

Estimate the true prevalence, as well as positive
and negative predictive values and
likelihood ratios from
survey testing results using a test of known
sensitivity and
specificity.
Confidence limits for both apparent and true prevalence estimates are calculated.
Values are also plotted for a range of possible survey results.
Confidence limits for apparent prevalence use the Wilson binomial approximation
from: Brown, LD, Cat, TT and DasGupta, A (2001). Interval Estimation for a proportion.
Statistical Science 16:101133.
True prevalence estimates are calculated as described by:
Rogan and Gladen (1978). Estimating prevalence from the results of a screening test.
American Journal of Epidemiology 107:7176. True prevalence estimates
that are less than zero or greater than one are not consistent with assumed sensitivity
and specificity values, and are indicated by '<0' and '>1', respectively.
If sample sizes for sensitivity and
specificity estimation are provided, confidence limits are based on variance estimates incorporating
additional uncertainty associated with these values, also as described by Rogan and
Gladen (1978). Otherwise, confidence limits assume
sensitivity and specificity are known exactly and use the method described by Greiner, M
and Gardner, IA (2000). Application of diagnostic tests in epidemiologic studies.
Preventive Veterinary Medicine 45:4359.
Blaker's, Sterne, ClopperPearson and Wilson confidence limits are
calculated as described by Reiczigel, Földi and Ózsvári (2010). Exact confidence limits for
prevalence of a disease with an imperfect diagnostic test,
Epidemiology and Infection 138:16741678. The authors recommend Blaker's interval for general use.
See the
author's home page or follow the link to
Confidence limits for prevalence adjusted for sensitivity/specificity for more information.
