智慧共同照護之實現: 以資料驅動為基礎之 AI 糖尿病個案管理模式;Smart Shared Care:Data-driven AI-based Diabetes Case Management Model

NCUIR > School of Management at National Central University > Graduate Institute of Information Management > Electronic Thesis & Dissertation > Item 987654321/93237

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/93237

Title:	智慧共同照護之實現: 以資料驅動為基礎之 AI 糖尿病個案管理模式;Smart Shared Care:Data-driven AI-based Diabetes Case Management Model
Authors:	李侑霖;Li, Yu-Lin
Contributors:	資訊管理學系
Keywords:	機器學習;深度學習;糖尿病;共同照護;智慧醫療;醫療資訊管理;Machine Learning;Deep Learning;Diabetes Mellitus;Co-Care;Smart Healthcare;Healthcare Information Management
Date:	2023-07-20
Issue Date:	2024-09-19 16:50:00 (UTC+8)
Publisher:	國立中央大學
Abstract:	近年生活習慣與飲食型態的改變，糖尿病的患病人口數逐年上升，且糖尿病是無法治癒的。現今台灣政府積極推廣「糖尿病共同照護網」的積極支援性照護模式，以往文獻指出共同照護的互相支持確能有效減少糖尿病併發症的發生，雖立意良好，但目前我國成效不佳。探究原因可能是廣泛性的收案照護，使得照護模式下的品質不若想像中良好。因此，若以病人為中心做考量，來了解其是否適宜加入共同照護網，才能真正精準地用有效的資源幫助需要幫助的糖尿病人。本研究共設計了三階段的實驗，實驗方法採用機器學習及深度學習中的五種分類器。第一階段訓練並預測不同糖尿病人血糖控制是否穩定;第二階段訓練並預測不同糖尿病人是否需接受積極照護;第三階段則是利用前兩階段模型之重要特徵變數，預測不同糖尿病人是否需加入糖尿病共同照護網。於實驗最後，也會以評估指標，來評估本研究所產出之模型效能。研究結果指出，最佳預測血糖穩定性模型為 RF，準確率達 90%;最佳預測糖尿病人積極治療必要性模型為 XGBoost，準確率 73%;最佳預測糖尿病人加入糖尿病共同照護網必要性模型為 Random Forest，準確率 86%。藉由本研究，可讓臨床端精準地判斷病人加入糖尿病共同照護網的必要性，並針對病人血糖控制穩定性、藥物或胰島素積極治療必要性提供照護的建議指引，以建構完整的智慧共同照護模式。;With the change of people′s living and eating habits in recent years, the number of people with diabetes are increasing year by year, and diabetes cannot be cured. Nowadays, Taiwan government is promoting the "Diabetes Shared Care Network", and previous studies have shown that "Diabetes Shared Care Network" can effectively reduce the occurrence of diabetes complications. However, the current effectiveness of this initiative in Taiwan is not satisfactory. The possible reason for this is the broad-based inclusion of patients, which may result in a lower quality of care than anticipated. Therefore, it is essential to consider the patient′s perspective to determine whether they are suitable for inclusion in the "Diabetes Shared Care Network". We designed a three-stage experiment. All experiments use five classifiers to construct the predictive models. In the first stage, the models are trained to predict the stability of blood glucose control in different individuals with diabetes. In the second stage, we use machine learning techniques to find out which diabetic patients need curative care. In the last stage, we use the results of the above two experiments to predict who should join the "Diabetes Shared Care Network". At the end of the experiments, the performance of the models developed in this study will be evaluated using various evaluation metrics. The research findings indicate that the Random Forest (RF) model achieved the highest accuracy of 90% in predicting blood glucose stability. The XGBoost model was found to be the best in predicting the necessity of curative care for individuals with diabetes, with an accuracy of 73%. Additionally, the Random Forest model performed the best in predicting the necessity of individuals joining the Diabetes Shared Care Network, with an accuracy of 86%. Through this study, we can make precise judgments regarding the necessity of patients joining the "Diabetes Shared Care Network". The findings also provide guidance on blood glucose control stability and the need for medication or insulin therapy, facilitating the development of a comprehensive intelligent shared care model.
Appears in Collections:	[Graduate Institute of Information Management] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	16	View/Open

社群 sharing

Loading...