Automatic knowledge extraction from EHRs
Abstract
Increasing efforts in the collection, standardization, and maintenance of large scale longitudinal elec- tronic health care records (EHRs) across the world provide a promising source of real world medical data with the potential of providing major novel insights of benefit both to specific individuals in the context of personalized medicine, as well as on the level of population-wide health care and policy. The present paper builds upon the existing and intensifying efforts at using machine learning to provide predictions on future diagnoses likely to be experienced by a particular individual based on the person’s existing diagnostic history. The specific model adopted as the baseline predictive framework is based on the concept of a binary diagnostic history vector representation of a patient’s diagnostic medical record. The technical novelty introduced herein concerns the manner in which transitions between diagnostic history vectors are learnt. We demonstrate that the proposed change prima fasciae enables greater learning specificity. We present a series of experiments which demon- strate the effectiveness of the proposed techniques, and which reveal novel insights regarding the most promising future research directions.
Citation
Vasiljeva , I & Arandelovic , O 2016 , Automatic knowledge extraction from EHRs . in IJCAI 2016 - Workshop on Knowledge Discovery in Healthcare Data : New York City, USA, 10 July 2016 . IJCAI 2016 - Workshop on Knowledge Discovery in Healthcare Data , New York , United States , 10/07/16 . < https://sites.google.com/site/ijcai2016kdhealth/accepted-papers > workshop
Publication
IJCAI 2016 - Workshop on Knowledge Discovery in Healthcare Data
Type
Conference item
Rights
© 2016, the Author(s). This work is made available online in accordance with the publisher’s policies. This is the author created, accepted version manuscript following peer review and may differ slightly from the final published version.
Collections
Items in the St Andrews Research Repository are protected by copyright, with all rights reserved, unless otherwise indicated.