Designs for Testing Group-Based Interventions with Limited Numbers of Social Units: The Dynamic Wait-Listed and Regression Point Displacement Designs.
Designs for Testing Group-Based Interventions with Limited Numbers of Social Units: The Dynamic Wait-Listed and Regression Point Displacement Designs.
复制标题
用于测试具有有限数量的社会单位的基于群体的干预的设计：动态等待列表和回归点位移设计。
DOI：
10.1007/s11121-014-0535-6
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
Brown,CHendricks
中科院分区：
文献类型：
--
作者：
Wyman,PeterA;Henry,David;Knoblauch,Shannon;Brown,CHendricks
动态等待列表设计（DWLD）和回归点位移设计（RPDD）解决了当组数量有限时评价基于组的干预措施的几个挑战。DWLD和RPDD都利用了提高统计能力的效率，并可以加强社区需求和研究优先事项之间的平衡。DWLD比传统的等待列表设计阻止更多的时间单位，从而增加了研究期间的比例，在此期间可以比较干预和控制条件，还可以改善在多个站点实施干预的后勤工作，并加强保真度。我们讨论了更大的背景下，推出随机设计的DWLD，并比较它与它的表弟阶梯楔形设计。RPDD使用从其中选择干预单元的设置人群的档案数据，为接受干预的单元创建预期的后测分数，并与实际后测分数进行比较。高的前测-后测相关性使RPDD具有评估干预影响的统计能力，即使在一个或几个设置接受干预时。RPDD在干预前后数年的档案数据可用时效果最佳。如果干预单元不是随机选择的，则倾向分数可用于控制非随机选择因素。DWLD和RPDD用于评估，分别在32所学校和暴力预防计划（停火）在两个芝加哥警区在10年期间的自杀预防培训（QPR）提供的例子。DWLD和RPDD如何解决内部和外部有效性的共同威胁，以及它们的局限性，进行了讨论。
The dynamic wait-listed design (DWLD) and regression point displacement design (RPDD) address several challenges in evaluating group-based interventions when there is a limited number of groups. Both DWLD and RPDD utilize efficiencies that increase statistical power and can enhance balance between community needs and research priorities. The DWLD blocks on more time units than traditional wait-listed designs, thereby increasing the proportion of a study period during which intervention and control conditions can be compared, and can also improve logistics of implementing intervention across multiple sites and strengthen fidelity. We discuss DWLDs in the larger context of roll-out randomized designs and compare it with its cousin the Stepped Wedge design. The RPDD uses archival data on the population of settings from which intervention unit(s) are selected to create expected posttest scores for units receiving intervention, to which actual posttest scores are compared. High pretest–posttest correlations give the RPDD statistical power for assessing intervention impact even when one or a few settings receive intervention. RPDD works best when archival data are available over a number of years prior to and following intervention. If intervention units were not randomly selected, propensity scores can be used to control for non-random selection factors. Examples are provided of the DWLD and RPDD used to evaluate, respectively, suicide prevention training (QPR) in 32 schools and a violence prevention program (CeaseFire) in two Chicago police districts over a 10-year period. How DWLD and RPDD address common threats to internal and external validity, as well as their limitations, are discussed.