Contrasting Association Results between Existing PheWAS Phenotype Definition Methods and Five Validated Electronic...

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Phenome-Wide Association Studies (PheWAS) link genetic variation to health outcomes.
• Electronic health records (EHRs) and International Classification of Diseases (ICD) codes are common data sources for PheWAS.
• The accuracy of ICD code-derived phenotypes versus rigorously defined phenotypes in PheWAS requires investigation.

Purpose of the Study:

• To compare PheWAS results derived from high-quality, algorithmically defined phenotypes against those derived from International Classification of Diseases, Ninth Revision (ICD-9) codes.
• To assess the impact of different ICD-9 code abstraction levels (3-digit, 5-digit, PheWAS codes) on association study outcomes.

Main Methods:

• Developed high-quality algorithms for five distinct phenotypes.
• Conducted PheWAS using these "gold standard" phenotypes and over 4.6 million genetic variants (minor allele frequency > 0.01).
• Compared PheWAS results obtained using the gold standard phenotypes with results derived from 3-digit, 5-digit, and PheWAS code abstractions of ICD-9 codes.

Main Results:

• Identified similarities in patient populations across phenotyping methods for certain diseases.
• Observed differences in PheWAS results depending on the phenotyping strategy (high-quality algorithms vs. ICD-9 code abstractions).
• Highlighted the potential impact of phenotyping methodology on genetic association findings.

Conclusions:

• While patient cohorts may overlap, the choice of phenotyping method significantly influences PheWAS outcomes.
• High-throughput phenotyping using ICD-9 codes may yield different genetic associations compared to rigorously defined phenotypes.
• Further research is needed to optimize phenotyping strategies for robust EHR-based genetic association studies.