目的：健保局為了降低COX-2選擇性抑制劑(COX-2)不當之用藥過高，於2006年10月實施COX-2處方不當比率超過部份不予支付處理方式。本研究目的在探討藥品處理方式介入之影響，並分析此類藥品符合給付規定與否之影響因素。方法：本研究擷取2005年至2007年健保資料庫門診病患COX-2處方資料，分析藥品支付處理方式之影響，運用決策樹演算法建立COX-2處方符合規定與否之預測模式。結果：本研究發現2007年之COX-2處方不當比率仍高達90%與2005年相當，基層診所不符合COX-2處方規定為92%，高於醫院門診處方之81%。急性上呼吸道感染為COX-2處方最常見之適應症。決策樹演算法產生二個主要資訊增益變數(病患年齡與醫院評鑑等級別)，病患年齡為COX-2處方規則集之最主要預測因子。本研究進一步發現COX-2普遍被開立於醫院門診就診兒童之處方。結論：健保COX-2用藥不當處理方式之介入，以實施後一年處方資料橫斷面分析仍尚未得到明確的結論，建議後續研究者進行多年縱貫研究來探討相關此議題。 Objectives: In order to decrease improper COX-2 selective (COX-2) inhibitors prescribing, the portions of COX-2 inhibitors prescriptions in violation of the medications rules adopted by Taiwan National Health Insurance exceeding the recommended rate are not eligible for reimbursement in October 2006. The aim of this study was to explore the effects of this medication management and examine the factors influencing whether COX-2 inhibitors prescription following medications rules. Methods: We selected COX-2 prescriptions from the NHI outpatient database during the period between 2005 and 2007. This database contains detailed records on the utilization of patients’ health care services. Data analysis was performed to examine the effects of medication reimbursement management and applied the classification and regression tree classifier (algorithm) to extract rules used in predicting ruled COX-2 prescription. Results: This study found that improper COX-2 prescriptions in 2007 reached 90 percent, were still high compared with previous year. General practitioners (GP) were more likely than hospital physicians (HP) to prescribe improper COX-2, the improper COX-2 prescriptions of GPs and HPs were around 92 percent and 81 percent, respectively. Acute upper respiratory infection was the most common indication over three years. The classification and regression tree algorithm produced two major information gain variables (patient’s age and hospital accreditation level). Patient’s age was the strongest predictor of ruled COX-2 prescription. Furthermore, we found that COX-2 was used widely in children’ prescription by HP. Conclusions: NHI COX-2 improper management intervention had not reached definite conclusion regarding 2007 prescriptions cross-sectional analysis. A longitudinal research is suggested to investigate these issues.