潜在分类模型在无金标准诊断试验中的评价

Evaluation of Diagnostic Test in the Absence of Gold Standard with Latent Class Model

目的探讨无金标准诊断试验下灵敏度与特异度的估计方法。方法以三个诊断试验为例,利用潜在分类模型估计无金标准诊断试验的灵敏度、特异度及其95%可信区间。并将潜在分类模型估计的结果与Bayesian方法估计结果进行比较。结果三个诊断试验的灵敏度与特异度的95%可信区间均存在一定的重叠,表明两种方法估计结果较为相似。结论诊断试验的灵敏度与特异度在无金标准时是无法直接计算的,可利用潜在分类模型与Bayesian方法估算诊断试验的灵敏度与特异度。

Objective To explore the methods to evaluate sensitivity and specificity of diagnostic test without gold standard.Methods Taking three diagnostic test without gold standard for example,Sensitivity,specificity and their 95% confidence interval were computed by latent class model（LCM）.And results were compared with those evaluated by Bayesian model.Results The 95% confidence interval of sensitivity and specificity by LCM and those of Bayesian models had overlapped,their parameter were similar.Conclusion Without gold standard,sensitivity and specificity would not to be computed directly.But,latent class model and Bayesian model could be used to estimate sensitivity and specificity and their 95% confidence interval in absence of gold standard.