應用全民健康保險研究資料庫比較不同合併症指標對中風病患的醫療費用之預測表現

Comparison of different comorbidity measures in predicting the medical expenditures of stroke patients by utilizing the National Health Insurance Research Database

目標：比較不同合併症指標對中風病患的醫療費用之預測表現。方法：全民健康保險研究資料庫2005年承保抽樣歸人檔(LHID2005)為主要資料來源，以首次中風住院患者為研究對象，比較四種合併症指標［Deyo-Charlson comorbidity index(CCI)、Romano-CCI、D`Hoore-CCI及Elixhauser方法］對患者出院後一年產生高醫療費用情形之預測力。本研究的基本模式包含病患年齡、性別、首次中風住院有無手術及首次中風住院天數變項，並且擷取2005年及2008年兩個不同年度之樣本資料進行比較分析。最後使用邏輯斯蒂迴歸分析(logistic regression analysis)的c值(c-statistics)，以檢定增加不同合併症指標後對模式預測力的表現差異。結果：四種合併症指標對於基本模式的預測力均有增進效果，其中又以Romano-CCI表現最佳的預測力（c值：0.710-0.746）。結論：本研究的研究結果顯示，合併症指標為中風住院病患出院後的醫療費用之重要預測因素，而Romano-CCI的預測力表現則最佳。建議衛生主管機關在擬定有關中風病患的醫療費用等相關給付制度時，應考量病患因合併症所造成之醫療照護負擔，而採取不同程度的調整。

Objectives: This study compared the performance of different co-morbidity measures in predicting medical expenditures of stroke patients. Methods: Data were sourced from the Longitudinal Health Insurance Database 2005 (LHID205), and the study population comprised all patients who were hospitalized due to stroke for the first time. Four co-morbidity measures were compared regarding the performance of predicting medical expenditures of subjects within 1 year after discharge: the Deyo-Charlson comorbidity index (CCI); Romano-CCI; D`Hoore-CCI; and Elixhauser method. The baseline model included patient age and gender, whether or not surgery was undertaken when hospitalized, and the length of stay. Two target years (2005 and 2008) of data were compared. The discriminatory power of the co-morbidity measures was assessed using the c-statistics derived from multiple logistic regression models. Results: All four co-morbidity measures significantly improved the predictive capacity of the baseline model. Furthermore, the Romano-CCI performed best in predicting medical expenditures of subjects within 1 year after discharge (c: 0.710-0.746). Conclusions: This study suggested that co-morbidity measures are significant predictors of medical expenditures of stroke patients, and the Romano-CCI performed best among the four co-morbidity measures in the research. When designing the payment schemes for stroke patients, the Taiwanese health authority ought to make adjustments in accordance with the burden of health care caused by co-morbidities.