38639 #Cdcgrandrounds and #Vitalsigns: A Cross-Sectional Analysis of Twitter Data

Ashley Jackson, BS1, Lindsay Mullican, BS1, Jingjing Yin, PhD2, Zion Tsz Ho Tse, PhD3, Hai Liang, PhD4, King-Wa Fu, PhD5, Jennifer Ahweyevu, BS6, Jimmy Jenkins, BS1, Nitin Saroha, BS7 and Isaac Chun-Hai Fung, PhD1, 1Department of Epidemiology and Environmental Health Sciences, Georgia Southern University, Statesboro, GA, 2Department of Biostatistics, Georgia Southern University, Statesboro, GA, 3College of Engineering, University of Georgia, Athens, GA, 4School of Journalism and Communication, The Chinese University of Hong Kong, Hong Kong, Hong Kong, 5Journalism and Media Studies Centre, University of Hong Kong, Hong Kong, Hong Kong, 6Department of Epidemiology & Environmental Health Sciences, Georgia Southern University, Statesboro, GA, 7Department of Computer Science, The University of Georgia, Athens, GA

Theoretical Background and research questions/hypothesis:  The CDC hosts monthly panel presentations with webcast titled ‘Grand Rounds’ since September 2009. CDC also publishes a monthly report known as Vital Signs. The CDC uses two respective hashtags #CDCGrandRounds and #VitalSigns respectively to promote their monthly event and report on Twitter. Our research question is to quantify the effect of attaching images or videos to #CDCGrandRounds or #VitalSigns tweets on their retweet frequency.

Methods:  Data was retrieved via Twitter Search Application Programming Interface. All original tweets containing the hashtag #CDCGrandRounds dated from April 21, 2011 to October 25, 2016 were retrieved (n=6,966).  All original tweets containing the hashtag #VitalSigns dated from March 19, 2013 to October 31, 2016 were retrieved (n=15,015). Each corpus was then sub-set into cycles (#CDCGrandRounds: n=58, #VitalSigns: n= 42). A #CDCGrandRounds cycle is defined as all tweets referring to the pre-specified topic for that particular cycle.  A #VitalSigns cycle is defined as the first day of the publication release, which is the first Tuesday of each month, until the day before the next publication is released. Any irrelevant tweets were excluded. We manually coded the 30 tweets with the highest number of retweets for each cycle, as whether it contained a form of media (a still image or a video). Univariable negative binomial regression models were applied to compute the probability ratio of each cycle, with the outcome variable being the retweet frequency and the predictor variable being whether a tweet contains media.

Results:  Of the 58 #CDCGrandRounds cycles, 29 were found to have statistically significant difference between tweets with and without media. Of these 29 cycles, one had a probability ratio (PR) estimate <1; twenty-four had PR >1 but <3; and four had PR>3. Two cycles were outliers: “Preventing Suicide: A Comprehensive Public Health Approach” (September 2015) with PR = 36.353 (95% CI, 4.869 – 343.845, P<0.001) and “Understanding the Causes of Major Birth Defects: Steps to Prevention” (January 2015) with PR=34.713 (95% CI, 7.662 – 261.591, P<0.001). Of the 42 #VitalSigns cycles, 19 were statistically significant.  The PR estimates of six of these 19 cycles were >1 and <3; and for 13 cycles, PR were >3. There were three outliers: “Prescription Painkiller Overdoses” (July 2, 2013) with PR = 33.514 (95% CI, 8.715, 133.357, P<0.01), “Preventing Norovirus Outbreaks” (June 3, 2014) with PR=29.536 (95% CI, 1.330, 326.283, P<0.01), and “Trucker Safety” (PR=10.27, 95% CI, 2.992, 37.010, P<0.01).

Conclusions:  The effect of attaching images or photos increasing retweet frequency varied across cycles for original tweets with #CDCGrandRounds and #VitalSigns. Images or photos may or may not increase retweet frequency.

Implications for research and/or practice:  Future research is needed to determine the optimal choice of images or photos attached to a tweet to maximize the influence of public health messages.