- 學位論文
應用馬氏田口系統於醫療急救之預測
Applying MTS to Cardiopulmonary Resuscitation Prediction
摘要
住院病人安全是醫院服務品質之重要指標，而院內心跳停止是醫療機構中常見且具高風險性的問題，會嚴重威脅住院病人安全。心肺復甦之成效關係到能否立即拯救危急瀕死之病患，與是否能提高病患急救後之出院存活率。院內臨床警示系統合併住院醫師交接班作業，可及早發現狀況不穩定之病患並給予醫療處置，能夠有效的減少院內心跳停止事件的發生。由於醫療領域資料多屬不平衡型態，而馬氏田口系統對於不平衡資料擁有較好的穩健性，且可透過篩選變項降低預測之成本與加快預測之速度，符合醫療急救的特性，因此本論文以MTS分析個案醫院之一般病房病患資料，以建立院內急救的臨床警示系統。本研究最終建立的臨床警示系統使用舒張壓、體溫、脈搏及呼吸作為變項，並以馬氏距離的臨界值作為系統的警示標準。另外，本研究也以其他常見的分類方法如類神經網路、決策樹以及邏輯斯迴歸進行分析，結果顯示MTS對於不平衡資料確實擁有較優越的分類能力。期盼本研究能夠拋轉引玉，使得MTS更廣泛地應用在醫療領域與不平衡資料問題中。
並列摘要
Safety of inpatients is an important index for service quality of hospital. In-Hospital Cardiac Arrest (IHCA) is a common and high risk problem for medical institutions. The effect of Cardio-Pulmonary Resuscitation (CPR) is related to if it could save dying patients immediately and if it could increase the survival rate of patients given first aid afterward. It could discover unstable patients early and provide them medical care if we combine clinical alert system (CAS) and residents' working shifts; also, it could reduce IHCA event. Because most of the data in medical field is imbalanced data type; in addition, there is a better robustness for imbalanced data in MTS, this paper aims to analyze the data of patients in ordinary ward of case study hospital by MTS method in order to build up CAS for IHCA. After we selected the system features: diastolic blood pressure, Temperature, Pulse rate and Respiratory rate, we used the Mahalanobis-distance to be the alert standard of system. Moreover, three commonly used classification method: artificial neural network, decision tree and logistic regression, are implemented using the same data set. The result showed that there is a better classification rate in MTS for imbalanced data type.