{"database":"biostudies-literature","file_versions":[],"scores":null,"additional":{"submitter":["Yildiz Y"],"funding":["National Institute for Health Research (NIHR)"],"pagination":["1715506"],"full_dataset_link":["https://www.ebi.ac.uk/biostudies/studies/S-EPMC12929515"],"repository":["biostudies-literature"],"omics_type":["Unknown"],"volume":["8"],"pubmed_abstract":["Introduction
Transformer-based models have shown strong potential for clinical prediction using electronic health record data, yet their performance can vary depending on modelling decisions and data characteristics.Methods
In this study, we trained a BEHRT model on hospital-based UK Biobank data and evaluated its performance across four clinical prediction tasks, including next-visit diagnosis and longer-term diagnosis prediction up to five years. We exhaustively assessed the impact of model size, medical terminology (CALIBER vs ICD-10), and data split strategies.Results
The large model consistently outperformed the smaller one in long-term prediction tasks (AUROC = 0.874 vs 0.858 at 5 years), while differences were marginal in 6-months prediction tasks. Performanc"],"journal":["Frontiers in digital health"],"pubmed_title":["Extending BEHRT to UK Biobank: assessing transformer model performance in clinical prediction."],"pmcid":["PMC12929515"],"funding_grant_id":["NIHR301413"],"pubmed_authors":["Jenkins DA","Yildiz Y","Nenadic G","Jani M"],"additional_accession":[]},"is_claimable":false,"name":"Extending BEHRT to UK Biobank: assessing transformer model performance in clinical prediction.","description":"Introduction
Transformer-based models have shown strong potential for clinical prediction using electronic health record data, yet their performance can vary depending on modelling decisions and data characteristics.Methods
In this study, we trained a BEHRT model on hospital-based UK Biobank data and evaluated its performance across four clinical prediction tasks, including next-visit diagnosis and longer-term diagnosis prediction up to five years. We exhaustively assessed the impact of model size, medical terminology (CALIBER vs ICD-10), and data split strategies.Results
The large model consistently outperformed the smaller one in long-term prediction tasks (AUROC = 0.874 vs 0.858 at 5 years), while differences were marginal in 6-months prediction tasks. Performanc","dates":{"release":"2026-01-01T00:00:00Z","publication":"2026","modification":"2026-07-16T22:08:08.881Z","creation":"2026-07-11T03:11:59.357Z"},"accession":"S-EPMC12929515","cross_references":{"pubmed":["41743675"],"doi":["10.3389/fdgth.2026.1715506"]}}