目標：本研究結合現行的家庭醫師整合性照護計畫與論人計酬支付模式，分析高屏地區鄉村型與都市型家醫群診所就醫病患樣本，探討影響醫療費用之風險校正因子，並建構適合預測醫療費用模式。方法：本研究資料來源主要為2012-2013年全民健保申報資料，和高屏地區鄉村與都市二個區域有參加「家庭醫師整合性照護計畫」診所就診的病人資料。先隨機抽樣一個地區家醫群診所就醫病患病人作為訓練樣本，建立預測醫療費用的風險校正模型，並與其他兩組隨機抽樣家醫群診所樣本進行模型組內和組間的交叉驗證，以評估及確認風險調整模型之預測能力。結果：鄉村型與都會型家醫群驗證樣本人口學特性或醫療利用特質差異影響實際醫療資源利用。鄉村型病人六十歲以上比例和疾病嚴重度相對都會型家醫群樣本較高，反應在2013年實際醫療總費用分別為29,658點和25,087點。另外，風險校正費用預測模型隨著加入變項數而影響醫療費用的預測力，校正後預測力R^2從6.19%增加到40.20%。結論：當進行風險計價模型時，應考量適切的風險因子和合理的成長率，以增加風險計價的準確度和推估的敏感度，並且視地區費用成長特性和個別族群病人風險特質，適切地反應其醫療資源耗用狀況，以確保醫療照護就醫公平性及醫療資源有效分配與利用。

Objectives: The aims of this study were to investigate the risk factors associated with patients' health expenditures and to apply risk-adjusted medical expenditure prediction models using the primary care physician integrated care plan and the capitation payment model. Methods: Data were drawn from the 2012-2013 national health insurance (NHI) claims data and the primary care physician integrated care plan data for two primary care groups (one in a rural and the other in an urban area of the Kaoping area in southern Taiwan). We randomly selected one group of patients from one primary care group as a training set to conduct the risk-adjusted models, and compared two random samples from rural and urban primary care groups to assess the within and between group validity of the prediction models. Results: This study supported an association between patients' demographic or comorbid characteristics and health expenditures in the rural and urban primary care patient groups. Patients in the rural area were age sixty or older and their health expenditures in 2013 were greater than those in the urban group (NTD 29,658 vs. 25,087). In addition, the model's goodness-of-fit measured as adjusted R^2 increased from 6.19% to 40.20% when more risk factors were added to the risk-adjusted health expenditure models. Conclusions: When risk-adjusted models to predict medical costs are used, appropriate risk factors and reasonable growth rates need to be considered in order to increase the accuracy and sensitivity of the prediction of medical costs. Moreover, geographic variability also needs to be considered in order to reflect the consumption of medical resources in different areas and to ensure the horizontal equity and efficient allocation of medical resources across rural and urban areas.