SCH: INT: Collaborative Research: Enhancing Context-Awareness and Personalization for Intensively Adaptive Smoking Cessation Messaging Interventions

根据疾病控制中心的数据，烟草使用仍然是美国主要的可预防死亡原因，每年造成约48万人死亡，并产生超过1500亿美元的医疗费用。因此，显然需要可扩展的，低成本的，有效的戒烟干预措施。这项研究旨在开发和验证新的基于信息的戒烟支持干预系统，该系统将利用智能可穿戴技术的最新进展来显着提高疗效。该系统将集成可穿戴传感器，持续估计个人的压力和渴望程度以及吸烟的发生。这些信息将用于增强干预系统的环境意识，使其能够不断调整每个人的干预组件的内容和交付时间。通过开发可扩展的基于信息的戒烟支持干预措施，改善个人相关性，这项研究有可能通过更有效地帮助个人戒烟，为社会带来直接利益。为了实现提供有效的个性化戒烟干预措施的目标，本研究将开发和评估支持所需的上下文推断，个性化和交付时间优化所需的模型，算法和可穿戴手机云计算基础设施。从该团队广泛的先前工作开始，这项研究将有助于（1）低成本，低功耗传感器的移动的健康感测和上下文推理的进步;（2）用于个性化上下文推理模型的实时，基于流的主动学习的进步;（3）基于推断上下文的个性化消息选择的上下文推荐系统的进步;以及（4）强大的实时可穿戴手机云数据分析系统的进步。这项工作还将通过开源软件版本为增强研究基础设施做出重大贡献，研究界可以利用这些软件版本在其他备受瞩目的健康领域（包括心脏病，肥胖和成瘾）产生效益。

According to the Centers for Disease Control, tobacco use remains the leading preventable cause of death in the US, causing approximately 480,000 deaths each year, and incurring over $150 billion in healthcare costs. As a result, scalable, low cost, and effective smoking cessation interventions are clearly needed. This research aims to develop and validate new messaging-based smoking cessation support intervention systems that will leverage recent advances in smart wearable technologies to significantly enhance efficacy. The system will integrate wearable sensors that continuously estimate an individual's level of stress and craving as well as the occurrence of smoking. This information will be used to enhance the context awareness of the intervention system, allowing it to continuously adapt both the content and delivery timing of intervention components for each individual. By developing scalable messaging-based smoking cessation support interventions with improved personal relevance, this research has the potential to lead to direct benefits to society by more effectively helping individuals to quit smoking. To accomplish the goal of providing effective, personalized smoking cessation interventions, this research will develop and evaluate the models, algorithms, and wearable-phone-cloud computational infrastructures required to support the context inferences, personalization, and delivery timing optimizations required. Starting from the team's extensive prior work, this research will contribute to (1) advances in mobile health sensing and context inference with low-cost, low-power sensors; (2) advances in real-time, stream-based active learning for personalizing context inference models; (3) advances in contextualized recommender systems to personalize message selection based on inferred contexts; and (4) advances in robust, real-time wearable-phone-cloud data analytics systems. This work will also make substantial contributions to enhancing research infrastructure through open source software releases that can be leveraged by the research community to yield benefits in other high-profile health areas including heart disease, obesity, and addiction.