Background

The coronavirus disease (COVID-19) pandemic, which began in Wuhan, China in December 2019, is rapidly spreading worldwide with over 1.9 million cases as of mid-April 2020. Infoveillance approaches using social media can help characterize disease distribution and public knowledge, attitudes, and behaviors critical to the early stages of an outbreak.

Objective

The aim of this study is to conduct a quantitative and qualitative assessment of Chinese social media posts originating in Wuhan City on the Chinese microblogging platform Weibo during the early stages of the COVID-19 outbreak.

Methods

Chinese-language messages from Wuhan were collected for 39 days between December 23, 2019, and January 30, 2020, on Weibo. For quantitative analysis, the total daily cases of COVID-19 in Wuhan were obtained from the Chinese National Health Commission, and a linear regression model was used to determine if Weibo COVID-19 posts were predictive of the number of cases reported. Qualitative content analysis and an inductive manual coding approach were used to identify parent classifications of news and user-generated COVID-19 topics.

Results

A total of 115,299 Weibo posts were collected during the study time frame consisting of an average of 2956 posts per day (minimum 0, maximum 13,587). Quantitative analysis found a positive correlation between the number of Weibo posts and the number of reported cases from Wuhan, with approximately 10 more COVID-19 cases per 40 social media posts (P<.001). This effect size was also larger than what was observed for the rest of China excluding Hubei Province (where Wuhan is the capital city) and held when comparing the number of Weibo posts to the incidence proportion of cases in Hubei Province. Qualitative analysis of 11,893 posts during the first 21 days of the study period with COVID-19-related posts uncovered four parent classifications including Weibo discussions about the causative agent of the disease, changing epidemiological characteristics of the outbreak, public reaction to outbreak control and response measures, and other topics. Generally, these themes also exhibited public uncertainty and changing knowledge and attitudes about COVID-19, including posts exhibiting both protective and higher-risk behaviors.

Conclusions

The results of this study provide initial insight into the origins of the COVID-19 outbreak based on quantitative and qualitative analysis of Chinese social media data at the initial epicenter in Wuhan City. Future studies should continue to explore the utility of social media data to predict COVID-19 disease severity, measure public reaction and behavior, and evaluate effectiveness of outbreak communication.

Introduction

The state of California has enacted progressive anti-tobacco policies, including Proposition 56 in 2016. In response, the alternative and emerging tobacco product (ATP) industry has increased its political activity. This study explores the association between the proportion of people voting against Proposition 56 and tobacco/ATP retail density.

Methods

We conducted a retrospective analysis using data on licensed California tobacco retailers, which were then cross-referenced for categorization using Yelp. Proportion voting against Proposition 56 was obtained from the Secretary of State's website. A series of linear regression tests were performed between population-normalized retailer density and voting proportion at the county level before and after adjusting for covariates such as age, gender, race/ethnicity and median household income.

Results

The total number of licensed tobacco retailers increased by 29.31% from 2015 to 2019. Association between proportion voting against Proposition 56 and retail density was significant during voting and during periods of policy implementation and post-implementation (2016-2018) for non-specialized tobacco retailers. For specialized/ATP retailers, significance was only detected during the post-implementation period (2018-2019) after normalization. Proportion voting against Proposition 56 was also a significant predictor of increase in total number of non-specific (β=0.48, p=0.008) as well as specialized tobacco and/or ATP retail storefronts (β=0.21, p=0.001) from 2016 to 2018.

Conclusions

This study provides initial evidence of the association between tobacco retail density and voting patterns for anti-tobacco policy. Future research should examine the role of tobacco retail density on variation in local support for state tobacco control initiatives, including tailoring outreach to specific voting census blocks in communities with heavy retail presence.

Background

Increased public health and regulatory scrutiny concerning the youth vaping epidemic has led to greater attention to promotion and sales of vaping products on social media platforms.

Objectives

We used unsupervised machine learning to identify and characterise sale offers of electronic nicotine delivery systems (ENDS) and associated products on Instagram. We examined types of sellers, geographic ENDS location and use of age verification.

