Dusseldorp Elise, van Genugten Lenneke, van Buuren Stef, Verheijden Marieke W, van Empelen Pepijn
Expertise Group Life Style, Netherlands Organization for Applied Scientific Research (TNO).
Expertise Group Life Style, Netherlands Organization for Applied Scientific Research.
Health Psychol. 2014 Dec;33(12):1530-40. doi: 10.1037/hea0000018. Epub 2013 Nov 25.
Many health-promoting interventions combine multiple behavior change techniques (BCTs) to maximize effectiveness. Although, in theory, BCTs can amplify each other, the available meta-analyses have not been able to identify specific combinations of techniques that provide synergistic effects. This study overcomes some of the shortcomings in the current methodology by applying classification and regression trees (CART) to meta-analytic data in a special way, referred to as Meta-CART. The aim was to identify particular combinations of BCTs that explain intervention success.
A reanalysis of data from Michie, Abraham, Whittington, McAteer, and Gupta (2009) was performed. These data included effect sizes from 122 interventions targeted at physical activity and healthy eating, and the coding of the interventions into 26 BCTs. A CART analysis was performed using the BCTs as predictors and treatment success (i.e., effect size) as outcome. A subgroup meta-analysis using a mixed effects model was performed to compare the treatment effect in the subgroups found by CART.
Meta-CART identified the following most effective combinations: Provide information about behavior-health link with Prompt intention formation (mean effect size ḡ = 0.46), and Provide information about behavior-health link with Provide information on consequences and Use of follow-up prompts (ḡ = 0.44). Least effective interventions were those using Provide feedback on performance without using Provide instruction (ḡ = 0.05).
Specific combinations of BCTs increase the likelihood of achieving change in health behavior, whereas other combinations decrease this likelihood. Meta-CART successfully identified these combinations and thus provides a viable methodology in the context of meta-analysis.
许多促进健康的干预措施结合了多种行为改变技术（BCTs）以实现效果最大化。虽然从理论上讲，行为改变技术可以相互促进，但现有的荟萃分析未能确定能产生协同效应的特定技术组合。本研究通过以一种特殊方式（称为Meta-CART）将分类与回归树（CART）应用于荟萃分析数据，克服了当前方法中的一些缺点。目的是确定能解释干预成功的行为改变技术的特定组合。
对米基、亚伯拉罕、惠廷顿、麦卡蒂尔和古普塔（2009年）的数据进行了重新分析。这些数据包括针对身体活动和健康饮食的122项干预措施的效应量，以及将这些干预措施编码为26种行为改变技术。以行为改变技术作为预测变量，治疗成功（即效应量）作为结果进行CART分析。使用混合效应模型进行亚组荟萃分析，以比较CART发现的亚组中的治疗效果。
Meta-CART确定了以下最有效的组合：提供行为与健康关联信息并促使意图形成（平均效应量ḡ = 0.46），以及提供行为与健康关联信息并提供后果信息及使用跟进提示（ḡ = 0.44）。最无效的干预措施是那些在不使用提供指导的情况下提供绩效反馈的措施（ḡ = 0.05）。
行为改变技术的特定组合增加了实现健康行为改变的可能性，而其他组合则降低了这种可能性。Meta-CART成功地识别了这些组合，因此在荟萃分析的背景下提供了一种可行的方法。
