採用遠距照護之決策樹分類-以臺灣遠距照護系統為例

Abstract

在臺灣，政府必須發展一套系統性的老人福利政策以因應高齡化社會的來臨。透過 資通訊技術與健康保險的幫助，遠距照護成為協助高齡化社會的一種可能方案。然而遠 距照護除了使用成本仍居高不下外，其適用的人數也相對稀少，因此我們必須要能找出 那些少數適合遠距照護服務的民眾。 本篇研究中，我們提出一個使用決策樹做為主要資料探勘的啟發式模型(HDTTCA)， 用來找出符合適用遠距照護民眾的條件。我們假設遠距照護服務能被納入在台灣特有的 健康保險制度(全民健康保險)裡，因此必須設計出一套找出全臺兩千三百萬人中適用的 民眾的演算法以降低人工分類的負擔。 由於決策樹的易讀性讓民眾能明確地了解適用遠距照護的條件，也提供政府在擬定 相關遠距照護政策時參考的規則。在這個模型中共有三個步驟：資料前處理、建構決策 樹、預測並解釋分類結果。本篇研究也找出三個在遠距照護產業影響甚大的關鍵因素： 醫療紀錄、就醫方便性與申請人的社經地位。

In Taiwan, it is top priority to develop a systematic policy for the elderly in the coming era of ageing society. Telecare is one possible solution that combine the help of information technology and health policy. However, the telecare service is expensive and is only adoptable for chronic citizens nowadays. This study proposes a Heuristic Decision Tree based Telecare Classification Approach (HDTTCA), a systematic structure to figure out those who are applicable for telecare service. We assume that National Health Insurance scheme subsidy the cost. Therefore, it is necessary to classify all the people in the country, we apply classification algorithm to reduce the research time. The main process of HDTTCA has three steps: (1) datasets preprocessing; (2) decision tree model building; and (3) prediction and explanation. In this study, we find out three of most important dimensions in telecare industry: clinical records, convenience, and social-economic status. We apply HDTTCA in a real world dataset and find that decision tree is a potential classification algorithm in solving this problem because of its interpretability. The see-through rules also give government a clear guideline to improve the decision-making process.