The 2014 Ebola outbreak was first recognized during March 2014 within West Africa, which became the largest Ebola outbreak since the virus was discovered. Shortly thereafter, the US Centers for Disease Control and Prevention (CDC) confirmed three cases of Ebola in Dallas, TX and one case in New York, NY. 1 During the outbreak, there was a surge in interest about Ebola online. Ebola was the third most searched term in the United States on Google in 2014.Reference Alicino, Bragazzi and Faccio 2 Similarly, tweets mentioning Ebola peaked multiple times in the United States throughout the outbreak on Twitter.Reference Fung, Tse and Cheung 3 Despite high levels of online conversation about Ebola, numerous studies have indicated that the discussion of Ebola on news and social media may have prompted fears and misinformation among the public.Reference Fung, Tse and Cheung 3
Local health departments (LHDs) are typically considered the first line of defense in protecting the health of their communities by providing information about health issues, and preparing for and responding to public health emergencies.Reference Duchin 4 One way LHDs communicate with their constituents is through social media, such as posting on Facebook or sending a tweet via Twitter to quickly disseminate information.Reference Harris, Mueller and Snider 5 From our previous study,Reference Wong, Harris and Staub 6 about 77% of LHDs using Twitter had tweeted at least once about Ebola for a total of 1648 Ebola-related tweets during the data collection period. Among the LHDs that tweeted about Ebola, there was an average of 7 Ebola tweets per health department, with most tweets providing information about the virus or directing people to resources that would enable them to learn more about Ebola through a web site or an infographic.Reference Wong, Harris and Staub 6
Our previous study indicated that being in geographic proximity to the nearest Ebola case (Texas or New York) was not significantly associated with LHDs tweeting about EbolaReference Wong, Harris and Staub 6 ; however, we did not examine how the geospatial distribution of LHD tweets aligned with LHD constituent interest in Ebola. This is an important topic for public health as LHDs may not be sending messages that align with their constituents’ current public health concerns. This study aims to answer three primary research questions: (1) What was the geospatial distribution of LHDs tweeting about Ebola? (2) What was the geospatial distribution of online searches about Ebola in the United States? and (3) What is the association between the geospatial distribution of Ebola tweets from LHDs and online search activity about Ebola?
Methods
Between September 3 and November 2, 2014, we used the NCapture tool of NVivo 10 to collect all tweets sent by 287 LHDs known to be using Twitter (bit.ly/1FrqEZe). The tool retrieved information such as coordinates, dates, and Twitter profile descriptions for each tweet sent by LHDs. Ebola-related tweets were identified by searching for the term “Ebola” in the tweets collected. Online search activity about Ebola was retrieved from Google Trends (bit.ly/1IiN9Oa), restricted by the search term “Ebola,” and limited to searches that took place in the United States between September 1 and October 31, 2014. Google Trends assigns each state a relative search volume (RSV) on a scale between 0 and 100, which enables a comparison of search popularity between states. The search data are adjusted by geographical location and time range, relative to the data point with the highest search queries. For example, the state with the highest frequency of Ebola searches would be assigned an RSV value of 100, whereas other states would be assigned RSV values relative to this state.
The geospatial distribution of LHD Ebola tweets was analyzed by importing coordinates from each tweet into ArcGIS 10.2.2. A portion of the tweets (n=166, 10.1%) was missing coordinates. Consequently, addresses for LHDs missing coordinates were obtained by matching LHD names identified from their Twitter profiles to their addresses listed in the 2013 National Association of County and City Health Officials Profile study. 7 Missing coordinates were then obtained for these LHDs by inputting their respective addresses into Google Maps. Coordinates for one Ebola tweet, however, could not be assigned as the Twitter account was a collaboration of three LHDs in separate locations. A Pearson’s correlation test was performed in SPSS 23 to assess the relationship between Google RSV values and per capita number of LHD Ebola tweets by state. Per capita number of LHD Ebola tweets was computed by dividing the total number of Ebola tweets for each state by its respective total number of LHDs.
Results
Ebola Tweets
Ebola tweets from LHDs were primarily concentrated in multiple cities across the northeast region of the United States (Figure 1). The 5 cities with the highest number of Ebola tweets were Philadelphia (n=127), New York City (n=103), Cincinnati (n=95), Buffalo (n=84), and Baltimore (n=53). A high concentration of Ebola tweets was observed in New York state, where the fourth Ebola case was confirmed in New York City. In contrast, few LHDs tweeted about Ebola in Texas, where the first 3 cases were confirmed in Dallas.
Geospatial Distribution of Ebola Tweets from Local Health Departments Between September 3, 2014, and November 2, 2014.