Methods

Our methodology was composed of three phases: data collection, topic modelling and content analysis. We used data mining approaches to query hashtags related to ENDS product use among young adults to collect Instagram posts. For topic modelling, we applied an unsupervised machine learning approach to thematically categorise and identify topic clusters associated with selling activity. Content analysis was then used to characterise offers for sale of ENDS products.

Results

From 70 725 posts, we identified 3331 engaged in sale of ENDS products. Posts originated from 20 different countries and were roughly split between individual (46.3%) and retail sellers (43.4%), with linked online sellers (8.8%) representing a smaller volume. ENDS products most frequently offered for sale were flavoured e-liquids (53.0%) and vaping devices (20.5%). Online sellers offering flavoured e-liquids were less likely to use age verification at point of purchase (29% vs 64%) compared with other products.

Conclusions

Instagram is a global venue for unregulated ENDS sales, including flavoured products, and access to websites lacking age verification. Such posts may violate Instagram's policies and US federal and state law, necessitating more robust review and enforcement to prevent ENDS uptake and access.

Introduction

Tobacco retailer density may be associated with greater youth initiation and reduced success during quit attempts; however, the extent to which tobacco retailer density has changed overtime across multiple categories of retailers has not been reported.

Methods

Data on licensed tobacco retailers within California from 2015-2019 were obtained from the California Department of Tax and Fee Administration. Store type was categorized by automated cross-referencing with Yelp. Geolocations were aggregated at county level for analyzing longitudinal trends in changes in tobacco retail density including demographic characteristics.

Results

The number of active CA tobacco retailer licenses increased from 19825 in 2015 to 25635 in 2019. The highest percent increase in tobacco retailer licenses (9.1%) was observed in 2017. The number of specialized tobacco stores was highest in Los Angeles, San Diego, and Riverside counties. We observed a significant increase in the number of active licenses for non-specialized and specialized tobacco stores, both overall and after controlling for the size of populations within each region. Time was a statistically significant predictor for the number of active licenses for only non-specialized stores, after adjusting for covariates. Regional volume of retailers was positively associated with higher proportion of women, lower median household income, and higher proportion of Hispanic residents.

Conclusions

Monitoring the changes in tobacco retail density and associated sociodemographic factors over time can help to identify communities at higher risk for tobacco and nicotine product exposure and access, and its associated health disparities.

Introduction

Early reports of COVID-19 cases and deaths may not accurately convey community-level concern about the pandemic during early stages, particularly in the United States where testing capacity was initially limited. Social media interaction may elucidate public reaction and communication dynamics about COVID-19 in this critical period, during which communities may have formulated initial conceptions about the perceived severity of the pandemic.

Methods

Tweets were collected from the Twitter public API stream filtered for keywords related to COVID-19. Using a pre-existing training set, a support vector machine (SVM) classifier was used to obtain a larger set of geocoded tweets with characteristics of user self-reporting COVID-19 symptoms, concerns, and experiences. We then assessed the longitudinal relationship between identified tweets and the number of officially reported COVID-19 cases using linear and exponential regression at the U.S. county level. Changes in tweets that included geospatial clustering were also assessed for the top five most populous U.S. cities.

Results

From an initial dataset of 60 million tweets, we analyzed 459,937 tweets that contained COVID-19-related keywords that were also geolocated to U.S. counties. We observed an increasing number of tweets throughout the study period, although there was variation between city centers and residential areas. Tweets identified as COVID-19 symptoms or concerns appeared to be more predictive of active COVID-19 cases as temporal distance increased.

Conclusion

Results from this study suggest that social media communication dynamics during the early stages of a global pandemic may exhibit a number of geospatial-specific variations among different communities and that targeted pandemic communication is warranted. User engagement on COVID-19 topics may also be predictive of future confirmed case counts, though further studies to validate these findings are needed.

Objectives

According to current reporting, the number of active coronavirus disease 2019 (COVID-19) infections is not evenly distributed, both spatially and temporally. Reported COVID-19 infections may not have properly conveyed the full extent of attention to the pandemic. Furthermore, infection metrics are unlikely to illustrate the full scope of negative consequences of the pandemic and its associated risk to communities.

