目標：了解同儕資訊回饋對醫師開立TG（三酸甘油脂）檢驗的影響。方法：選取內科共18位門診看診醫師，隨機分派各9位醫師進入實驗組或與對照組。於92年8月起到12月，進行4次資訊回饋，資訊內容為該月門診平約每百張處方箋開立TG次數，將每位醫師開立檢驗的次數依序排列於單張報表中。控制組醫師則未接受任何資訊回饋。以介入前實驗組醫師的門診處方資料（91年12月到92年3月及92年7月與8月）為回饋前的對照基線資料。並於實驗介入結束後半年（93年6月／8月），對實驗組醫師進行深入訪談，以暸解其對實驗介入的主觀看法。結果：實驗介入後，實驗組與對照組醫師平均每百張門診處方箋開立TG檢驗次數無統計顯著差異，但實驗組醫師在實驗介入後，開立TG檢驗次數有降低趨勢，介入後平均值較介入前減少0.134，對照組則有上升趨勢，開立TG檢驗次數較之前平均增加0.504次，與文獻結果符合。就個別醫師的訪談結果，在回饋前檢驗次數平均值偏高的醫師，回饋後檢驗次數平均值則呈下降趨勢，而回饋前檢驗次數平均值偏低的醫師，回饋後檢驗次數平均值則呈上升趨勢，同儕資訊對位於極端值的醫師有影響，支持社會比較理論。醫師對同儕資訊回饋的內容較關切，多數醫師(67%)亦肯定同儕資訊回饋對醫師行為可能產生影響。結論：同儕資訊回饋是否可以做為修正醫師行為的方法，其合理的操作方式或比較基準需要進一步探討。

Objective: The purpose of this study is to investigate the effects of peer comparison feedback on changing TG test ordering among internists in a teaching hospital. Methods: A quasi-randomized time series controlled trial was conducted. Eighteen physicians were randomly assigned into the experimental and control groups. The experiment was conducted during a period of six months. In-depth interviews with six physicians in the experimental group were carried out afterwards. Results: The results indicate that physicians in the experimental and control groups show no significant difference on the average counts of ordering TG tests per 100 orders. Physicians in the experimental group show a slight decrease in the average counts of ordering TG tests per 100 orders, yet physicians in the control group show otherwise. In-depth interviews show that physicians who have relatively higher or lower average counts of ordering TG tests tend to change their behaviors after intervention. The results support the social comparison arguments. Physicians interviewed indicate positive attitudes toward the effect of peer comparison information feedback on changing physician behaviors, yet show more concerns about the information from feedback. Conclusion: More studies are needed to determine whether or not peer comparison feedback can be used to change physician behaviors.

醫師 ； 同儕比較 ； 資訊回饋 ； 開立TG檢驗行為

physicians, peer comparison, information feedback, test-ordering behaviors