The estimation of dental caries incidence in the presence of diagnostic error

Diagnostic error in large scale screening programs for dental caries is frequent, as when teeth initially classified as carious are later diagnosed as never having been affected by caries. This paper presents a general formulation of diagnostic error in dental caries screening. The basic parameter of the formulation is the caries incidence rate. The general formulation of this paper permits an explicit comparison, in a common notation, of two specific models of diagnostic error--one due to Carlos and Senning, the other to Lu. Each model gives rise to a consistent estimator of the incidence rate. The distributional properties of these estimators had not previously been examined because of the dependence of teeth in the same mouth with respect to caries experience. Using the present formulation, the sampling scheme may be regarded as a one-stage cluster sample with mouths as clusters. This approach accounts for intracluster dependence thus permitting the derivation of an estimator of the relevant covariance matrix and a confidence interval for the incidence rate.