全民健保實施以來,因支付制度的不適用、醫療資源分配不當及病患不當用藥習慣等原因,導致健保當局財務日趨惡化。如何針對以上種種可能導致財務問題的原因,作出適當的處置與改善,為目前健保當局極需思量的問題。 台灣自1993年正式邁入高齡化社會,人體隨著年齡的增長而日趨老化虛弱,各種直接或間接的疾病症狀因而產生,致使老年人口的健保診治費用成為健保費用支出的大宗。老年人常因骨質疏鬆、骨強度降低、骨脆性增加的情況,稍不小心發生意外而骨折,當中又以髖部骨折為常見問題。髖部骨折是老年醫療照護的重要議題。本研究以髖部骨折中最常見之股骨轉子間骨折為主要研究對象。 老年人常患有糖尿病、高血壓及心臟病等疾病,而這些疾病皆可能會影響股骨轉子間骨折診治,使得醫療費用較一般病患來的高。本研究希望能藉由病患臨床資料以醫療費用與住院天數為指標,分別以案例式推理及類神經網路建立預估模型,並分別評估模型績效,最後再與健保當局現今實施之醫療資源配置實例比較,探討股骨轉子間骨折手術的醫療資源分配與利用率。 本研究應用類神經網路與案例式推理以作為探討醫療資源分配之工具。研究結果顯示,預測住院天數部分,神經網路之均方根誤差為0.0518;案例式推理在誤差為一天時,準確率可達92.89%。預測醫療費用部分,案例式推理在誤差為2000元,率可達83.89%。
Ever since the implementation of National Health Insurance, the unsuitable on payment system, medical resource allocation and the medicine used, resulted in the aggravation of financial. How to handle and improve the above mentioned problems become an important issue for the authorities. As the age growth people getting weakness and ageing, also the direct or indirect disease arise, result in the tremendous diagnosis expense of health insurance for the elders. It is often for the elders to fracture because of osteoporosis, bone strength decrease and bone brittle increase. Intertrochanteric fracture of femur will be applied in this research. The elders often suffer from diseases such as the diabetes、high blood pressure and heart disease, and these diseases may influence the intertrochanteric fracture of femur treatment. This research plans to apply important factors that influence the intertrochanteric fracture of femur then based on the index of total medical expenditure and length of stay to establish the prediction model with case-based reasoning and neural network techniques, the performance of the model will be evaluated and compared with the real case of the Bureau of National Health Insurance to analyze the medical resource allocation and utilization of the intertrochanteric fracture of femur. In this study, we propose Neural Network and Case-Based Reasoning for assistance in medical resource planning. Results of prediction for length of stay showed that the classification of the RMSE of Neural Network is at 0.0518, and the accuracy with absolute tolerance at one day of Case-Based Reasoning reached 92.89%. And the results of expense, the accuracy with absolute tolerance at 2,000 dollars of Case-Based Reasoning reached 92.89%.