Enny Das*, Radboud University, Centre for Language Studies, Nijmegen, Netherlands Emma Broekhuizen, Radboud University, Centre for Language Studies, Nijmegen, Netherlands Florian Kunneman, Radboud University, Centre for Language Studies, Nijmegen, Netherlands Ali Hurriyetoglu, Radboud University, Centre for Language Studies, Nijmegen, Netherlands Track: Research Presentation Topic: Blogs, Microblogs, Twitter Presentation Type: Oral presentation Submission Type: Single Presentation Building: Sheraton Maui Resort Room: C - Napili Date: 2014-11-14 11:50 AM – 12:35 PM Last modified: 2014-09-04

If you are the presenter of this abstract (or if you cite this abstract in a talk or on a poster), please show the QR code in your slide or poster (QR code contains this URL).

Background & Objective
Public responses to health epidemics such as H1N1 and SARS follow unpredictable patterns. Analyzing social media data can increase insight into public opinion and health behavior. Research thus far analyzed tweet content and sentiment about a single disease (Chew & Eysenbach, 2010; Signorini, Segre & Polgreen, 2011; Salathé & Khandelwal, 2011). The present research extends these findings by looking at generalizable patterns in social media content across epidemics (H1N1, SARS, and mumps) as a function of the key factors that determine epidemic threat level: severity (Witte, 1992) and temporal and spatial distance (cf. Henderson, Fujita, Trope & Liberman, 2006; vulnerability in Witte, 1992). Based on previous theorizing that perceptions of personal threat are defined more by perceptions of vulnerability than by perceptions of severity, we hypothesized that expressions of threat and humor would vary mainly as a function of temporal and spatial distance of a health epidemic (e.g., Das, de Wit, & Stroebe, 2003).

Method
8.986 relevant Dutch tweets were gathered via Twiqs.nl. We used the markers ‘(#)SARS’, ‘(#)H1N1’ and ‘(#)(de)bof’ (mumps). These epidemics differ in severity (H1N1 and SARS > mumps), temporal distance (mumps > H1N1, SARS at time of measurement) and physical distance (H1N1, mumps > SARS). For mumps we also compared tweets sent within > outside a radius of 30 km of affected locations based on ID-location. Two independent coders manually coded tweets on eight categories, e.g., threat/fear, humor, secondary consequences and protection measures (H1N1: κ=.75, SARS: κ=.75, mumps: к =.73). In order to assess agenda setting patterns, we also coded articles from five main Dutch newspapers about H1N1 (N=100; κ=.75) and SARS (N=21; κ=.94).

Results
Tweets about H1N1 (severe, small physical distance) contained more serious categories (e.g. threat (36.41% versus 26.00%, χ2(1)= 38.80, p<.001), protection measures (11.04% versus 3.09%, χ2(1)= 78.08, p<.001) than tweets about SARS (severe, large physical distance), which mostly contained jokes (40.66% versus 18.78% in H1N1, χ2(1)= 116.32, p<.001) or use of the disease as a swear word (24.19% versus 0.82% in H1N1, χ2(1)= 177.24, p<.001). Tweets with a small physical distance contained marginally more threat (25.00%) than tweets sent from further away (14.18%, χ2(1)=3.10, p=.078). No significant differences in humor for small (24.04%) and larger physical distance were observed (30.50%, χ2(1)=0.89, p=.345), although the pattern was in the predicted direction. No significant differences in tweet content were found as a function of disease severity. Tweet peak patterns across time closely followed newspaper patterns, suggesting a classic agenda setting effect.

Conclusions
Expressions of threat and humor in social media appear especially informative to monitor public perceptions of personal vulnerability to a health epidemic. Threat tweets were most prominent, and humorous tweets slightly less prominent when an epidemic was psychologically or physically close, even when it was not very severe. Findings inform the monitoring function of social media.