Extracting Severity Markers From Unstructured Clinical Data of Congestive Heart Failure Patients Using a Pretrained Text-To-Text Transfer Transformer Model
An OMNY Health™ Poster
A method of extracting severity markers from unstructured data could improve disease management and retrospective clinical research.
On May 15 – May 18, 2022, this poster was on display at ISPOR 2022.
Conclusions:
- These results suggest that T5 models are capable of extracting disease specific knowledge from clinical notes.
- Areas for further research include studying how variations in question
wording and note splitting can influence results, determining sensitivity and specificity of the T5 model, and studying ways to improve performance, sensitivity, and process automation. - Applications include incorporation into health economic models and
improving predictive accuracy for CHF readmission and mortality.