Using Twitter to Track Unplanned School Closures: Georgia Public Schools, 2015-17

To aid emergency response and pandemic preparedness, CDC researchers monitor online news sources for reports of unplanned school closures (USCs). We examined the potential of expanding the available data by analyzing Twitter posts on unplanned closures of public schools in Georgia, USA. Georgia public school data were obtained from the National Center for Education Statistics, which included school locale (“urban-centric” classification system). We identified school Twitter accounts and USC-related tweets. Reports of USCs from online news sources were provided by the CDC. Descriptive statistics, univariate, and multivariable logistic regression were computed. Of 2,299 public schools in Georgia, 1,864 schools had active Twitter accounts. Approximately 45% of these schools were in suburban Georgia. In the 2015-16 and 2016-17 sessions, respectively, 52.25% and 20.37% of schools that used Twitter to announce USCs were not captured by online news sources. Conversely, 51.44% and 13.33% of schools with USCs captured by online news sources were not captured by Twitter in 2015-16 and 2016-17 respectively. Compared to city schools, suburban schools were 14% more likely (adjusted relative risk, aRR=1.14; 95% CI=1.08, 1.18) and town schools 50% less likely (aRR=0.50; 95% CI=0.39, 0.62) to have active Twitter accounts. Town schools were 2.7 times (aRR=2.70; 95% CI=2.20, 3.17) as likely as city schools to announce USCs on Twitter in 2015-16; and in 2016-17, city, suburban, town and rural schools were equally likely (p>0.05) to announce USCs on Twitter. Compared with city schools, USCs for town schools are 83% and 62% more likely to be identified by online news sources in 2015-16 (aRR=1.83; 95% CI=1.43, 2.26) and 2016-17 (aRR=1.62; 95% CI=1.43, 1.79) respectively. Our study shows that simultaneously searching for USC announcements on Twitter and online news sources will likely provide a more comprehensive dataset. For global health practitioners, this pilot study demonstrated the potential utility of social media data in advancing surveillance for outbreak management and emergency responses.