Yoder Paul, Compton Don
Vanderbilt University, Nashville, Tennessee 37203, USA.
Ment Retard Dev Disabil Res Rev. 2004;10(3):162-8. doi: 10.1002/mrdd.20013.
This article provides a rationale for considering predictors of growth in a treatment group as inadequate to identifying predictors of treatment response. When we interpret predictors of growth in a treatment group as synonymous with predictors of treatment response, we implicitly attribute all of the treated children's growth to the treatment, an untenable assumption under most conditions. We also contend that the use of standard scores in predictors of growth studies does not allow us to differentiate growth from treatment, from growth from other factors. We present two research methodologies that are appropriate methods of identifying predictors of treatment response: (a) single-subject experimental logic utilized to identify the specific participants in which treatment responses (not just growth) were found, combined with follow-up group comparison logic to identify the characteristics on which responders and nonresponders differ, and (b) statistical interactions among child/family/context characteristics and randomly assigned group membership. Principles for selecting potential predictors of treatment response are provided.
本文提供了一个理由，即认为将治疗组中生长的预测因素视为不足以识别治疗反应的预测因素。当我们将治疗组中生长的预测因素解释为与治疗反应的预测因素同义时，我们隐含地将所有接受治疗儿童的生长归因于治疗，这在大多数情况下是一个站不住脚的假设。我们还认为，在生长研究的预测因素中使用标准分数并不能使我们区分治疗导致的生长与其他因素导致的生长。我们提出了两种适用于识别治疗反应预测因素的研究方法：(a) 单受试者实验逻辑，用于识别发现治疗反应（不仅仅是生长）的特定参与者，并结合后续的组间比较逻辑来识别反应者和无反应者之间存在差异的特征；(b) 儿童/家庭/背景特征与随机分配的组成员身份之间的统计交互作用。本文还提供了选择治疗反应潜在预测因素的原则。
