Department of Civil and Environmental Engineering, A. James Clark School of Engineering, University of Maryland, College Park, MD, USA.
Department of Veterinary and Biomedical Sciences, College of Agricultural Sciences, The Pennsylvania State University, University Park, PA, USA; Department of Statistics, Eberly College of Science, The Pennsylvania State University, University Park, PA, USA.
J Biomed Inform. 2022 May;129:104054. doi: 10.1016/j.jbi.2022.104054. Epub 2022 Mar 21.
Vaccination is the most effective way to provide long-lasting immunity against viral infection; thus, rapid assessment of vaccine acceptance is a pressing challenge for health authorities. Prior studies have applied survey techniques to investigate vaccine acceptance, but these may be slow and expensive. This study investigates 29 million vaccine-related tweets from August 8, 2020 to April 19, 2021 and proposes a social media-based approach that derives a vaccine acceptance index (VAI) to quantify Twitter users' opinions on COVID-19 vaccination. This index is calculated based on opinion classifications identified with the aid of natural language processing techniques and provides a quantitative metric to indicate the level of vaccine acceptance across different geographic scales in the U.S. The VAI is easily calculated from the number of positive and negative Tweets posted by a specific users and groups of users, it can be compiled for regions such a counties or states to provide geospatial information, and it can be tracked over time to assess changes in vaccine acceptance as related to trends in the media and politics. At the national level, it showed that the VAI moved from negative to positive in 2020 and maintained steady after January 2021. Through exploratory analysis of state- and county-level data, reliable assessments of VAI against subsequent vaccination rates could be made for counties with at least 30 users. The paper discusses information characteristics that enable consistent estimation of VAI. The findings support the use of social media to understand opinions and to offer a timely and cost-effective way to assess vaccine acceptance.
接种疫苗是提供针对病毒感染的持久免疫力的最有效方法；因此，快速评估疫苗接受度是卫生当局面临的紧迫挑战。先前的研究已经应用调查技术来研究疫苗接受度，但这些方法可能既缓慢又昂贵。本研究调查了 2020 年 8 月 8 日至 2021 年 4 月 19 日的 2900 万条与疫苗相关的推文，并提出了一种基于社交媒体的方法，该方法可以得出疫苗接受指数（VAI），以量化 Twitter 用户对 COVID-19 疫苗接种的看法。该指数是根据借助自然语言处理技术确定的意见分类计算得出的，为美国不同地理尺度的疫苗接受程度提供了定量指标。VAI 可以根据特定用户和用户组发布的正面和负面推文的数量轻松计算得出，它可以针对县或州等地区进行编制，以提供地理空间信息，并且可以随着时间的推移进行跟踪，以评估疫苗接受度随媒体和政治趋势的变化。在国家层面，它表明 VAI 在 2020 年从负面转为正面，并在 2021 年 1 月后保持稳定。通过对州和县一级数据的探索性分析，可以对至少有 30 名用户的县进行 VAI 与后续疫苗接种率的可靠评估。本文讨论了能够一致估计 VAI 的信息特征。研究结果支持使用社交媒体来了解意见，并提供及时且具有成本效益的方法来评估疫苗接受度。
