Toward optimizing digital mental health interventions: A clinical trial aimed at understanding what drives patient engagement.

项目摘要 抑郁和焦虑是高度共病和昂贵的疾病。循证心理治疗是一线 治疗，但未得到充分利用，不可扩展。数字心理健康干预（DMHI），通过 因特网和/或移动的应用程序已经发展成为有效的和潜在可扩展的治疗。然而，迄今为止， 常规护理的有效性受到患者参与不足的限制。为了实现变革性的 DMHI的潜力，我们必须确定战略，让病人参与，而不增加人类的支持， 这将限制可扩展性。自动激励推送消息（AMM）和轻触式人工教练 支持（CS）提供了两种这样策略。拟议的研究测试这些策略，同时绘制初步的 关于DMHI参与的假设模型的结论，该模型基于技术采用和治疗 坚持文学。该模型假定，两个系统层面的结构（社会影响和促进 条件）和三个患者水平的结构（态度，自我效能，习惯强度）驱动DMHI参与。在 研究1（N = 20），我将采用以用户为中心的设计，开发和完善一套针对三个AMM 假设患者水平的增强驱动结构（目标1）。在研究2中，我将招募N = 76名初级保健 抑郁和/或焦虑患者通过提供者转诊参加为期8周的2x2析因临床试验， 所有参与者都将获得具有已知疗效的DMHI，并随机接受参与策略 条件（即，先前验证的CS协议、新开发的AMM、两者或两者都没有）。进一步了解 AMM如何运作，AMM臂中的消息传递将是微随机的：每天参与者将被 随机分配接收或不接收消息，使得他们平均每周接收4.2条消息。微型 随机化允许关于消息传递的近期影响的因果推断（即，AMM是一种暗示， 动作）以及消息影响和上下文之间的关系（例如，一天的时间，消息传递）。 测量的结果数据将包括参与程度（可操作为DMHI使用分钟数），每周 关于五个成就驱动结构的自我报告，以及每周的临床结果自我报告。我将测试 每种策略对测量结果数据的影响（目标2），并探索 实现驱动结构和DMHI参与（目标3）。然而，将评估临床结果， 与实验治疗模型一致，该研究利用了具有已知功效的DMHI， 允许关注上游目标（患者参与），而不是临床结果本身。 总体目标是影响目标，以最终提高临床有效性。该项目将 建立我在临床试验设计方面的专业知识，并建立我在以用户为中心的设计方面的熟练程度（即，快速、原型 通过实地研究进行测试）和数据科学（即，分析密集的、相关的纵向数据）方法 常用于DMHI优化研究。研究结果将为R01的优化奠定基础 DMHI的参与，并最终有效性，当整合到日常护理。

PROJECT SUMMARY Depression and anxiety are highly comorbid and costly diseases. Evidence-based psychotherapy is the first-line treatment but is underutilized and not scalable. Digital mental health interventions (DMHIs), delivered via the internet and/or mobile apps, have evolved as efficacious and potentially scalable treatments. To date, however, effectiveness in routine care is limited by insufficient patient engagement. In order to achieve the transformative potential of DMHIs, we must identify strategies to keep patients engaged without adding human support in a form that would limit scalability. Automated motivational push messaging (AMM) and light-touch human coach support (CS) offer two such strategies. The proposed research tests these strategies, while drawing preliminary conclusions about a hypothesized model of DMHI engagement based on the technology adoption and treatment adherence literature. The model posits that two systems-level constructs (social influence and facilitating conditions) and three patient-level constructs (attitude, self-efficacy, habit strength) drive DMHI engagement. In Study 1 (N=20), I will employ user-centered design to develop and refine a set AMMs targeting the three hypothesized patient-level engagement-driving constructs (Aim 1). In Study 2, I will recruit N=76 primary care patients with depression and/or anxiety via provider referral to an 8-week 2x2 factorial clinical trial whereby participants will all receive access to a DMHI with known efficacy and be randomized to an engagement strategy condition (i.e., a previously-validated CS protocol, newly-developed AMM, both or neither). To further understand how AMMs function, message delivery in the AMM arms will be micro-randomized: each day participants will be randomized to receive a message or not, such that they receive an average of 4.2 messages/week. Micro- randomization allows causal inference about the near-term impact of message delivery (i.e., are AMMs a cue to action) and the relationship between message impact and context (e.g., time of day the message is delivered). Measured outcome data will comprise level of engagement (operationalized as minutes of DMHI use), weekly self-reports on the five engagement-driving constructs, and weekly self-reports of clinical outcomes. I will test im pacts of each strategy on m easured outcom e data (Aim 2) and explore the hypothesized relationships between engagement-driving constructs and DMHI engagement (Aim 3). Clinical outcomes will be assessed, however, consistent with the experimental therapeutics model, this research leverages a DMHI with known efficacy, allowing the focus to be an upstream target (patient engagement) rather than the clinical outcomes themselves. The overarching goal is to influence the target so as to ultimately enhance clinical effectiveness. This project will build my expertise in clinical trial design and build my proficiency in user-centered design (i.e., rapid, prototype testing via field studies) and data science (i.e., analysis of intensive, correlated longitudinal data) methods commonly applied in DMHI optimization research. Findings will lay a foundation for R01s aimed at optimizing DMHIs for engagement, and ultimately effectiveness, when integrated into routine care.