OBJECTIVE: This project aimed to quantify and compare Massachusetts and Georgia public school districts' 2017-2018 winter-storm-related Twitter unplanned school closure announcements (USCA). METHODS: Public school district Twitter handles and National Center for Education Statistics data were obtained for Georgia and Massachusetts. Tweets were retrieved using Twitter application programming interface. Descriptive statistics and regression analyses were conducted to compare the rates of winter-storm-related USCA. RESULTS: Massachusetts had more winter storms than Georgia during the 2017-2018 winter season, but Massachusetts school districts posted winter-storm-related USCA at a 60% lower rate per affected day (adjusted rate ratio, aRR = 0.40, 95% confidence intervals, CI: 0.30, 0.52) than Georgia school districts after controlling for the student enrollments and Twitter followers count per Twitter account. A 10-fold increase in followers count was correlated with a 118% increase in USCA rate per affected day (aRR = 2.18; 95% CI: 1.74, 2.75). Georgia school districts had a higher average USCA tweet rate per winter-storm-affected day than Massachusetts school districts. A higher number of Twitter followers was associated with a higher number of USCA tweets per winter-storm-affected day. CONCLUSION: Twitter accounts of school districts in Massachusetts had a lower tweet rate for USCA per winter-storm-affected days than those in Georgia.

OBJECTIVE: Researchers at the Centers for Disease Control and Prevention monitor unplanned school closure (USC) reports through online systematic searches (OSS) to assist public health emergency responses. We counted the additional reports identified through social media along with OSS to improve USC monitoring. METHODS: Facebook and Twitter data of public-school districts and private schools in counties affected by California wildfires in October and December of 2017 and January of 2018 were retrieved. We computed descriptive statistics and performed multivariable logistic regression for both OSS and social media data. RESULTS: Among the 362 public-school districts in wildfire-affected counties, USCs were identified for 115 (32%) districts, of which OSS identified 104 (90%), Facebook, 59 (52%), and Twitter, 37 (32%). These data correspond to 4622 public schools, among which USCs were identified for 888 (19.2%) schools, of which OSS identified 722 (81.3%), Facebook, 496 (55.9%), and Twitter, 312 (35.1%). Among 1289 private schools, USCs were identified for 104 schools, of which OSS identified 47 (45.2%), Facebook, 67 (64.4%), and Twitter, 29 (27.9%). USC announcements identified via social media, in addition to those via OSS, were 11 public school districts, 166 public schools, and 57 private schools. CONCLUSION: Social media complements OSS as additional resources for USC monitoring during disasters.

OBJECTIVES: To aid emergency response, Centers for Disease Control and Prevention (CDC) researchers monitor unplanned school closures (USCs) by conducting online systematic searches (OSS) to identify relevant publicly available reports. We examined the added utility of analyzing Twitter data to improve USC monitoring. METHODS: Georgia public school data were obtained from the National Center for Education Statistics. We identified school and district Twitter accounts with 1 or more tweets ever posted ("active"), and their USC-related tweets in the 2015-16 and 2016-17 school years. CDC researchers provided OSS-identified USC reports. Descriptive statistics, univariate, and multivariable logistic regression were computed. RESULTS: A majority (1,864/2,299) of Georgia public schools had, or were in a district with, active Twitter accounts in 2017. Among these schools, 638 were identified with USCs in 2015-16 (Twitter only, 222; OSS only, 2015; both, 201) and 981 in 2016-17 (Twitter only, 178; OSS only, 107; both, 696). The marginal benefit of adding Twitter as a data source was an increase in the number of schools identified with USCs by 53% (222/416) in 2015-16 and 22% (178/803) in 2016-17. CONCLUSIONS: Policy-makers may wish to consider the potential value of incorporating Twitter into existing USC monitoring systems.

