背景與目的:在疾病篩檢診斷工具正確性評估研究上,一般會考量以同一組實驗對象接受多種不同篩檢或診斷工具。對於此種施測工具分類正確性評估,本文將使用多個實際例子作為說明。分析方法:當探討使用二種(k=2)施測工具分類二元資料準確性研究時,本文使用ROC曲線分析方法,並引用Hanley and McNeil(1983)研究結果;對於多個(k=2)施測工具檢驗分類正確性的比較,本文進一步提出使用Bonferroni法,進行分類二元資料準確性之研究。結果與討論:依據本研究結果,當研究人員使用多種篩檢或診斷工具進行分類時,將可選擇適當的統計方法評估何種篩檢或診斷工具有較高的分類正確性。
Background: In this research, we evaluated the accuracy of different screening methods and diagnostic tools for a disease with the same set of patients. We use several examples to illustrate the accuracy of these screening tools. Method: Receiver Operating Characteristic Curve Analysis (ROC curve analysis) was used to assess the accuracy of different screening tools for a disease using the same set of patients. The result of Hanley and McNeil (1983) can be applied in the comparison of two (k>2) screening tools. However, for the comparison of multiple (k>2) screening tools, we suggest the use of Bonferroni method instead. Result: This paper provides a suggestion as to which statistical method should be used when evaluating the accuracy of screening tools.