A Three-Population Model for Sequential Screening for Bacteriuria12

Levy, P. S. and E. H. Kass (Channing Laboratory, Boston City Hospital, 774 Albany St., Boston, Mass. 02118). A three-population model for sequential screening for bacteriuria. Amer. J Epid., 1970, 97: 148–154.—A probabilistic model is derived for a sequential screening procedure widely used in the detection of bacteriuria. The model assumes the existence of three populations, namely true bacteriurics, nonbacteriurics who submit contaminated specimens and non-bacteriurics who do not submit contaminated specimens. Under the assumption that the sensitivity of the test is known, this model is used to derive estimators of the prevalence of bacteriuria in the population and the overall specificity of the test. In particular, the model is applied to a three-stage version of the test used at the Boston City Hospital. Estimation of the predictive value of a positive test is discussed in terms of the sensitivity and specificity of the screening test and the prevalence of bacteriuria in the population. A three-stage screening test is the minimum sequence required for estimation of these parameters and, in the Boston City Hospital data, appears to be the minimum necessary to obtain a high predictive value of a positive test.