의약품 추천 시스템에서 전자 건강 기록(EHR) 그래프의 효과 분석

초록

The medication recommendation aims to recommend effective medications at a patient's current visit, utilizing Electronic Health Records (EHR) data. A number of medication recommender systems have been proposed to solve this problem. In this paper, we first point out that these methods just utilize the co-prescription relations between medications represented by the EHR graph without verifying their effect in recommending effective medications. Then, we compare and analyze the recommendation accuracies of various state-of-the-art medication recommender systems equipped with or without the EHR graph. Our experimental results show that utilizing the EHR graph may not necessarily enhance the accuracy of medication recommendation, rather incurring a negative effect. This work is meaningful since it helps existing and future medication recommender systems to provide more accurate medication recommendation for patients.

키워드