醫療費用的支出不成比例地集中於老年人且每一位老年人的使用型態也有很大的差異,因此有必要找出醫療資源的高使用者,進而設法降低他們的使用量,因此本的研究主題是探討醫療使用的預測因子。我們收集了美國紐約州羅徹斯市某一家公司保險公司,在1988年對老年保險人所做的問卷調查及其後兩年(1988-1989)之醫療保險費用。有效樣本在1988年為3425人,1989年為3014人。將資料隨機分成兩組,一組做模式建立,一組做驗證組,利用複回歸方法計算依變項的R2及並觀察其相互間之關聯,且依變項為經對數轉換之年度醫療保險支出;而年度醫療支出的計算方式分為捨去法和推估法。結果發現本研究之推估法比捨去法有較高的R2(在1988年為16.4%比上15.11%,1989年為18.93%比18.11%)。而病患過去醫療的支出及現在所患的疾病和過去的醫療使用佔R2的90%。所以當我們試圖去預測醫療支出時,病患過去醫療的支出及現在所患的疾病和過去的醫療使用是主要探討對象,雖然本研究找出醫療使用因子的方法比其他方式有較高的R2,但R2仍小於20%。它意味著找出醫療使用因子不是那麼容易。
The elderly utilize a disproportionately large amount of health services in the U.S. It is also noted that the utilization expenditure pattern is not evenly distributed over aged propel. Hence, the possibility of finding the potential high users of health services emerges. This can help estimate ongoing health expenditures and possibly decrease utilization by high users. We collected enrollees of an Senior plan in Rochester NY who answered a survey in 1988 who have the claims data for 1988 and 1989.The effective sample size is 3425 for 1988 and 3014 for 1989. The data was randomly split into two sets, one for model building and one for validation. Multiple regression was used to calculate the R2 for the dependent variables. The dependent variable was the logarithmically transformed expenditures for a single year. The expenditures are calculated by two different methods. A validation procedure was applied to the regression model .Result of the study revealed the proposed extrapolated method has a consistently higher R2 than that of the other method(16.4% VS 15.11%1988,18.93% vs 18.11 %1989 ). Past expenditures, current diseases, and past utilization account for 90% of total R2. So When one designs a survey to predict health services expenditures, these three components (past expenditures, current diseases, and past utilization) should be considered first. Although the R2s in this study are higher than that of other studies, they are less than 20%.lt means that it will be not easy to find the potential high users.