Suppr超能文献

转诊模式对诊断测试特征的影响。

The influence of referral patterns on the characteristics of diagnostic tests.

作者信息

Knottnerus J A, Leffers P

机构信息

Department of General Practice, University of Limburg, The Netherlands.

出版信息

J Clin Epidemiol. 1992 Oct;45(10):1143-54. doi: 10.1016/0895-4356(92)90155-g.

Abstract

The discrimination of a diagnostic test--characterized by sensitivity, specificity, likelihood ratio and ROC curve--may be influenced by referral patterns of general practitioners. Symptoms and test results in particular will affect the probability of referral, while the degree of development of the pathological process directly influences the probability of positive test results. Using numerical examples, we analyse and discuss a few specific situations: (1) referral depends only on symptoms; (2) referral depends both on symptoms and on test results; (3) referral depends only on test results. In the first situation, test characteristics and predictive values are invariant over the strata of symptomatology, while in the third situation the predictive values are unchanged. If there is a positive relationship between positive test results and referral probability, overall sensitivity will increase while specificity and likelihood ratio will decrease. A general representation is given for the evaluation of the direction of change of the likelihood ratio as a function of referral probabilities. The shape of receiver-operating characteristic curves is less sensitive to bias, but at the level of specific cut-off points considerable changes may occur.

摘要

诊断试验的判别——以灵敏度、特异度、似然比和ROC曲线为特征——可能会受到全科医生转诊模式的影响。症状和检测结果尤其会影响转诊的概率,而病理过程的发展程度直接影响检测结果为阳性的概率。我们通过数值示例分析和讨论了几种具体情况:(1)转诊仅取决于症状;(2)转诊既取决于症状也取决于检测结果;(3)转诊仅取决于检测结果。在第一种情况下,检测特征和预测值在症状分层中是不变的,而在第三种情况下预测值不变。如果检测结果为阳性与转诊概率呈正相关,总体灵敏度会增加,而特异度和似然比会降低。给出了一个通用表示法,用于评估似然比随转诊概率变化的方向。接受者操作特征曲线的形状对偏差不太敏感,但在特定临界点水平可能会发生相当大的变化。

文献检索

告别复杂PubMed语法,用中文像聊天一样搜索,搜遍4000万医学文献。AI智能推荐,让科研检索更轻松。

立即免费搜索

文件翻译

保留排版,准确专业,支持PDF/Word/PPT等文件格式,支持 12+语言互译。

免费翻译文档

深度研究

AI帮你快速写综述,25分钟生成高质量综述,智能提取关键信息,辅助科研写作。

立即免费体验