Artificial Intelligence in a Mobile Intervention for Depression (AIM)

描述(由申请人提供)：本提案的主要目的是开发和评估在移动干预应用程序中使用最先进的机器学习方法来治疗严重抑郁障碍(MDD)。机器学习是人工智能的一个分支，专注于开发基于收集的数据自动改进和进化的算法。机器学习模型可以学习检测数据中复杂的、潜在的模式，并将这些知识实时应用于决策。拟议的干预措施名为IntelliCare，将使用从患者和干预措施收集的持续数据 应用程序不断调整干预内容、内容形式和激励性信息，以创建高度定制和用户响应的治疗系统。行为干预技术(BITS)，包括基于网络的干预和移动干预，已经开发出来，并越来越多地被用于治疗MDD。BITS在治疗抑郁症方面是适度有效的，特别是在通过电子邮件或电话进行人工指导的情况下。然而，缺乏个性化和无法适应患者的需求或偏好，导致感觉缺乏相关性，导致较差的依从性和结果。IntelliCare将被设计为一款移动应用程序，但可以通过计算机网络浏览器和平板电脑访问。IntelliCare机器学习框架将使用从使用数据(例如使用治疗组件的时间长度)、手机中嵌入的传感器(例如GPS)和用户的自我报告(例如治疗组件的“赞”和有用性评级)获得的个人数据来提供高度定制的干预，该干预可以向患者学习，并使干预和激励材料适应患者的偏好和状态。低强度的教练将作为支持坚持的后盾。该项目将包括三个阶段。第一阶段将涉及IntelliCare的开发和通过可用性测试对其进行优化。第二阶段将对200名用户进行现场试验，他们将接受为期12周的IntelliCare。现场试验有两个目标：一是完成治疗框架的可用性测试和优化，二是开发机器学习模型和算法。第三阶段将对IntelliCare进行双盲随机对照试验，将其与MobilCare进行比较。MobilCare将与IntelliCare相同，只是它将使用标准演示 和展示，而不是机器学习，以提供治疗和激励材料。我们将从初级保健机构招募一半的参与者，因为这是美国事实上的抑郁症治疗网站，另一半是通过互联网招募的，互联网是进入健康应用程序的主要门户。将自适应机器学习分析应用于移动干预可能会创建新一代BIT，从而彻底改变此类干预的概念化、设计和部署方式。这些创新将产生广泛的后果，并可扩展到更广泛的双边投资条约，包括基于网络的干预措施，以及针对广泛的健康和精神健康问题的其他干预措施。

DESCRIPTION (provided by applicant): The primary aim of this proposal is to develop and evaluate the use of state of the art machine learning approaches within a mobile intervention application for the treatment of major depressive disorder (MDD). Machine learning, a branch of artificial intelligence, focuses on the development of algorithms that automatically improve and evolve based on collected data. Machine learning models can learn to detect complex, latent patterns in data and apply such knowledge to decision making in real time. The proposed intervention, called IntelliCare, will use ongoing data collected from the patient and intervention application to continuously adapt intervention content, content form, and motivational messaging to create a highly tailored and user-responsive treatment system. Behavioral intervention technologies (BITs), including web-based and mobile interventions, have been developed and are increasingly being used to treat MDD. BITs are moderately effective in treating depression, particularly when guided by human coaching via email or telephone. However, lack of personalization and inability to adapt to patient needs or preferences, which results in a perceived lack of relevance, contributes to poorer adherence and outcomes. IntelliCare will be designed as a mobile application, but will be accessible via computer web browsers and tablets. The IntelliCare machine learning framework will use individual data obtained from use data (e.g., length of time using a treatment component), embedded sensors in the phone (e.g., GPS), and the user's self-reports (e.g., "like" and usefulness ratings of treatment components) to provide a highly tailored intervention that can learn from the patient and adapt intervention and motivational materials to the patient's preferences and state. Low intensity coaching will serve as a backstop to support adherence. This project will contain three phases. Phase 1 will involve the development of IntelliCare and its optimization through usability testing. Phase 2 will be a field trial of 200 users who will receive IntelliCare for 12 weeks. The field trial has two aims: first to complete usability testing and optimization of the treatment framework, and second to develop the machine learning models and algorithms. Phase 3 will subject IntelliCare to a double blind, randomized controlled trial, comparing it to MobilCare. MobilCare will be identical to IntelliCare except that it will use standard presentation and presentation, rather than machine learning, to provide treatment and motivational materials. We will recruit half the participants from primary care settings, as this is the de facto site for treatment of depression in the United States, and half through the Internet, which is the main portal to health apps. The application of adaptive machine learning analytics to a mobile intervention has the potential to create a new generation of BITs that could revolutionize the way that such interventions are conceptualized, designed, and deployed. These innovations would have broad consequences and could be extended a broader range of BITS, including web- based interventions, and to other interventions targeting a wide range of health and mental health problems.