建置資訊模組篩選費用異常以因應健保DRGs之實施

A Module for Screening out the Most Unexpected Reimbursement under DRGs

DRGs的實施，醫院將面臨如何用科學客觀且簡易的資訊模組篩選出健保費用的異常，及早檢討漏帳或是出院診斷書寫不完整等的現象，以改善內部流程上的可能錯誤。利用試題反應理論的Rasch(1960)分析，建置網路化的資訊模組：(1)檢視以年或月為DRGs綜效潛能參數估計是否對篩選結果有顯著性差異；(2)網路模組的建立，協助醫院每日篩選出統計顯著性申報費用異常的案件，及早檢討及防範系統上的可 作業偏差。以2004年國家衛研究院發行之17家醫學中心健保住院資料庫，針對健保第一期將導入實施的6個MDC 94,536筆個案，對於各醫院導入期的不同DRGs費用，利用Winsteps軟體分析出各醫院DRG費用效標及各DRGs的閾難度估計。超出±1.96外的標準化殘差被視作費用極端值的個案。結果顯示，以年或月為醫院DRG費用效標參數估計，對篩選申報費用異常的案件數沒有統計顯著性差異；每家醫院每月平均12件申報費用需予再做檢討；網路篩選的資訊模組，為醫院疾病分類人員提供線上即時的申報費用異常報表；健保局也可利用此一資料探勘方式，糾舉可疑的醫院DRGs取巧或編碼不當昇級行為，以維護醫院總額支付制度下的資源公平與正義。

How to screen out the most unexpected reimbursement for DRGs is a controversial issue in the fields of healthcare and hospital management. The aims of this study consist of (l) examining significant differences which could be affected by data collection based on months or annuals; (2) building a module to help hospital managers finding unexpected reimbursements for DRGs implementation in order to instantly correct the possible errors under daily routines. We adopted Rasch's, a Danish mathematician, measurement model to illustrate the development of screening module under global budgeting through DRGs payment system. A total of 94,536 discharge cases of 17 medical centers in Taiwan in 2004 were analyzed using Winsteps software to measure the latent trait of each hospital in medical reimbursement under DRGs and to calibrate threshold difficulties across all the DRGs items. Standardized residual analysis utilized in this study to detect unexpected reimbursements in terms of standard errors beyond the value of ±1,96 was demonstrated via a web-module. The results showed that there is no any statistically significant difference in the abnormal reimbursement cases screened out through the measures based on either months or annuals. A total of averaged 140 unexpected cases each month were needed to be rechecked in those 17 medical centers in year 2004. The module implemented on internet could help DRGs grouping clerks efficiently find out the possible unexpected matching cases beforehand for instant correction. The Taiwan's Bureau of National Health Insurance (BNHI) is expected to use this kind of data mining in abnormal case selection for fairness and justness under global budgeting through DRGs prospective payment system.