DISC: Describe Smoking Cessation in RCT Multi-Component Behavioral Intervention

描述(申请人提供)：行为干预通常用于促进戒烟。它们通常有多个组件，并随着时间的推移而实施。吸烟者的参与和反应行为在干预过程中发生变化，导致结果的个体差异很大。然而，在纵向多成分干预中，描述吸烟者复杂行为的方法还不够成熟。基于互联网和面对面的文化定制干预是两种很有前途的干预措施， 但相对未经探索的行为干预。第一种方法对普通吸烟人群来说具有成本效益，但我们对如何通过干预充分衡量个人的动态在线参与度或检查其有效性知之甚少。第二个目标是特定人群，但我们需要了解种族/民族群体如何应对这种干预，以及文化定制在多大程度上是有用的。我们提出了一种新的模式识别方法来刻画基于互联网和文化定制的干预过程中复杂的参与/反应行为。我们的方法建立在PI的初步吸烟行为研究基础上， 她开发了一种基于多重归因的模糊聚类模型(MI-模糊)来识别怀孕吸烟行为模式，并处理吸烟者在多个聚类中具有成员资格并且他们的吸烟数据是纵向的、非正常的、高维的并且包含许多缺失值的现实世界情况。在这里，我们将增强MI-模糊的新功能，将其与典型模型进行比较，并将我们的模式方法扩展到两个纵向行为干预研究：(1)休斯顿博士针对普通吸烟人群的由NCI资助的大规模戒烟互联网干预，(2)由NIDA博士资助的针对少数吸烟人群的小规模认知、文化定制、基于临床的TDTA干预。我们将表征吸烟者的在线参与(Quit-Primo)和认知反应(TDTA)，评估干预措施的组成部分如何对不同的吸烟者起作用，阐明其有效性，并提供新的、详细的理解，了解吸烟者的轨迹模式如何与不同的戒烟结果相关。更好地了解吸烟者如何参与和应对干预措施，将有助于发现传统方法所遗漏的重要关系，为如何针对目标人群改进这些干预措施以及可能对早期干预具有临床重要性的高危行为模式提供新的证据。检查不同类型的行为干预也将有助于将我们的模式方法推广到其他物质使用研究和人群。通过提供分析原型和可获得的工具，本研究将推进一般模式识别方法，并加速其在物质使用行为研究中的应用。随着我们传播活动的扩大，这项工作可能会刺激类似的研究，以寻求更好的和有针对性的干预，最终受益于与药物使用相关的以患者为中心的护理。

DESCRIPTION (provided by applicant): Behavioral interventions are commonly used to promote smoking cessation. They typically have multiple components and are implemented over time. Smokers' engagement and response behaviors change over the course of interventions, resulting in substantial individual variations in outcomes. However, methods are underdeveloped for characterizing smokers' complex behaviors in longitudinal multi-component interventions. Internet-based and face-to-face culturally-tailored interventions are two promising, but relatively unexplored, behavioral interventions. The first is cost effective for reaching generl smoking populations, yet we know little about how to adequately measure individuals' dynamic online engagement with an intervention or examine its efficacy. The second targets specific populations, but we need to learn how racial/ethnic groups respond to such interventions and how much cultural tailoring is useful. We propose a new pattern-recognition approach to characterize complex engagement/response behaviors during Internet-based and culturally tailored interventions. Our approach is built on the PI's preliminary smoking behavior studies, for which she developed a multiple-imputation-based fuzzy clustering model (MI-Fuzzy) to identify pregnancy smoking behavioral patterns, and to cope with real-world situations where smokers have memberships in multiple clusters and their smoking data are longitudinal, non-normal, high dimensional and contain many missing values. Herein, we will enhance MI-Fuzzy with new features, compare it to typical models, and expand our pattern approach to two longitudinal behavioral intervention studies: (1) Dr. Houston's large-scale NCI-funded, Quit-Primo Internet intervention for a general smoking population, and (2) Dr. Kim's small-scale NIDA-funded cognitive, culturally tailored, clinic-based TDTA intervention for a minority smoking population. We will characterize smokers' online engagement (Quit-Primo) and cognitive responses (TDTA), evaluate how the interventions' components work for different smokers, clarify their efficacy, and provide a new, detailed understanding of how smokers' trajectory patterns relate to different cessation outcomes. Better understanding of how smokers engage with and respond to interventions will help uncover important relationships missed by traditional approaches, yield new evidence on how to improve these interventions for targeted populations and on high-risk behavioral patterns that may be clinically important for early intervention. Examining different types of behavioral interventions will also facilitate generalizing our pattern approach to other substance-use studies and populations. By providing analytical prototypes and accessible tools, this study will advance general pattern recognition methodology, and accelerate its utility in behavioral studies of substance use. As our dissemination activities expand, this work will likely stimulate similar studies for better and targeted interventions, ultimately benefiting patient-centered care related to substance use.