利用管制圖評估醫院醫療服務品質指標

目的：利用健保局網路上公佈的2008年第1季至2010年第2季（共10季的觀察期間）醫院總額24項醫療品質指標的執行成果，探討(1)2010年第2季非預期惡化偏離的指標分佈，在醫院層級及健保分局間是否相同？(2)過去10季觀察期間的指標改善趨勢，在醫院層級及健保分局間是否相似？方法：八種具有時間趨勢與無時間趨勢的管制圖，分析健保局最近一季的24項指標，以管制界線不同標準差的上下落點偏離位置，分別用0至8註記，形成558（醫院層級別指標）×8（方法）的評估資料。利用相關係數平方和與因素分析的因素得點（權重），選出最近一季資料的最適管制圖方法。卡方檢定2010年第2季的指標成果及過去10季觀察期間的指標改善趨勢，在醫院層級及健保分局間是否相同。發展一具管制圖資訊模組，協助醫務管理實務界，定期大量偵測指標落點的偏離情形。結果：研究發現：(1)選出標準差及其趨勢(X-Rm_trend)的組合管制圖，做為本次研究的評估工具；(2)2010年第2季的指標成果，在醫院層級間未呈統計顯著性的不同(p=.17)，但在健保分局間，呈現統計顯著性的差異(p=.0005)；(3)過去10季觀察期間的指標改善趨勢，在醫院層級間未呈統計顯著性的不同(p=.108)，但在健保分局間，呈現統計顯著性的差異(p=.025)。結論：本研究管制圖的資訊模組，協助偵測最近一季指標的偏離變化，也挑出呈現惡化趨勢的指標，及早提供檢討與改善。對於醫院總額預算支付執行委員會每季品質指標的檢討，提供資料內容更為充實的變革參考。
Objectives: This study examined 24 quality indicators for medical services published quarterly by Taiwan's Bureau of National Health Insurance (BNHI). BNHI data for 10 quarters from 2008 through the second quarter of 2010 were downloaded. The research objectives were 1) to investigate the association between hospital levels (or bureaus of BNHI) and outcomes of studied indicators for the latest quarter; 2) to determine the trend of studied indicators associated with hospital levels (or bureaus of BNHI) using data from the past 10 quarters..
Methods: We designated eight kinds of control charts where the latest performance indicators were classified on a nine-point scale scored from 0 to 8. A rectangular metric of 558 (attribute indicators) × 8 (alternative methods) was used to select an optimal method using both maximal sum squared correlation coefficients and factor analysis. Chi-square tests were used to verify the research questions.
Results: We found that 1) the control chart combined with standard deviation and X-Rm_trend was the best evaluation tool for the present study; 2) indicator outcomes were not significantly associated with hospital levels (p=.17) but were with bureaus of BNHI (p＜.0005) for the latest quarter; 3) no association existed between hospital levels and trends over the past 10 quarters (p=.108), but a significant difference was linked to bureaus of BNHI (p=.025).Conclusions: We suggest that BNHI apply the control chart module to detect trends and outliers for each indicator periodically monitored by BNHI. This model can be generalized to other kinds of indicators routinely measured in hospitals. Copyright © 2011 臺灣醫務管理學會.