대한신장학회

Skip Navigation: Skip to contents

KOR
ENG
전체메뉴보기

간행물 검색

현재 페이지 경로

HOME
간행물
간행물 검색

논문분류	춘계학술대회 초록집
제목	Developing an AI-powered prescription algorithm for maintenance hemodialysis patients
저자	Gwangeon Sim
출판정보	2024; 2024(1):
키워드
초록	Objectives: Nephrologists are enduring excessive overtime and workload for monthly prescription of ESRD patients, raising significant concerns about burnout. Specifically, the critical treatment in hemodialysis centers involves monthly examinations and medication prescriptions for dialysis patients, typically guided by the KDIGO and K-DOQI guidelines. However, practices may vary among attending physicians, with prescriptions influenced by individual preferences, knowledge, and beliefs. In this study, we developed an artificial intelligence model using data from approximately 180,000 past prescriptions and assessed its applicability. Methods: The artificial intelligence-based medication prescription system for hemodialysis patients is comprised of a model trained using past blood test and prescription data of hemodialysis patients. This model is then utilized to prescribe medications based on the blood test results of hemodialysis patients (Figure 1). The flow of the study involves extracting and analyzing past blood test result data and corresponding prescription data for hemodialysis patients to use as training data for the AI model. Additionally, an analysis of prescription protocols is carried out to perform the initial work for AI model development (Figure 2). Once the AI model is finalized, the patient's blood test results upon arrival are stored in the database (database layer). These results are then fed into the previously developed AI model as input variables, recommending suitable drugs for the test results (functional layer). Finally, the attending physician follows the process of prescribing the appropriate drugs to the patient through the system interface (presentation layer). Results: The AI-based drug prescription model demonstrated exceptional performance, with scores of 0.99 or higher across all four performance indicators of the classification model according to 7 types of protocol including bicarbonate, erythropoetin, vitamin D25, iPTH, potassium and phosporus. Conclusions: This innovation is anticipated to significantly mitigate physician burnout by reducing it by less than 10%, thereby making a substantial contribution to the healthcare system.
원문(PDF)	PDF 원문보기

목록

위로가기