Characteristics of Zika Behavior Discourse on Twitter

Zika is an important emerging illness that has been linked to neurological syndromes in adult patients, and birth defects in patients infected in-utero. Here, we use Twitter to explore the discourse of individuals tweeting about Zika. We labeled a sample of 500 English tweets to identify common themes, used keywords to track two themes, reproduction and travel, and identify spatial patterns in tweets based on the language of the tweet. We also observed tweets made in the first person that might be indicative of reflections of personal behavior on Twitter. Future work will delve further into first person tweets and continue to examine spatial and temporal trends in the themes temporally observed. Introduction Mosquito borne arboviruses (most notably dengue, chikungunya, and, now, Zika) cause massive outbreaks worldwide. In 2015 Zika emerged in South America, and caused a large outbreak throughout many countries in South, Central and North America.1–3 Although the majority of Zika cases are asymptomatic4 or cause mild disease, evidence emerged that there was an association between infection and neurological syndromes (e.g., Guillian-Barré5 or birth defects in fetuses infected in-utero.6) Specifically, women who were infected with Zika during pregnancy had fetuses with much higher rates of microcephaly, a condition where the head circumference of the fetus is very small (below the second percentile for gestation7), and has accompanying brain defects.6 The intersection between this outbreak and the impacts on human behavior are of substantial public health relevance. Social media data provide one way to gain insight into human behavior with respect to well publicized public health events. Social media data now has a rich history of use understanding health-related human behavior. Previous studies have used Twitter as a general way to understand patterns in human behavior with respect to possible risk factors for disease. Paul and Dredze used Twitter data to identify types of ailments discussed on Twitter.8 Others identified tweets related to human behaviors designed to control infection. For example, Signorini et al. found that an important minority of persons talking about H1N1 on Twitter, also talked about possible control measures (hand hygiene and mask wearing).9 As Zika emerged, researchers again used Twitter to find relevant patterns. McGough et al. used a combination of official incidence reports and Internet data streams (including Twitter) to build Zika forecasts for several Central and South America countries.10 Stefanidis et al. used data from the first 3 months of the outbreak to look at spatial clusters in discussions of Zika on Twitter.11 They also found tweets referencing pregnancy and abortion.11 This study builds upon previous work by looking at trends in discourse. We present preliminary evidence that examines temporal, spatial and keyword trends in a larger Zika Twitter corpus, and analyzes the demographics of Twitter users that talked about Zika. These analyses provide a discussion of Zika tweet content at a finer resolution, and broader spatial context than previous work. Methods Data collection Tweets were collected from Gnip, based on the keywords “zika”, “zica” (a common Portuguese spelling of the virus1, or ‘zikv’ (a common abbreviation of ‘Zika Virus’). Tweets were collected from March 1, 2015 until October 31, 2016. This resulted in just under 15.5 million tweets, 7 million of which are in English. For this initial work, we only coded English tweets. We identified the likely gender of person tweeting using Demographer12, and identified the likely location of the tweets using Carmen.13 1https://en.wiktionary.org/wiki/zica Data labeling This initial work has focused on the relationship between Zika and behavioral decisions (e.g., those relating to reproduction, travel, mosquito interventions etc.), as well as to important global events (e.g., 2016 Olympics). We identified a list of keywords related to these behaviors (see Table 1). We then filtered the dataset for tweets that included the keywords. Where applicable we included different plural forms (-ies rather than -s endings). For words that might match longer words or strings (e.g., ‘birth’ would match ‘birthday’), we included \b, a Python regular expression marker for a word boundary. Table 1: Behavior keywords.