A 14-step phenotyping process based on a standard operating procedure was developed.
• Bespoke R packages were used for codelist generation and review.
• Real-world data mapped to the OMOP Common Data Model was utilized.

Main Results:

• Codelists for pancreatic cancer and SLE were successfully generated.
• Cohorts were created in six OMOP-mapped databases, showing comparable incidence/prevalence to literature.
• Generated cohorts exhibited characteristics consistent with clinical expectations.

Conclusions:

• The developed phenotyping process, using bespoke tools, enables comprehensive codelist generation and cohort analysis.
• Structured and reproducible phenotyping is essential for the reliability of observational studies, particularly for regulatory decision-making.
• This framework supports large-scale exploration of cohort characteristics.