DRGs審查核減費用之預測分析：Rasch分析法之應用

Prediction of DRGs for reimbursement assessment in Taiwan: application of Rasch analysis

健保費用核減一直是醫院界所爭議的主題。檔案分析抽審的準確性，可以避免醫療院所的爭議。本研究試圖以標準化的費用殘差，比較相同DRGs下的醫療費用離群情形，利用Rasch模式的統計技術，來預測健保審查的費用核減案件，檢測其準確性。利用2004及2005年全國醫學中心住院申報資料，整理出20個各300件的DRGs，當做樣本常模參照。針對某醫學中心該2年健保住院被核減的194件出院個案資料，一件件與其常模參照的DRGs資料做比較，以標準化殘差大於2.0（即p<.05）做為檢測是否呈該項費用偏高的依據，因此探討：（1）被健保核減的醫療費用是否正確？（2）被核減的藥品、衛材及檢查費用的正確性，在兩個年度間是否有差異？（3）最可能被核減的醫療費用或DRGs是那些？研究結果發現，（1）標準化殘差偵測出來的核減正確性達0.82，ROC曲線下面積（正確性）顯著大於0.50；（2）藥品、衛材及檢查費用在兩年度間的被核減正確性沒有差異；（3）DRGs中以涎腺切除術（無合併症或併發症）（DRG05002）最易被核減，醫療費用中以DRG05002、放射與特殊材料費用被核減的可能性最高。以DRGs做為同質資料比較的基礎，搭配Rasch分析的統計技術，可以輔助健保申報資料檔案分析，而提昇其審查醫療費用的效率與正確性。

The accuracy and effects of reimbursement assessment determined by Bureau of National Health Insurance (BNHI) have been debated in past decade. Using techniques of Rasch model’s standardized residual analysis applied to 20 sets of norm referenced DRGs (each with 300 cases) in comparison with 194 respective examined DRGs reimbursement cases, we examined whether the results of prediction with Rasch modeling have a high probability associated with actually deducted cases. The discharged cases with their medical expenses deducted by BNHI in 2004 and 2005 were analyzed from a medical center in Taiwan. The results showed that 1) the effects deducted by BNHI assessment could be predicted with an accurate rate of 0.82 by the standardized residual approach of Rasch model, 2) the accuracies for drug, medical material and examination fees were not associated with the years, all of those areas under the ROC curve (AUC) were significantly greater than the randomized probability of 0.50, 3) DRG05002, X-ray and special material fees exhibited the greatest probability of deduction. The norm referenced DRGs developed to compare with a specific reimbursement discharged case using Rasch analysis and profile analysis of patient reimbursement data are recommended to BNHI and hospitals for monitoring unnecessary expenditure of medical healthcare.