Verdonschot E H, Wenzel A, Bronkhorst E M
Department of Cariology and Endodontology, University of Nijmegen, The Netherlands.
Community Dent Oral Epidemiol. 1993 Aug;21(4):203-8. doi: 10.1111/j.1600-0528.1993.tb00757.x.
The quality of a diagnostic tool for caries detection is usually evaluated by quantification of observer performance using sensitivity and specificity values calculated from data obtained from ordinal caries depth rating scales. The application of Receiver Operating Characteristic (ROC) analysis has been restricted to the use of confidence rating scales. The aim of this investigation was to study the appropriateness of ROC analysis using caries depth ratings by comparing sensitivity and specificity values to parameters of diagnostic performance obtained by ROC analysis. In two recent reports, sensitivity and specificity values were utilized to express diagnostic accuracy of observer performance from specified caries diagnostic systems. The raw data from these investigations were subjected to ROC analysis in the current study. The results illustrated that ROC analysis, producing estimates of sensitivities for all specificities, yielded more comprehensive measures of diagnostic performance than single values for sensitivity and specificity. In particular, the area under the ROC curve constituted a clearly interpretable parameter representing the quality of diagnostic performance.
用于龋齿检测的诊断工具的质量通常通过使用从序数龋齿深度评级量表获得的数据计算出的敏感性和特异性值来量化观察者表现进行评估。受试者工作特征(ROC)分析的应用仅限于置信度评级量表。本研究的目的是通过将敏感性和特异性值与通过ROC分析获得的诊断性能参数进行比较,研究使用龋齿深度评级进行ROC分析的适用性。在最近的两份报告中,敏感性和特异性值被用来表达特定龋齿诊断系统中观察者表现的诊断准确性。在本研究中,对这些调查的原始数据进行了ROC分析。结果表明,ROC分析产生了所有特异性的敏感性估计值,比敏感性和特异性的单一值能更全面地衡量诊断性能。特别是,ROC曲线下的面积构成了一个清晰可解释的参数,代表诊断性能的质量。