在院中DRGs費用之預測與管理

Prediction and Management of Medical Fees When Patients Receiving Cares in Hosiptals under DRGs

錢才瑋(Tsair-Wei Chien) ； 王文中(Wen-Chung Wang) ； 陳年興(Nian-Shing Chen) ； 林宏榮(Hung-Jung Lin)

《醫療資訊雜誌》 16卷4期 (2007/12) Pp. 13-24

https://doi.org/10.29829/TJTAMI.200712.0002

摘要

健保局計劃於2008年元月起導入台灣版本的DRGs，醫療院所除致力於出院後疾病分類的品質提昇，更期望建置在院中費用偵測與管理的模式，以因應DRGs的實施。本研究(1)介紹DOS程式碼介面，連結健保DRG編碼程式，實現單筆即時DRGs編碼的臨床需要；(2)模擬預測在院中費用是否超出DRG支付定額，比較3方案的敏感度與明確度，以ROC曲線下之面積判別在院中之預測準確度；(3)利用Rasch模式的殘差分析，與各自DRG費用常模做比較，挑出影響費用偏高的可能項目，並蒐集健保核減資料，驗證殘差分析所提示效果的準確度。研究發現，利用DOS程式批次指令，結合Excel與健保DRG編碼程式或是Rasch分析的WINSTEP軟體，可於10秒內完成一筆DRG的編碼或是超出預期費用項目之提示圖示分析報表；模擬分析以Rasch(1960)模式的羅吉斯迴歸預測的提示準確度最高；僅有20多行的程式，以機率原理及統計科學結合資訊模組的技術，對醫師於病患在院中的DRGs費用管理，可提昇其效率與效能；實證殘差分析提示效果達76%，以檢驗檢查最佳、其次為藥品費用、再次為醫材費用。

關鍵字

模擬預測 ； 殘差分析 ； 敏感度 ； 明確度 ； ROC曲線下面積

並列摘要

Taiwan's Bureau of National Health Insurance (BNHI) has planed to implement inpatient DRGs payment system scheduled effective in January of 2008. Many hospital managers urgently invent initiatives to decrease impacts of DRGs. Prediction of everyday medical fee in a hospital for a patient deserves more studies. Unexpected high fee should be submitted to attending physicians for possible deficits under DRGs. In this study, we proposed three alternatives of regression prediction models and examine their quality and accuracy using simulation data. Sensitivity and specificity with areas under the ROC curve (AUC) were used to determine the accuracy of these alternative models. A diagram that can be produced in ten seconds was shown to illustrate how a medical individual fee beyond expectation can be detected from Rasch (1960) model's residual analysis. Microsoft DOS and Windows shell commands that bridges Excel and BNHI DRG grouper and Rasch's WINSTEPS software were developed to facilitate efficacy and effectiveness in management of patient's hospitalization cares under DRGs. The effects on costs of medical care with computer-generated informational messages directed to physicians in hospitalization are foreseen and expected. The effect of Rasch residual analysis was as high as 76% in terms of reimbursement deduction with prompt onto any at least one item medical fee. The most prompt effect of residual analysis was medical examination, followed by drug consumption and medical materials.

並列關鍵字

Prediction ； residual analysis ； sensitivity ； specificity ； AUC