What question did this study set out to answer?

The study aims to assess the performance of a BEHRT model for clinical prediction using UK Biobank data.

Extending BEHRT to UK Biobank: assessing transformer model performance in clinical prediction

View Full Paper

Ask AI

Bookmark

Authors

Evaluates transformer model performance in clinical prediction tasks, highlighting significant findings in long-term outcomes.

The study aims to assess the performance of a BEHRT model for clinical prediction using UK Biobank data.
Trained a BEHRT model on UK Biobank hospital data
Evaluated performance across four clinical prediction tasks
Assessed impacts of model size, medical terminology, and data split strategies
Compared predictions of next-visit diagnosis and long-term diagnosis up to five years
The large model outperformed the smaller model in long-term prediction tasks (AUROC = 0.874 vs 0.858 at 5 years)
Differences in performance were marginal for 6-month predictions
CALIBER terminology yielded higher average precision scores (0.773 vs 0.678 with ICD-10)
Model performance varies significantly based on vocabulary size and modeling choices

Yildiz et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark