A computational framework for defining and validating reproducible phenotyping algorithms of 313 diseases in the UK Biobank
Ana Torralbo, Jonathan M. Davitte, Damien C. Croteau-Chonka, Cai R. Ytsma et al. (2025)
This publication is the first public output of the GSK-UCL Phenomics collaboration. My team at the UCL Institute of Informatics developed a methodology for identifying disease cohorts within electronic health records made available by the UK Biobank longitudinal study. These disease definitions utilise provider- and source-specific codes, and are an expansion of the definitions in the CALIBER study.
I began working in this collaboration in 2021 following my MSc in Health Data Science. Over the next four years, I generated disease algorithms, validated clinical codes, developed software to filter and analyse the disease cohorts, generate visualisations, and updated web portal pages of each disease.
The GSK-UCL Phenomics team at GSK Stevenage in September 2024.
Pictured, left to right: Jonathan Davitte, Cai Ytsma, Spiros Denaxas, Ana Torralbo, Julia Carrasco-Zanini, Natalie Fitzpatrick, Natalie Zelenka
Beyond the published work, over the course of this collaboration I led the generation of algorithms for diseases and conditions of specific interest to GSK, as well as developed complex algorithms for diseases that require the combination of lab results, prescription records, and clinical codes.
Another output of my work is the standalone methodology for harmonising primary care prescription records into a research ready dataset. Read more about this here.