BACKGROUND: For pandemic preparedness, researchers used online systematic searches to track unplanned school closures (USCs). We determine if Twitter provides complementary data. METHODS: Twitter handles of Michigan public schools and school districts were identified. All tweets associated with these handles were downloaded. USC-related tweets were identified using 5 keywords. Descriptive statistics and multivariable logistic regression were performed in R. RESULTS: Among 3469 Michigan public schools, 2003 maintained their own active Twitter accounts or belonged to school districts with active Twitter accounts. Of these 2003 schools, in 2015-2016 school year, at least 1 USC announcement was identified for 349 schools via the current method only, 678 schools via Twitter only, and 562 schools via both methods. No USC announcements were identified for 414 schools. Rural schools were less likely than city schools to have active Twitter coverage (adjusted relative risk [adjRR] = 0.3956, 95% confidence interval [CI] 0.3312-0.4671), and to announce USCs on Twitter (adjRR = 0.5692, 95% CI 0.4645-0.6823), but more likely to have USCs identified by the current method (adjRR = 1.4545, 95% CI 1.3545-1.5490). CONCLUSIONS: Each method identified USCs that were missed by the other. Our results suggested that identifying USCs on Twitter is complementary to the current method.

BACKGROUND: Information and emotions towards public health issues could spread widely through online social networks. Although aggregate metrics on the volume of information diffusion are available, we know little about how information spreads on online social networks. Health information could be transmitted from one to many (i.e. broadcasting) or from a chain of individual to individual (i.e. viral spreading). The aim of this study is to examine the spreading pattern of Ebola information on Twitter and identify influential users regarding Ebola messages. METHODS: Our data was purchased from GNIP. We obtained all Ebola-related tweets posted globally from March 23, 2014 to May 31, 2015. We reconstructed Ebola-related retweeting paths based on Twitter content and the follower-followee relationships. Social network analysis was performed to investigate retweeting patterns. In addition to describing the diffusion structures, we classify users in the network into four categories (i.e., influential user, hidden influential user, disseminator, common user) based on following and retweeting patterns. RESULTS: On average, 91% of the retweets were directly retweeted from the initial message. Moreover, 47.5% of the retweeting paths of the original tweets had a depth of 1 (i.e., from the seed user to its immediate followers). These observations suggested that the broadcasting was more pervasive than viral spreading. We found that influential users and hidden influential users triggered more retweets than disseminators and common users. Disseminators and common users relied more on the viral model for spreading information beyond their immediate followers via influential and hidden influential users. CONCLUSIONS: Broadcasting was the dominant mechanism of information diffusion of a major health event on Twitter. It suggests that public health communicators can work beneficially with influential and hidden influential users to get the message across, because influential and hidden influential users can reach more people that are not following the public health Twitter accounts. Although both influential users and hidden influential users can trigger many retweets, recognizing and using the hidden influential users as the source of information could potentially be a cost-effective communication strategy for public health promotion. However, challenges remain due to uncertain credibility of these hidden influential users.

BACKGROUND: The Office of Advanced Molecular Detection (OAMD), Centers for Disease Control and Prevention (CDC), manages a Twitter profile (@CDC_AMD). To our knowledge, no prior study has analyzed a CDC Twitter handle's entire contents and all followers. OBJECTIVE: This study aimed to describe the contents and followers of the Twitter profile @CDC_AMD and to assess if attaching photos or videos to tweets posted by @CDC_AMD would increase retweet frequency. METHODS: Data of @CDC_AMD were retrieved on November 21, 2016. All followers (N=809) were manually categorized. All tweets (N=768) were manually coded for contents and whether photos or videos were attached. Retweet count for each tweet was recorded. Negative binomial regression models were applied to both the original and the retweet corpora. RESULTS: Among the 809 followers, 26.0% (210/809) were individual health professionals, 11.6% (94/809) nongovernmental organizations, 3.3% (27/809) government agencies' accounts, 3.3% (27/809) accounts of media organizations and journalists, and 0.9% (7/809) academic journals, with 54.9% (444/809) categorized as miscellaneous. A total of 46.9% (360/768) of @CDC_AMD's tweets referred to the Office's website and their current research; 17.6% (135/768) referred to their scientists' publications. Moreover, 80% (69/86) of tweets retweeted by @CDC_AMD fell into the miscellaneous category. In addition, 43.4% (333/768) of the tweets contained photos or videos, whereas the remaining 56.6% (435/768) did not. Attaching photos or videos to original @CDC_AMD tweets increases the number of retweets by 37% (probability ratio=1.37, 95% CI 1.13-1.67, P=.002). Content topics did not explain or modify this association. CONCLUSIONS: This study confirms CDC health communicators' experience that original tweets created by @CDC_AMD Twitter profile sharing images or videos (or their links) received more retweets. The current policy of attaching images to tweets should be encouraged.

