Using Natural Language Processing and Crowdsourcing to Monitor and Evaluate Public Information and Communication Disparities about Colon Cancer Screening

项目摘要 美国黑人的结直肠癌（CRC）发病率和死亡率高于非西班牙裔 白色美国人。虽然与《儿童权利公约》有关的一些差异有所减少（例如，发病率和分期 尽管存在差异，但在CRC筛查（CRCS）的背景下仍然存在差异。研究表明，支持和 信息丰富的社交网络，包括在线和离线，可以改善美国黑人的CRCS。一 越来越多的证据表明在线健康信息来源的重要性， 扫描有关CRC和CRCS，但很少有人知道的影响，个人的信息是 在这些平台上相遇。白色和黑色信息的内容和数量研究 美国人从网上健康信息来源遇到的问题仍然不清楚，特别是关于任何 关于美国黑人寻求、扫描或分享哪些特定信息的差异。那里 是了解哪些信息在特定疾病风险人群中产生共鸣的关键需求（例如， CRC）以及可能对参与早期检测行为（例如，CRCS），并可能面临 在接触网上公共信息方面存在差异。该项目利用和应用新型癌症 通信监视方法（例如，自然语言处理和众包）来检查 关于CRC预防和控制的公共卫生宣传。目标1将使用计算，自然 语言处理方法，以捕获和分析有关CRC的数字和社交媒体信息， CRCS将确定主要信息、错误信息的来源和类型以及信息不平等。这 方法提供了一种高效、有效和反应灵敏的方法来监控（mis）信息和新兴的 关于CRCS的信息目标2将使用众包方法（维基调查）来评估人口 公众对CRCS信息和信息的看法。招募具有国家代表性的样本， 白色（N = 1，000）和美国黑人（N = 1，000）45-74岁的成年人，我们将使用创新的数据收集 一种称为wiki调查的程序，用于将候选消息列为战略工作中的潜在消息目标 促进CRCS。对于目标3，我们将进行一项随机对照消息试验（N = 1，600），以确定 维基调查方法为目标受众群体选择消息的有效性。我们将使用 从研究2中收集的数据，以确定四组信息，每个目标都有强有力的论据 受众排名：两个受众的最高评级消息，目标群体内的最高评级消息， 针对目标外群体的最高评级消息，以及中等/中值评级消息。我们会越过那些 目标受众（白色/美国黑人）的消息类别，以测试是否通过wiki选择消息 调查与未来坚持筛查建议和共享CRCS的意图相关 充分和传播该项目将提供证据，以帮助确定这些新方法的有效性和可扩展性， 这对今后创新形成性研究和评价方法至关重要。

PROJECT SUMMARY Colorectal cancer (CRC) incidence and death rates are higher among Black Americans than non-Hispanic White Americans. While some CRC-related disparities have decreased (e.g., incidence and stage of presentation), disparities persist in the context of CRC screening (CRCS). Studies suggest that supportive and information-rich social networks, both online and offline, could improve CRCS among Black Americans. A growing body of evidence indicates the importance of online sources of health information seeking and scanning about CRC and CRCS, but little is known about the impact of the messages that individuals are encountering on these platforms. Research on the content and volume of messages White and Black Americans encounter from online health information sources is still unclear—particularly regarding any disparities that exist about what specific information is sought, scanned, or shared by Black Americans. There is a critical need to understand which messages resonate among populations at-risk for specific diseases (e.g., CRC) and who may have concerns about engaging in early detection behaviors (e.g., CRCS) and may face disparities in exposure to public information online. The proposed project utilizes and applies novel cancer communication surveillance approaches (e.g., natural language processing and crowdsourcing) to examine public health communication about CRC prevention and control. Aim 1 will use computational, natural language processing approaches to capture and analyze digital and social media information about CRC and CRCS to identify prominent messages, sources and types of misinformation, and information inequalities. This approach offers an efficient, effective, and responsive method to monitor (mis)information and emerging messages about CRCS. Aim 2 will use a crowdsourcing approach (wiki surveys) to assess population perceptions of public information and messages about CRCS. Recruiting nationally representative samples of White (N = 1,000) and Black American (N = 1,000) adults ages 45-74, we will use an innovative data collection procedure known as wiki surveys to rank candidate messages as potential message targets in strategic efforts to promote CRCS. For Aim 3, we will conduct a randomized controlled message trial (N = 1,600) to determine the validity of the wiki survey approach to selecting messages for targeted audience segments. We will use data collected from Study 2 to identify four sets of messages with strong arguments respective to each target audience’s rankings: highest rated messages for both audiences, highest rated messages for target in-group, highest rated messages for target out-group, and middle-/median-rated messages. We will cross those message categories with target audience (White/Black American) to test if messages selected via the wiki survey are associated with intentions to adhere to screening recommendations in the future and share CRCS messages. The project will offer evidence to help determine the validity and scalability of these novel methods, which is essential to innovate formative research and evaluation approaches in the future.