Methods

In an effort to better understand the impacts of COVID-19, we concurrently assessed the geospatial and longitudinal distributions of Twitter messages about COVID-19 which were posted between March 3rd and April 13th and compared these results with the number of confirmed cases reported for sub-national levels of the United States. Geospatial hot spot analysis was also conducted to detect geographic areas that might be at elevated risk of spread based on both volume of tweets and number of reported cases.

Results

Statistically significant aberrations of high numbers of tweets were detected in approximately one-third of US states, most of which had relatively high proportions of rural inhabitants. Geospatial trends toward becoming hotspots for tweets related to COVID-19 were observed for specific rural states in the United States.

Discussion

Population-adjusted results indicate that rural areas in the U.S. may not have engaged with the COVID-19 topic until later stages of an outbreak. Future studies should explore how this dynamic can inform future outbreak communication and health promotion.

Background

The opioid epidemic has had a devastating impact on youth from American Indian and Alaska Native (AI/AN) Tribes and Villages, which also experience disparate suicide rates. The use of publicly available social media data originating from AI/AN communities may enhance public health response time to substance use disorder (SUD)-related overdose and augment Tribal public health surveillance systems, but these concepts have yet to be adequately explored. The goal of this exploratory analysis was to identify primary and secondary accounts of overdose and characterize relevant contextual factors in the AI/AN population on social media.

Methods

The Twitter application programming interface was queried for all Tweets containing geocoded data between March 2014 and June 2020 and filtered for the keyword ['overdose']. This sample of Tweets (n = 146,236) was then restricted to those geolocated from US Tribal lands (n = 619). Tweets were manually annotated for primary or secondary accounts of overdose as well as suicidal ideation, substance(s) used, stigma of drug use, and community-wide incidents.

Results

We collected a total of 146,235 tweets that were geocoded and contained the word 'overdose,' of which 9.5% were posted on Tribal lands (n = 619). 9.4% of these tweets (n = 58) met our study inclusion criteria and were mainly posted from Oklahoma (n = 26, 45%) and North Carolina (n = 13, 22.4%). Most Tweets (n = 41, 71%) described a primary account of an overdose and were mostly posted from 2014 to 2015. Less than half of the Tweets (n = 27, 46.5%) referenced a specific substance. Those substances mentioned included alcohol, marijuana, methamphetamine, heroin, laundry softener, cocaine, K2-Spice (synthetic cannabinoid), codeine, morphine, Nyquil, and Xanax.

Discussion

Though exploratory, our study identified SUD-related content self-reported by AI/AN communities on Twitter, especially in Oklahoma and North Carolina. These results may assist in the future design and detection of infodemiology trends and early warning signs that can better facilitate intervention specific to the ongoing Tribal opioid epidemic. While all data were collected from the public domain, additional care should be given to individual and community privacy.

Background

Access to tobacco products, including vape products, from local brick-and-mortar stores influences the exposure, uptake, and use of these products in local communities.

Methods

Licensed tobacco retailers in California were classified as specialized tobacco/vape stores or non-specialized stores by obtaining categories published on Yelp. California smoking and vaping prevalence data were obtained from the 500 cities project and ESRI community analyst tool respectively. A series of simple linear regression tests were performed, at the zip code level, between the retailer count in each store category and smoking/vaping population. The Getis-Ord Gi* and Anselin Local Moran's I statistics were used for characterization of tobacco retail density hotspots and cold spots.

Results

The association between CA smoking/vaping population and number of tobacco retailers was statistically significant for all store categories. Variability in smoking population was best explained by variability in non-specialized storefronts(R²=0.84). Spatial variability in tobacco-only storefronts explained the least proportion of variability in the overall smoking population. Similar results were obtained specific to vaping population, although the proportion of population explained by variability in the number of non-specialized storefronts was comparatively lower(R²=0.80).

Conclusions

Localities with greater numbers of non-specialized tobacco retailers had higher rates of smoking/vaping populations, and this association was much stronger for localities with greater numbers of specialized retailers. Non-specialized storefronts may represent convenient access points for nicotine products, while specialized storefronts may represent critical access points for initiation. Hence, regulations that address the entirety of the tobacco/vaping retail environment by limiting widespread access from non-specialized stores and reducing appeal generated by specialized retailers should be incorporated in future tobacco regulatory science and policymaking.