Routine use of genetic data in healthcare is much-discussed, yet little is known about its performance in epidemiological models including traditional risk factors. Using severe COVID-19 as an exemplar, we explore the integration of polygenic risk scores (PRS) into disease models alongside sociodemographic and clinical variables. PRS were optimized for 23 clinical variables and related traits previously-associated with severe COVID-19 in up to 450,449 UK Biobank participants, and tested in 9,560 individuals diagnosed in the pre-vaccination era. Associations were further adjusted for (i) sociodemographic and (ii) clinical variables. Pathway analyses of PRS were performed to improve biological understanding of disease. In univariate analyses, 17 PRS were associated with increased risk of severe COVID-19 and, of these, four remained associated with COVID-19 outcomes following adjustment for sociodemographic/clinical variables: hypertension PRS (OR = 1.1, 95%CI 1.03-1.18), atrial fibrillation PRS (OR = 1.12, 95%CI 1.03-1.22), peripheral vascular disease PRS (OR = 0.9, 95%CI 0.82-0.99), and Alzheimer's disease PRS (OR = 1.14, 95%CI 1.05-1.25). Pathway analyses revealed enrichment of genetic variants in pathways for cardiac muscle contraction (genes N = 5; beta[SE] = 3.48[0.60]; adjusted-P = 1.86 × 10-5). These findings underscore the potential for integrating genetic data into epidemiological models and highlight the advantages of utilizing multiple trait PRS rather than a single PRS for a specific outcome of interest.
© 2025. The Author(s).