目的：國內DRGs制度即將全面推行，對於超長住院人次較多的醫院將會帶來衝擊，本研究將探討哪些類型醫院及哪些專科較會發生超長住院情形。方法：本研究使用2002年全民健保急性一般住院資料，以住院日數超過90百分位爲概念，分別以超過全體住院人次90百分位（簡稱14日法）、超過同一DRG住院人次90百分位（簡稱DRGs法）、超過迴歸風險校正病人性別、年齡及DRGs後殘差的90百分位（簡稱迴歸法）三種方式定義超長住院，了解醫院及醫院各專科超長住院人次佔率表現，並使用複迴歸探討影響醫院超長住院人次佔率之相關因素。結果：計算各醫院住院病人有多少人次會超長住院，得到以「14日法」定義為7.4%，以「DRGs法」或「迴歸法」定義爲7.2%，三者結果相近；在變異係數部分，分析醫院間的變異情形，以「14日法」定義之164.5%最大，以「DRGs法」之112.1%最小；公立、有呼吸照護病床或一日內再住院率高之醫院其超長住院人次佔率顯著較高，醫院各專科別中以復健科之超長情形最爲嚴重，使用「DRGs法」或「迴歸法」的定義下，婦產科、兒科、耳鼻喉科及眼科之超長住院情形明顯提高，顯示此類專科亦需重視其住院日數之控制。結論：雖然迴歸法校正更多變項，但以「DRGs法」探討醫院超長住院情形在行政審查上仍較爲方便可行；超長住院人次較多的醫院或科別，應及早因應DRGs制度帶來的衝擊而加強住院日控制；建議政府應配套發展亞急性與急性後期照護，提供病人連續性照護，減少病患滯留醫院的情形。

Objectives: To investigate the characteristics of hospitals and departments which were associated with a higher rate of overstay in hospitals in Taiwan. Methods: The subjects were acute hospitals drawn from the Taiwan National Health Insurance (NHI) databases in 2002. The three different definitions of overstay are: 1) over the 90th percentile of hospital days (the 14 days method), 2) over the DRGs adjusted days of hospitalization (the DRGs method), and 3) over the regression risk adjusted days of hospitalization (the regression method). We used these definitions to determine the conditions for overstay in hospitals. Associated factors of hospital characteristics were investigated with multiple regression analysis. Results: The rate of overstay was 7.4% as defined by the 14 days method. This was similar to 7.2% as defined by both the DRGs and the regression methods. The highest coefficient of variation was 164.5% as defined by the 14 days method, while the lowest was 112.1% as defined by the DRGs method. With regard to hospital characteristics, public hospitals, having respiratory care beds, and a readmission rate within one day above 1.56% were associated with higher rates of overstay. The rate of overstay in rehabilitation medicine was higher than that in other departments. Utilizing the definitions of the DRGs method and the regression method, the rates of overstay in the departments of obstetrics and gynecology, pediatrics, otolaryngology and ophthalmology highlighted the importance of controlling length of stay (LOS). Conclusions: Although the regression method accounted for more variables, the DRGs method was more convenient and practical for administrative review. Hospitals and departments which have a higher rate of overstay need to enhance control of LOS. An appropriate subacute/post acute care system should be established and developed to ensure quality of continuous care and reduction in the LOS for acute patients.