Developing Optimal Dynamic Behavioral Intervention in Community-Based Studies.

描述(申请人提供)：中风的预防可以通过改变生活方式来实现，比如身体活动和服药依从性。因此，开发和传播行为干预计划作为预防中风的公共卫生措施的重要性怎么强调都不为过。出于同样的原因，行为干预计划自然包括解决各种问题的多个组成部分；一个成功的多组成部分计划很可能是根据中间健康结果以最佳顺序实施每个干预组成部分的直接结果，从而最大化最终的健康结果，如在12个月内降低血压。这种类型的治疗方案根据个人的特点定制干预序列，有时被称为动态治疗方案(DTR)。这项研究旨在开发、验证和传播统计方法，通过精心设计的随机社区研究来确定最佳DTR。我们计划分四步实现这一研究目标。首先，我们将开发一种名为Q-学习的数据分析技术，它将使我们能够使用基于社区的研究数据以公正的方式确定最佳DTR。Q-学习是一种起源于计算机科学文献的尖端技术；这项研究将把这种创新的想法应用到临床应用中，其中数据是以高水平的变异性(噪声)观察的。其次，我们将开发统计设计，以促进通过Q学习发现最佳DTR，同时使试验参与者受益。这将涉及两个临床试验设计概念的新综合：序贯多分配随机试验(SMART)和自适应随机化(AR)。第三，我们将通过计算机模拟和实际行为干预研究的数据分析来验证所提出的理论和方法。第四，我们将通过构建可供公众访问的软件来传播这些方法，并将这些方法用于下一阶段干预研究的规划；这一步骤旨在结束新方法与其临床应用之间的滞后时间。我们的长期公共卫生目标是提高开发最佳行为干预课程的能力。 与公共卫生相关：中风是全球主要残疾的主要原因和第三大死亡原因，可以通过改变生活方式来预防。因此，开发和传播有效的行为干预计划对预防中风非常重要。这项研究旨在通过精心设计的随机社区研究来扩展我们的统计能力，以开发最佳的、个性化的行为干预课程。

DESCRIPTION (provided by applicant): Stroke prevention may be achieved through lifestyle changes on a variety of issues such as physical activities and medication adherence. It is therefore difficult to overstate the importance of developing and disseminating behavioral intervention programs as a public health measure to prevent strokes. For the same reason, a behavioral intervention program naturally involves multiple components addressing the various issues; and a successful multi-component program is likely a direct result of administering each interventional component in an optimal sequence, based on the intermediate health outcomes, so as to maximize the eventual health outcome such as blood pressure reduction over 12 months. This type of treatment program tailors the intervention sequence according to an individual's own characteristics, and is sometimes called dynamic treatment regime (DTR). This research aims to develop, validate, and disseminate statistical methods to identify optimal DTR through carefully designed randomized community-based studies. We plan to achieve this research goal in four steps. First, we will develop a data analytical technique, called Q-learning, that will enable us to identify an optimal DTR in an unbiased fashion using data from community- based studies. Q-learning is a cutting-edge technique originating from the computer science literature; this research will adapt this innovative idea to clinical applications where data are observed with high level of variability (noise). Second, we will develop statistical designs that facilitate the discovery of optimal DTR through Q-learning while benefiting the trial participants. This will involve novel synthesis of two clinical trial design concepts: sequential multiple assignment randomized trial (SMART) and adaptive randomization (AR). Third, we will validate the proposed theory and methods by using computer simulation and analyzing data from an actual behavioral intervention study. Fourth, we will disseminate the methods by building software with public access and employ the methods in the planning of the next stage of intervention study; this step is intended to close the lag time between novel methods and its clinical applications. Our long-term public health goal is to enhance the capability of developing optimal behavioral intervention curriculums. PUBLIC HEALTH RELEVANCE: Stroke, the leading cause of major disability and the third leading cause of death worldwide, can be prevented through lifestyle changes. It is therefore important to develop and disseminate effective behavioral intervention programs for stroke prevention. This research aims to extend our statistical capacity to develop optimal, personalized behavioral intervention curriculums through carefully designed randomized community-based studies.