臨床路徑之建立機制—應用資料採礦技術 研究生：李淑芬 指導教授：王偉華 博士 東海大學工業工程與經營資訊研究所 摘要 中央健保局為控制醫療費用上漲之範圍，逐步採用各項前瞻性支付制度，如：總額支付制度、論人計酬制、論病例計酬制等，迫使各醫療院所重視醫療資源及費用之有效運用。現階段管理目標為維持原有醫療品質、控制費用發生於合理範圍內及減少醫療資源的浪費。現今醫療院所多推行臨床路徑(Clinical Path)管理，以達成上述之管理目標。臨床路徑管理為標準化某種疾病與處置的診斷作業，讓病人由住院到出院者依此模式接受治療。確保醫療品質、減少合併症的發生、縮短住院日數，有效控制醫療成本及病人滿意度。 傳統臨床路徑制定方法，需要臨床上各領域專業人員共同參與。以大量統計性資料為參考，經由設計、選擇、評估等階段制定而成。制定過程中，需反覆以人力協商方式達成共識，使得路徑的發展相當耗費人力與時間。為此，本研究應用資料採礦技術，期望建立系統化臨床路徑之建立機制。藉以提高臨床路徑建立之效率、發展病患之適切療程，及提供予醫師臨床療程建議之分類器。研究過程可分為兩個部分：首先，尋找某疾病隱含多種療程的可能性；繼而，根據病患的特徵判別所屬療程類別。本研究之目的為 (1) 應用群集技術找出療程類別及其臨床路徑基礎藍圖。再由醫療專業人員進一步規劃完整路徑，達成資訊面與領域知識的整合；(2) 透過類神經分類器判別病患所屬療程類別。 本研究選取兩種疾病進行機制驗證，分別為「剖腹生產」與「小兒肺炎」。剖腹生產屬「論病例計酬案件」，藉此分析機制之正確性與可行性；再以小兒肺炎為例，評估施行於「非論病例計酬案件」之可行性。本研究之結果顯示，兩種疾病皆有滿意表現。剖腹生產案件，透過此機制得到一個臨床路徑，其分類器之準確率亦達98%；小兒肺炎案件，藉由機制找到二個臨床路徑，分類器可達77%之準確率。 A Study for Developing the Clinical Path — The Data Mining Approach Student: Shu-Fen Lee Advisor: Dr. Wei-Hua Andrew Wang Institute of Industrial Engineering and Enterprise Information Tunghai University Abstract Due to the increasing medical practicing cost, the Bureau of National Health Insurance (BNHI) decided to implement the concept of Prospective Payment System (PPS) to help the medical hospitals in planning and controlling their medical care cost and quality. Several payment system have been implemented and are scheduling to be implemented in these one or two years. These systems are trying to constraint the hospitals’ incomes and in a way to accelerate the cost and quality control efforts in the hospitals. In the PPS concept, the medical process design and control is a key way to guarantee the efforts and the results. Clinical path is one of the most important tool in making PPS functional and helping the BNHI and the hospitals to reach the expected goals. Traditional way in developing clinical paths is to organize some related medical teams. And, within the teams, the related medical experiences could be integrated and hopefully reach a common agreement in the medical treatments procedure. After that, some general clinical paths could be formed. However, this way wastes lots of time and energy in negotiations and argues and can hardly reach the medical practice agreements. In this research, a data mining approach is developed to provide a new way in developing the first and common-agreed, in the sense of data, blueprint of clinical path. A common blueprint is provided for the members in the medical teams to design the appropriate clinical paths for the patients. In this research, two difference examples have been selected as the test bed for the proposed method. One is the Caesarian operation, which is a case payment example. The other is the Pneumonia of Pediatrics in which no clinical path has been reported before. We have implemented the proposed method in these two examples and received the satisfied results. In the Caesarian operation, a clinical path is generated and the accurate rate of ANN classifier is above 98%. In the Pneumonia of Pediatrics example, two clinical paths are generated and the accurate rate of ANN classifier is above 77%. In the above examples, the generated clinical pathways have been approved by the domain experts as well.
