English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recent evidence shows that deep learning models trained on electronic healthrecords from millions of patients can deliver substantially more accuratepredictions of risk compared to their statistical counterparts. While thisprovides an important opportunity for improving clinical decision-making, thelack of interpretability is a major barrier to the incorporation of theseblack-box models in routine care, limiting their trustworthiness and preventingfurther hypothesis-testing investigations. In this study, we propose twomethods, namely, model distillation and variable selection, to untangle hiddenpatterns learned by an established deep learning model (BEHRT) for riskassociation identification. Due to the clinical importance and diversity ofheart failure as a phenotype, it was used to showcase the merits of theproposed methods. A cohort with 788,880 (8.3% incident heart failure) patientswas considered for the study. Model distillation identified 598 and 379diseases that were associated and dissociated with heart failure at thepopulation level, respectively. While the associations were broadly consistentwith prior knowledge, our method also highlighted several less appreciatedlinks that are worth further investigation. In addition to these importantpopulation-level insights, we developed an approach to individual-levelinterpretation to take account of varying manifestation of heart failure inclinical practice. This was achieved through variable selection by detecting aminimal set of encounters that can maximally preserve the accuracy ofprediction for individuals. Our proposed work provides a discovery-enablingtool to identify risk factors in both population and individual levels from adata-driven perspective. This helps to generate new hypotheses and guidesfurther investigations on causal links.

Risk factor identification for incident heart failure using neural network distillation and variable selection