全國醫院健保申報費用公開揭露之改善

Improving Mechanism of Information Disclosure in Reimbursement Claims for Taiwanese Hospitals

目標：建議健保揭露醫院費用申報資料的改善，輔以揭露各醫院：（1）當月申 報費用的期望落差、（2）歷次申報費用的變異度、（3）申報費用的排名與勝率值、 （4）費用申報的圖示分析。 方法：利用Rasch等級量表模式的參數估計與檢測，探討健保網站資訊揭露549家 醫院「2006年元月至2008年9月的33個月醫院的門急住診申報費用」資料的單向度性， 茲以利用其卡方統計量來鑑別醫院申報費用的變異度、標準化殘差分數，進而偵測當 月申報費用的期望差異度，再以申報費用的醫院排名，揭露其勝率值，且圖示分析各 醫院費用的申報情形。 結果：經移除95年的資料後，而達成單向度的測量，模式解釋量達99%，測量信 度為0.98；有26家醫院（4.74%）歷次申報費用的變異度太過度（Outfit MNSQ>2.0）， 2008年9月份有102家（18.58%）Z分數大於1.96（p<.05），其中以地區醫院及東區分 局的比率最高，分別達21.75%及37.50%；申報費用的期望差異度以地區醫院最多，變 異度以區域醫院與醫學中心最小。 結論：健保有必要將資訊揭露的醫院申報費用，再進一步顯示其（1）當月Z分 數、（2）長期偏離均方誤、（3）醫院的量尺分數、勝率值與排名、及（4）超連結各 醫院上述三項指標的圖示系統，以強化資訊公開全民共同的監督機制。

Objectives: To improve the disclosure mechanism in reimbursement claims for Taiwanese hospitals, we aim to complement information on 1) goal unexpected performance in month, 2) the historic stability in change over time, 3) ranking ordered hospital in magnitude of reimbursement, 4) depicting hospital reimbursement blueprint of interest. Methods: Appling rating scale model of Rasch analysis to examine data unidimensional property, we incorporate with Chi-square statistics, standardized residual Z-score, interval logit score derived from Rasch modeling to depict a detailed hospital’s pattern of reimbursement claims. Results: After removing data of year 2006, the month data fitted to the Rasch model’ s requirement of a single construct. Variance of model’s explanation was 99%, and Cronbach’ s Alpha was 0.98. There were 26 hospitals (4.74%) which long-term variation of reimbursement claims were shown under-fitting (Outfit MNSQ>2.0). In September of 2008, there were 102 hospitals (18.58%) which Z-score greater than 1.96(p<.05) indicates significantly aberrant distortion. Among them, both local area and region hospitals showed highest ratio, accounting for 21.75% and 37.50%, respectively. The local area hospital showed most in Z-score aberrance. Greater under data-fit demonstrated with Outfit MNSQ was shown less in both region hospitals and medical centers. Conclusions: BNHI could apply the four disclosed elements of Z-score, Outfit mean square, hospital ranks with logit scaled scores, and a hyperlink of individual performances for hospital reimbursement to enhance the efficacy of disclosure mechanism under a closer cooperative supervision by civics, academics and hospitals themselves for which hospital global budgeting has been implemented in Taiwan.