Skip to main content

Table 5 Simulation study with 2000 replications

From: Statistical methods to correct for verification bias in diagnostic studies are inadequate when there are few false negatives: a simulation study

Percentage Verified Proportion with n false negatives AUC over 2000 replications Coverage Probability over 2000 replications
  n = 0 n = 1 n = 2 n > 2 Mean 2.5th – 97.5th percentile  
Probabilities of verification likely to be encountered in a screening study
True n/a n/a n/a n/a 0.750 0.728, 0.774 95%
10% 83% 16% 2% 0% 0.758 0.577, 0.860 77%
30% 41% 37% 17% 6% 0.752 0.677, 0.813 89%
60% 1% 6% 12% 80% 0.750 0.713, 0.786 93%
Probabilities of verification were adjusted such that false negatives were more (less) likely to be present for 10% (60%) verified
10% 25% 35% 23% 16% 0.751 0.677, 0.820 93%
60% 98% 3% 0% 0% 0.728 0.552, 0.819 67%
  1. 5000 participants are enrolled in a screening study with an underlying incidence of disease of 10%. False negatives are defined as verified participants with a diagnostic test result less than the 20th percentile of the diagnostic test results and with a positive gold standard result.