Using social media data for e-cigarette surveillance and policy research

社交媒体数据为公共卫生研究人员提供了前所未有的机会，以了解与酒精，烟草和其他药物（ATOD）相关的新兴使用模式，后果和背景因素。电子烟在不受管制的环境中迅速扩散，尽管其安全性的证据有限，但成年人和青年人越来越多地使用电子烟。这项拟议中的研究考察了Twitter上分享的电子烟信息的性质，并将个人层面的Twitter数据与调查数据联系起来，以研究个人接触和分享这些信息的程度，以及社交媒体帖子是否反映了实际的看法和行为。结果将告知我们对电子烟的理解以及社交媒体数据在监控中的效用 和ATOD的调节。目标。目标1：使用文本挖掘和情感分析，描述Twitter上分享的有关电子烟的信息，以了解（a）消费者的看法和使用;（B）广告商的营销策略，包括垃圾邮件行为;以及（c）改变联邦，州或地方法律。目标二：将个人层面的Twitter数据与自我报告的调查数据关联起来，以（a）描述当前吸烟者和电子烟用户可能接触和分享的电子烟信息的数量和类型，（B）评估信息接触是否与回忆和信息分享相关，（c）验证Twitter上分享的信息是否反映了实际的感知和使用。目标1：将使用迭代更新的搜索语法来识别有关电子烟的推文，以捕捉不断变化的电子烟术语。文本挖掘和情感分析将使用监督和非监督算法进行。将采用检测推特行为、网络和帖子内容中异常模式的计算方法来识别垃圾邮件或假冒消费者。第一年开发的算法将在第二年和第三年每季度进行改进，以表征新兴主题。目标二：采用旋转面板设计，将招募1，800名成年吸烟者和1，800名电子烟和香烟双重使用者的全国便利样本，以完成六项基线和后续调查，其中两项是专注于政策或监管行动的快速反应调查。我们将分析受访者的公开Twitter帖子及其Twitter网络，以及他们可能接触和分享的电子烟信息的程度和类型。该调查将衡量参与者对Twitter网络中电子烟信息的回忆，电子烟相关的感知和使用以及戒烟行为。我们将把Twitter数据与调查数据联系起来，以验证社交媒体帖子，并评估信息曝光是否与回忆和信息共享有关，以及这些模式和关系如何因受访者特征而异（例如，电子烟使用、戒烟意向、社交媒体使用）。

DESCRIPTION (provided by applicant): Social media data give public health researchers unprecedented opportunity to understand emerging use patterns, consequences, and contextual factors related to alcohol, tobacco, and other drugs (ATOD). Electronic cigarettes (e-cigarettes) are rapidly proliferating in an unregulated environment and increasingly used by adults and youth despite limited evidence of their safety. The proposed study examines the nature of e-cigarette information shared on Twitter and links individual-level Twitter data to survey data to examine the extent to which individuals are exposed to and share this information and whether social media posts reflect actual perceptions and behaviors. Results will inform our understanding of e-cigarettes and the utility of social media data for surveillance and regulation of ATOD. Aims. Aim 1: Using text mining and sentiment analysis, characterize information about e-cigarettes shared on Twitter to understand (a) consumer perceptions and use; (b) advertisers' marketing strategies, including spamming behavior; and (c) changing federal, state, or local laws. Aim 2: Link individual-level Twitter data to self-reported survey daa to (a) describe the amount and type of e-cigarette information that current smokers and e-cigarette users are potentially exposed to and share, (b) assess whether information exposure is associated with recall and information sharing, and (c) validate whether information shared on Twitter reflects actual perceptions and use. Aim 1: Tweets about e-cigarettes will be identified using an iteratively updated search syntax to capture the changing terminology of e-cigarettes. Text mining and sentiment analysis will be conducted using supervised and unsupervised algorithms. Computational methods that detect abnormal patterns in tweeting behavior, networks, and post content will be employed to identify spamming or fake consumers. The algorithms developed in Year 1 will be refined quarterly in Years 2 and 3 to characterize emerging topics. Aim 2: Using a rotating panel design, a national convenience sample of 1,800 adult current smokers and 1,800 dual e-cigarette and cigarette users will be recruited to complete six baseline and follow-up surveys, with two being rapid response surveys focused on policy or regulatory action. We will analyze public Twitter posts of respondents and their Twitter networks and the extent and type of e-cigarette messages they were potentially exposed to and shared. The survey will measure participants' recall of e-cigarette messages from their Twitter network, e-cigarette-related perceptions and use, and cessation behaviors. We will link Twitter data to survey data to validate social media posts and assess whether information exposure is associated with recall and information sharing and how these patterns and relationships vary by respondent characteristics (e.g., e-cigarette use, quit intentions, social media use).