BACKGROUND: Early insights into the timing of the start, peak, and intensity of the influenza season could be useful in planning influenza prevention and control activities. To encourage development and innovation in influenza forecasting, the Centers for Disease Control and Prevention (CDC) organized a challenge to predict the 2013-14 Unites States influenza season. METHODS: Challenge contestants were asked to forecast the start, peak, and intensity of the 2013-2014 influenza season at the national level and at any or all Health and Human Services (HHS) region level(s). The challenge ran from December 1, 2013-March 27, 2014; contestants were required to submit 9 biweekly forecasts at the national level to be eligible. The selection of the winner was based on expert evaluation of the methodology used to make the prediction and the accuracy of the prediction as judged against the U.S. Outpatient Influenza-like Illness Surveillance Network (ILINet). RESULTS: Nine teams submitted 13 forecasts for all required milestones. The first forecast was due on December 2, 2013; 3/13 forecasts received correctly predicted the start of the influenza season within one week, 1/13 predicted the peak within 1 week, 3/13 predicted the peak ILINet percentage within 1 %, and 4/13 predicted the season duration within 1 week. For the prediction due on December 19, 2013, the number of forecasts that correctly forecasted the peak week increased to 2/13, the peak percentage to 6/13, and the duration of the season to 6/13. As the season progressed, the forecasts became more stable and were closer to the season milestones. CONCLUSION: Forecasting has become technically feasible, but further efforts are needed to improve forecast accuracy so that policy makers can reliably use these predictions. CDC and challenge contestants plan to build upon the methods developed during this contest to improve the accuracy of influenza forecasts.

To aid decision-making during CDC's response to the 2014-2016 Ebola virus disease (Ebola) epidemic in West Africa, CDC activated a Modeling Task Force to generate estimates on various topics related to the response in West Africa and the risk for importation of cases into the United States. Analysis of eight Ebola response modeling projects conducted during August 2014-July 2015 provided insight into the types of questions addressed by modeling, the impact of the estimates generated, and the difficulties encountered during the modeling. This time frame was selected to cover the three phases of the West African epidemic curve. Questions posed to the Modeling Task Force changed as the epidemic progressed. Initially, the task force was asked to estimate the number of cases that might occur if no interventions were implemented compared with cases that might occur if interventions were implemented; however, at the peak of the epidemic, the focus shifted to estimating resource needs for Ebola treatment units. Then, as the epidemic decelerated, requests for modeling changed to generating estimates of the potential number of sexually transmitted Ebola cases. Modeling to provide information for decision-making during the CDC Ebola response involved limited data, a short turnaround time, and difficulty communicating the modeling process, including assumptions and interpretation of results. Despite these challenges, modeling yielded estimates and projections that public health officials used to make key decisions regarding response strategy and resources required. The impact of modeling during the Ebola response demonstrates the usefulness of modeling in future responses, particularly in the early stages and when data are scarce. Future modeling can be enhanced by planning ahead for data needs and data sharing, and by open communication among modelers, scientists, and others to ensure that modeling and its limitations are more clearly understood. The activities summarized in this report would not have been possible without collaboration with many U.S. and international partners (http://www.cdc.gov/vhf/ebola/outbreaks/2014-west-africa/partners.html).

