| Abstract: | 手術?除膽囊是根治急性膽囊炎和膽囊結石最佳方法，隨著醫療技術的進步，腹腔鏡膽囊切除術逐漸取代傳統的開腹式膽囊?除術，然而全民健康保險制度的開辦，住院醫療費用逐年上升，隨著總額預算支付制度全面實施後，醫院為達到節省成本，增加病床使用率，控制住院天數等目標，病患在未得到完全診療下出院，短時間內反覆住院，相對醫療資源的耗用增加且影響病患完善醫療的權益。目標：探討膽囊切除手術術後三十日內非計劃性住院之相關影響因素及應用資料探勘技術建立再住院之預測模式。方法：本研究以2005年「全民健康保險研究資?庫」承保抽樣歸人檔，執行膽囊切除手術為研究對象，以決策樹、邏輯斯迴歸及支援向量機3種分類技術，建構預測膽囊切除術後三十日內非計劃性再住院之分類預測模式，並進行效能評估，選取最佳效能之分類預測模型。實驗結果：對於膽囊切除手術術後三十日內非計劃性再住院的正確預測率，以邏輯斯迴歸、決策樹模式、支援向量機分別為85.4%、83.6%、79.4%。J48 unpruned tree (風險樹)中醫師專科年資、有無重大傷病、病患年齡及醫院層級共有四個變項，分析其重要因素，將患者劃分為六個節點。因此本研究應能協助建構適當的預測模型，而產生有效的篩選與評估決策建議。 結?：資?探勘技術能協助處?龐大的資?萃取出關鍵的資訊，並可建立預測模型以供醫療臨床使用，以補一般統計分析之不足，有效率地篩選與掌握術後再住院高危險群病患，不僅可減少就醫次數，減緩醫療人員負擔並讓病患獲得適當的醫療照護。
Excising gallbladder with surgery is the best method for completecuring Acute cholecystitis and Gallstone.With the advances iv medical technology. Laparoscopic cholecystectomy has replaced traditional Open cholecystectomy gradually. However, the system has started of National Health Insurance, the hospital charges of admission goes up year by year, with the Global Budges Payment has been implemented, hospital must save hospital cost, increase the usage rate of beds and control admission days. Some uncured patients have to discharge early, readmission in a few time. Relatively, that will be increase the wastage of medical resource and affect patients to get complete medical benefits.Objectives：Investigate after the surgery of Cholecystectomy within thirty days of unplanned admission relevant factors, and the application of data mining technology to establish predication model of readmission.Methods：In this study, in 2005 the "National Health Insurance Research Database" Insurance Beneficiaries who file their cases for Cholecystectomy for the study, the dicision tree, Logistic Regression and Support Vector Machine three kinds of classification techniques constructed to predict the classificatory predication model for after the surgery of Cholecystectomy within thirty days of unplanned admission, and implement the performance assessment, which selects the best classification performance predication model.Results：The correct predication rates for after the surgery of Cholecystectomy within thirty days unplanned admission, with Logistic Regression, the dicision tree and Support Vector Machine was 85.4%, 83.6%, 79.4%. There are four changing items in the unpruned tree(J48): working eyperience of doctor, catastrophic illbess or not, ages of patients, levels of hospitals to analyze the important factors, and devided patients into six nodes. Therefore, this study should be assist to constructe suitable predication model, and produce effective screening and evaluation of plicy recommendations.Conclusion：Data mining techniques can help deal with huge data, which extract critical information, and to make predication model for medical clinical use, can make up for general statistical analysis of deficiencies, that will efficiently filter after the surgery with readmission and grasp patients at high risk. Not only can reduce the number of medical treatments, reduce the workload of medical personnel and patients can obtain appropriate medical care. |
