Assessing equivalence: An alternative to the use of difference tests for measuring disparities in vaccination coverage
Assessing equivalence: An alternative to the use of difference tests for measuring disparities in vaccination coverage
复制标题
DOI：
10.1093/aje/kwf149
复制
发表时间：
2002-12-01
影响因子：
5
通讯作者：
Chu, SY
中科院分区：
文献类型：
--
作者：
Barker, LE;Luman, ET;Chu, SY
消除不同群体在疫苗接种覆盖率方面的健康差距是公共卫生政策的基石。然而，传统上使用的统计检验不能证明组之间存在无差异的状态。人们不会问“人群间免疫覆盖率的差异--或差异--已经消除了吗？”，人们可以问，“是否实现了实际的对等？”一种称为等价性测试的方法可以表明，组之间的差异小于可容忍的小量。本文演示了该方法，并介绍了对确定可容忍差异水平有影响的公共卫生考虑因素。使用2000年全国免疫调查的数据，作者测试了白人和其他种族/民族成员之间的疫苗接种率在统计上的显著差异，以及白人和这些相同群体之间的等效性。对于一些少数群体和一些疫苗，覆盖率在统计上明显低于白人；然而，对于其中一些群体和疫苗，等效性检验表明实际上是等同的。为了使用等价性测试来评估差异是否仍然对公共健康构成威胁，研究人员必须了解何时使用该方法，如何建立关于可容忍的微小差异的假设，以及如何解释测试结果。
Eliminating health disparities in vaccination coverage among various groups is a cornerstone of public health policy. However, the statistical tests traditionally used cannot prove that a state of no difference between groups exists. Instead of asking, "Has a disparity-or difference-in immunization coverage among population groups been eliminated?," one can ask, "Has practical equivalence been achieved?" A method called equivalence testing can show that the difference between groups is smaller than a tolerably small amount. This paper demonstrates the method and introduces public health considerations that have an impact on defining tolerable levels of difference. Using data from the 2000 National Immunization Survey, the authors tested for statistically significant differences in rates of vaccination coverage between Whites and members of other racial/ethnic groups and for equivalencies among Whites and these same groups. For some minority groups and some vaccines, coverage was statistically significantly lower than was seen among Whites; however, for some of these groups and vaccines, equivalence testing revealed practical equivalence. To use equivalence testing to assess whether a disparity remains a threat to public health, researchers must understand when to use the method, how to establish assumptions about tolerably small differences, and how to interpret the test results.