Background: Twitter is used by officials to distribute public health messages and by the public to post information about ongoing afflictions. Because tweets originate from geographically and socially diverse sources, scholars have used this social media data to analyze the spread of diseases like flu [Alessio Signorini 2011], asthma [Philip Harber 2019] and mental health disorders [Chandler McClellan, 2017]. To our knowledge, no Twitter analysis has been performed for Valley fever. Valley fever is a fungal infection caused by the Coccidioides organism, mostly found in Arizona and California. Objective: We analyzed tweets concerning Valley fever to evaluate content, location, and timing. Methods: We collected tweets using the Twitter search application programming interface using the terms “Valley fever,” “valleyfever,” “cocci” or “‘Valleyfever” from August 6 to 16, 2019, and again from October 20 to 29, 2019. In total, 2,117 Tweets were retrieved. Tweets not focused on Valley fever were filtered out, including a tweet about “Rift valley fever” and tweets where “valley” and “fever” were separate and not one phrase. We excluded tweets not written in English. In total, 1,533 tweets remained; we grouped them into 3 categories: original tweets, hereafter labeled “normal” (N = 497), retweets (N = 811), and replies (N = 225). We converted all terms to lowercase, removed white space and punctuation, and tokenized the tweets. Informal messaging conventions (eg, hashtag, @user, RT, links) and stop words were removed, and terms were lemmatized. Finally, we analyzed the frequency of tweets by season, state, and co-occurring terms. Results: Tweet frequency was 228.5 per week in summer and 113.4 per week in the fall. Users tweeted from 40 different states; the most common were California (N = 401; 10.1 per 100,00 population) and Arizona (N = 216, 30.1 per 100,000 population), New York (N = 49), Florida (N = 21), and Washington, DC (N = 14). Term frequency analysis showed that for normal tweets, the 5 most frequent terms were “awareness,” “Arizona,” “disease,” “California,” and “people.” For retweets, the most common terms were “Gunner” (a dog name), “vet,” “prayer,” “cough,” and “family.” For replies, they were “dog,” “lung,” “vet,” “day,” and “result.” Several symptoms were mentioned: “cough” (normal: 8, retweets: 104, and replies: 7), “sick” (normal: 21, retweets: 42, replies: 7), “rash” (normal: 2, retweets: 6, replies: 1), and “headache” (normal: 1, retweets: 3, replies: 0). Conclusions: Valley fever tweets are potentially sufficient to track disease intensity, especially in Arizona and California. Data collection over longer intervals is needed to understand the utility of Twitter in this context.Disclosures: NoneFunding: None
Serverless Geospatial Data Processing Workflow System Design
