Digital Markers in Relapse and Recovery

关于复发的文献包括许多方法上的不一致，在复发的定义、评估方法和复发相关因素的模型方面存在很大差异。收集复发数据的传统方法包括：1）回顾性审查，要求参与者回忆复发的情况和发生在他们之前的因素; 2）前瞻性报告，在基线或定期收集有关潜在前因的信息，然后检查与检测到的复发的关联;和3）接近实时的报告，其中要求或电子提示参与者报告接近实际复发时间的因素。缺乏使用日常生活行为观察的研究主要是因为收集这些数据几乎是不可能的。 然而，近实时报告是最佳的，因为复发脆弱性因素，如自我效能，药物线索，焦虑，压力，药物渴望和社会支持可以在几分钟内发生变化。在这些挑战的背景下，该项目将实时报告作为一种工具，用于检测和预测参加药物使用治疗计划的患者和正在康复的患者的复发。 公共卫生研究和实践刚刚开始利用通信媒体的新兴变化，通过使用分析社交媒体语言和智能手机和可穿戴设备生成的数据的方法和工具。该项目将采用先进的数据分析技术，以检查物质使用治疗中的个人和正在经历长期恢复的个人留下的数字足迹。我们将使用自然语言处理和机器学习技术来构建预测未来复发和长期恢复的模型。被动测量将用于捕获比通常使用传统方法实现的更精细细节级别的行为数据。 大多数复发预防方法仅利用通常通过调查和访谈收集的有关参与者的可用信息的一小部分。即使复发风险随着时间的推移被反复测量，复发脆弱性通常也是基于最后一次可用的测量。然而，这种方法丢弃了关于复发脆弱性因素的动态变化性质的有价值的信息，并且没有使用来自恢复中的其他患者的信息来改进预测。该项目将导致与复发风险快速变化相关的复发脆弱性的动态实时预测。 我们在这个实验室的长期目标是开发一个自动化的，连续的系统，用于监测数字来源（社交媒体语言，智能手机传感器数据，可穿戴设备的数据），以预测每日复发脆弱性评分。然后，我们将开发一个复发脆弱性反馈工具，供成瘾治疗提供者、治疗中的人和康复中的人使用。这将通过开发在患者处于危险时自动干预的应用程序来实现临床研究和实践的新方法的使用。

The literature on relapse includes numerous methodological inconsistencies, with wide variation in the definition of relapse, assessment methodologies, and models of relapse-related factors. Traditional methodologies for collecting data on relapse included: 1) retrospective reviews in which participants are asked to recall instances of relapse and the factors preceding them; 2) prospective reports, in which information about potential antecedents is collected at baseline or periodically and then examined for association with a detected relapse; and 3) near real-time reports, in which participants are asked or electronically prompted to report on factors near the actual time of relapse. The lack of research using behavioral observation of daily life is mainly because collecting this data has been almost impossible. However, near real-time reports are optimal because relapse vulnerability factors such as self-efficacy, drug cues, anxiety, stress, drug craving, and social support can change over a period of a few minutes. In the context of these challenges, this project will real-time reports as a tool to detect and predict relapse in patients attending substance use treatment programs and in patients who are in recovery. Public health research and practice are just beginning to taken advantage of emerging changes in communication media by using methods and tools that analyze social media language and data generated from smartphones and wearable devices. This project will adapt advanced data analytic techniques to examine the digital footprints left by individuals in substance use treatment and in those who are experiencing long-term recovery. We will use natural language processing and machine learning techniques to build models that predict future relapse and long-term recovery. Passive measurement will be used to capture behavioral data at a finer level of detail than is typically achieved using conventional methods. The majority of relapse prevention approaches utilize only a fraction of the available information about a participant typically gathered through surveys and interviews. Even when relapse risk is measured repeatedly over time, relapse vulnerability is typically based on the last available measurement. However, this approach discards valuable information on the dynamically changing nature of relapse vulnerability factors and does not use information from other patients in recovery to improve predictions. This project will result in the dynamic, real-time predictions of relapse vulnerability linked to rapid changes in relapse risk. Our long-term goal in this lab is to develop an automated, continuous system for monitoring digital sources (social media language, smartphone phone sensor data, data from wearable devices) to forecast daily relapse vulnerability scores. We will then develop a relapse vulnerability feedback tool to be used by addiction treatment providers, people in treatment, and people in recovery. This will enable the use of novel approaches to clinical research and practice by developing applications that automatically intervene when a patient is at risk.