Effect of disease prevalence on percent bias. Effect of disease prevalence on percent bias in observed sensitivity (A) and specificity (B). Parameter definitions are as in "Parameters" (Table 6), except that the disease prevalence is allowed to vary. Percent bias is the bias in observed sensitivity or specificity divided by the true sensitivity or specificity. The observed results for Test 2 in a test if negative series design are denoted by "Test 2 Series". The observed results for a single test design are denoted by "Single". The observed sensitivity is biased upwards by 14% for the single test design and 12% for the series design.
Ringham et al. BMC Medical Research Methodology 2010 10:3 doi:10.1186/1471-2288-10-3