以Rasch分析評估醫院層級間的品質指標

Using Rasch Analysis to Assess Quality Indicators in Different Level of Hospitals

錢才瑋(Tsair-Wei Chien) ； 王文中(Wen-Chung Wang)

《醫務管理期刊》 8卷3期 (2007/09) Pp. 249-260

https://doi.org/10.6174/JHM2007.8(3).249

摘要

目的：利用健保局在網路上公佈的2002年至2006年第二季醫院總額醫療品質的11項指標執行成果，以便：(l)探討品質指標在醫院層級問是否有族群功能差異(DIF)的現象，(2)評估指標是否具到量單一向度以及這些指標對三種醫院層級間的信度，(3)發展一套Z分數KIDMAP，以協助診斷非預期反應指標之檢討。
方法：利用Rasch分析的WINSTEPS軟體，進行評定量尺模式的資料分析與模式參數的估計，探討DIF現象與信度，比較醫院層級問的品質執行的表現，並用KIDMAP呈現非預期反應。
結果：大部分的品質指標在醫院層級間呈DIF的現象，醫院層級三個分量表信度以區域醫院最低，品質績效以醫學中心最高，地區醫院最低。以區域醫院爲參考基準，醫學中心勝算比爲1.3，地區醫院爲0.32。KIDMAP非常簡單有用。
結論：健保局可再(l)擴增更多且難度分佈均勻的品質指標，且直接以醫院指標成果而非其彙整資料做分析，以提昇信度；(2)增加更符醫療品質的指標而非僅是醫療費用節約的指標，以提昇效度；(3)研究論品質績效計酬的可能工具與機制，以提昇醫療品質。

關鍵字

試題差異功能 ； 羅許分析 ； 醫院層級 ； 品質績效 ； 測量信度

並列摘要

Objectives: This study examined 11 quality indicators of medical services quarterly disclosed by Taiwan's Bureau of National Health Insurance (BNHI), with objectives of 1) investigating whether these indicators exhibit differential item functioning (DIF) over three hospital levels? 2) ascertaining whether these indicators constitute unidimensionality and yield reliable measures within hospital levels? and 3) developing a Z-score KIDMAP to help identify unexpected response patterns.
Methods: The rating scale model was fitted to the data using the WINSTEPS program. Parameters were calibrated. Reliability and DIF were assessed. The performances were compared across hospitals. KIDMAP was constructed to identify unexpected response patterns.
Results: Most indicators had DIF across hospital levels. For the reliability of the three subscales, the regional hospitals had the lowest reliability. Medical centers had the highest quality performance, while district hospitals had the lowest. Taking regional hospitals as a reference, the odds-ratios for medical centers and district hospitals were 1.3 and 0.32, respectively. KIDMAP was found very indicative.
Conclusions: It is suggested that BNHI take the following policies: (1) develop more quality indicators with appropriate difficulty levels and release raw data rather than aggregated data to increase subscale reliability; (2) recruit more concrete medical quality indicators other than those of cost containment; and (3) investigate feasibility of PFP (pay-for-performance) and develop self assessment mechanism to upgrade medical quality.

並列關鍵字

DIF ； Rasch analysis ； hospital level ； quality performance ； measurement reliability

參考文獻

醫院總額專業醫療服務品質報告

Rasch G.(1960).Probabilistic models for some intelligence and attainment tests.Copenhagen:Institute of Educational Research.

Rasch G.(1980).Probabilistic models for some intelligence and attainment tests.Chicago:The University of Chicago Press.

王文中(2004)。Rasch測量理論與其在教育和心理之應用。教育與心理研究。27，637-694。

Holland PW,Wainer H.(1993).Differential item functioning.Hillsdale, NJ:Erlbaum.