Online Searches
From September to October 2014, Texas registered the highest number of Google searches about Ebola among all states, with an RSV value of 100. In addition, Dallas registered an RSV value of 98, the second highest number of searches about Ebola among all US cities. Other states with high RSV values for Ebola searches included Oklahoma (99), District of Columbia (89), Louisiana (80), and New York (77). There was a weak, negative, non-significant correlation (r=−.03, P=0.83, 95% CI: −0.30, 0.25) between Google RSV values and per capita number of LHD Ebola tweets by state, indicating no association between online searches for Ebola and LHDs tweeting about Ebola.
Discussion
This study examined the geospatial distribution of LHD tweets and online searches about Ebola. The findings were consistent with our previous study, in which there was no association between states with confirmed Ebola cases and LHDs tweeting about Ebola (Figure 1).Reference Wong, Harris and Staub 6 Many LHDs in cities within a 200-mile radius of New York City increased Ebola tweeting when Ebola cases were confirmed between September and October 2014 by the CDC. Few LHDs in Texas, particularly around Dallas, tweeted about Ebola although there were 3 confirmed cases.
The initial public health response to the 2014 Ebola outbreak bears some resemblance to the events that took place during the emergence of the HIV–AIDS epidemic in 1981. A common adage describing the initial public health response to the HIV–AIDS epidemic was, “What do we think? What do we know? What can we prove?”Reference Shilts 8 Although there was substantial evidence indicating a possible HIV–AIDS epidemic, public health officials intentionally delayed releasing official statements about the outbreak until they could prove that the nationwide deaths of drug addicts and homosexual men were caused by HIV–AIDS. Confirmation linking these deaths to HIV–AIDS, however, was deemed essential as it would avert unnecessary panic among the public. This public health approach of delaying communication until there is confirmed evidence continues to persist today for other emerging infectious diseases. We speculate this may also explain the low frequency of Ebola tweets in Texas during the outbreak. Although the United States anticipated a possible Ebola outbreak, LHDs in Texas likely delayed tweeting about Ebola until they could prove the initial cases in Texas were indeed caused by the Ebola virus.
A previous study suggested that there was a moderate, positive correlation between the number of new weekly global Ebola cases and weekly search activity about Ebola on Google.Reference Alicino, Bragazzi and Faccio 2 A similar situation may have occurred in the United States. For instance, 3 southern states (Texas, Oklahoma, and Louisiana) were among the top 5 states with the highest RSV values for online searches about Ebola, likely attributed to the multiple confirmed Ebola cases in Texas. Despite the high levels of search activity about Ebola, southern states were shown to have considerably lower levels of LHD Ebola tweeting (Figure 1). Similarly, our findings indicated a poor correlation between the number of Ebola tweets sent by LHDs and the RSV values for their respective states. As the frequency of disease outbreaks and the number of unique infectious diseases continue to rise globally,Reference Smith, Goldberg and Rosenthal 9 it is crucial for LHDs and other health authorities to disseminate information about disease outbreaks in a timely manner. We recommend LHDs consider using online search data to tailor their messages on social media and other communications so that they align with health topics of interest among their constituents to increase their engagement,Reference Kreuter and Wray 10 and mitigate potential fears and confusion spread from misinformation.Reference Fung, Tse and Cheung 3
This study was limited by a sample that was restricted to only LHDs. Likewise, we did not investigate state health departments on Twitter that may have tweeted about Ebola. In addition, it is possible that we missed the Twitter pages for some LHDs when we first collected our sample through a web search conducted in July 2012.Reference Harris, Mueller and Snider 5 Finally, Google online search data was only available at the state level; hence, we were unable to examine patterns at the local level. Given that LHDs vary widely within and across states, state-level analyses may have obscured local patterns that differ by the varying structure. Despite these limitations, the findings are important as this is the first study to explore the geospatial distribution of LHDs communicating about a disease during an outbreak, and on whether the distribution of LHD tweets aligned with that of online search activity for the disease.
Conclusions
Events of recent disease outbreaks, including the 2009 H1N1 influenza pandemic, 2014 Ebola outbreak, and the current emergence of the Zika virus may suggest gaps in the capacity for the public health system to prepare for and respond to public health emergencies.Reference Duchin 4 Although delayed communications from public health officials are warranted, they may unintentionally harm public health by enabling infectious diseases to spread among those uninformed or misled by inaccurate information posted online.Reference Fung, Tse and Cheung 3 Therefore, it is critical that LHDs communicate possible disease outbreaks with their constituents in a timely manner before a crisis occurs.Reference Duchin 4 , Reference Wong, Harris and Staub 6 The results of this study also highlight the potential for LHDs to utilize online search data to assess health topics of interest among their constituents and to tailor their communications to effectively engage the community in implementing emergency response strategies.Reference Alicino, Bragazzi and Faccio 2 , Reference Kreuter and Wray 10 This is especially beneficial considering that high online search activity for a particular disease may be indicative of the increased information needs of Internet users, who may be seeking relevant information based on the symptoms of the users themselves or those of their friends.Reference Alicino, Bragazzi and Faccio 2 Further research is needed to better understand how health authorities can use information from online search data for these purposes.
