Reed G F, McHugh R B
Biometrics. 1979 Jun;35(2):473-8.
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.
在大规模龋齿筛查项目中，诊断错误很常见，比如最初被归类为龋齿的牙齿后来被诊断为从未患过龋齿。本文提出了龋齿筛查中诊断错误的一般公式。该公式的基本参数是龋齿发病率。本文的一般公式允许用一种通用符号对两种特定的诊断错误模型进行明确比较——一种是卡洛斯和森宁提出的，另一种是卢提出的。每个模型都产生了发病率的一致估计量。由于同一口腔中的牙齿在龋齿经历方面存在相关性，这些估计量的分布特性以前尚未得到研究。使用当前的公式，抽样方案可被视为以口腔为聚类的单阶段聚类抽样。这种方法考虑了聚类内的相关性，从而允许推导相关协方差矩阵的估计量和发病率的置信区间。
