When is a test positive? The use of decision analysis to optimize test interpretation.

Abstract

Clinical laboratory are often provided as numerical values that are then interpreted as being positive or negative. While this approach might simplify interpretation, it also makes interpretation contingent on a standard test cutoff point. Alternatively, test results can be interpreted for a specific patient with reference to the particular patient's probability of disease, the benefit of detecting disease when it is present, and the cost of mistakenly making the diagnosis when the disease is a absent. This paper explains the analysis of laboratory test results using techniques from decision analysis and receiver operator characteristic (ROC) curve analysis to define a positive result. The relationship between the ROC curve and likelihood ratios is illustrated using the diagnosis of urinary tract infection (UTI) to illustrate these concepts.