Sensitivity and specificity in diagnostic accuracy: ROC solid or lost in translation?
- 1 Department of Surgery, Stavanger University Hospital, Stavanger, Norway
- 2 Department of Surgical Sciences, University of Bergen, Bergen, Norway
- Correspondence to Dr Kjetil Søreide, Department of Gastroenterologic Surgery, Stavanger University Hospital, POB 8100, Armauer Hansens vei 20 Stavanger N-4068, Norway; ksoreide{at}mac.com
- Accepted 1 July 2009
I thank Bharti and Bharti for their appreciation of the article on receiver-operating characteristics curve (ROC) analysis in biomarker research.1 In the ROC article, I mentioned the diagnostic accuracy as defined by sensitivity and specificity, and how a shift in any chosen test cut-off may influence the sensitivity and specificity, illustrated in a figure.2 Indeed, and as pointed out by Bharti et al, the figure displaying the shift in cut-off threshold for a given test and the influence on sensitivity and sensitivity contains an error, in that it has mixed the terms “sensitivity” and “specificity” in the text boxes. I apologise to the readers for causing this confusion. The concept of diagnostic accuracy, including the terms sensitivity and specificity, although used on a global scale, every day in clinical decision-making, may be easily confused, and I shall easily admit …