BACKGROUND: School closures may delay the epidemic peak of the next influenza pandemic, but whether school closure can delay the peak until pandemic vaccine is ready to be deployed is uncertain. METHODS: To study the effect of school closures on the timing of epidemic peaks, we built a deterministic susceptible-infected-recovered model of influenza transmission. We stratified the U.S. population into 4 age groups (0-4, 5-19, 20-64, and ≥65 years), and used contact matrices to model the average number of potentially disease transmitting, nonphysical contacts. RESULTS: For every week of school closure at day 5 of introduction and a 30% clinical attack rate scenario, epidemic peak would be delayed by approximately 5 days. For a 15% clinical attack rate scenario, 1 week closure would delay the peak by 9 days. Closing schools for less than 84 days (12 weeks) would not, however, reduce the estimated total number of cases. CONCLUSIONS: Unless vaccine is available early, school closure alone may not be able to delay the peak until vaccine is ready to be deployed. Conversely, if vaccination begins quickly, school closure may be helpful in providing the time to vaccinate school-aged children before the pandemic peaks.

BACKGROUND: To support policy making, we developed an initial model to assess the cost-effectiveness of potential strategies to increase influenza vaccination rates among children in China. METHODS: We studied on children aged 6 months to 14 years in four provinces (Shandong, Henan, Hunan, and Sichuan), with a health care system perspective. We used data from 2005/6 to 2010/11, excluding 2009/10. Costs are reported in 2010 U.S. dollars. RESULTS: In comparison with no vaccination, the mean (range) of Medically Attended Cases averted by the current self-payment policy for the two age groups (6 to 59 months and 60 months to 14 years) was 1,465 (23 approximately 11,132) and 792 (36 approximately 4,247), and the cost effectiveness ratios were $ 0 (-11-51) and $ 37 (6-125) per case adverted, respectively. In comparison with the current policy, the incremental cost effectiveness ratio (ICER) of alternative strategies, OPTION One-reminder and OPTION Two-comprehensive package, decreased as vaccination rate increased. The ICER for children aged 6 to 59 months was lower than that for children aged 60 months to 14 years. CONCLUSIONS: The model is a useful tool in identifying elements for evaluating vaccination strategies. However, more data are needed to produce more accurate cost-effectiveness estimates of potential vaccination policies.

During the 2013 outbreak of human infections of avian influenza A(H7N9), the Centers for Disease Control and Prevention (CDC) used official data released by the World Health Organization (WHO) and the Chinese government to keep United States public health officials informed of updates of the outbreak.1 The Chinese central government released official avian influenza A(H7N9) data via its web sites (e.g. National Health and Family Planning Commission2), their official news agency (Xinhua News Agency) and their official newspapers (e.g. People’s Daily, Beijing). In addition, official avian influenza A(H7N9) information was released by Chinese provincial and municipal governments such as Shanghai Municipal Bureau of Health,3 Jiangsu Department of Health4 and Zhejiang Department of Health.5 Prior studies have discussed the role of social media in the early detection of disease outbreaks6–9 and the facilitation of community-level discussion.10 In this perspective, we focus on the use of social media by public health agencies to disseminate and obtain official outbreak information during a public health emergency response. | | Weibo (literally, microblog) is a category of Chinese microblogging sites that are similar to Twitter. Both Twitter and weibo are social media that allow users to post a 140-character long message online. Weibo has become popular in China since August 2009 when Twitter became unavailable to users in mainland China. As of December 2012, 309 million people were reported to be weibo users in China as compared to the global 500 million registered Twitter users as of July 2012. There are several different providers of weibo, including Sina Weibo, Tencent (QQ) Weibo, Sohu Weibo, Baidu Weibo, ifeng Weibo, NetEase Weibo and others. Most weibo users live in China; a random sample of users of Sina Weibo found that 1.6% of users were from countries other than China.11

In 2010, toxigenic Vibrio cholerae was newly introduced to Haiti. Because resources are limited, decision-makers need to understand the effect of different preventive interventions. We built a static model to estimate the potential number of cholera cases averted through improvements in coverage in water, sanitation and hygiene (WASH) (i.e., latrines, point-of-use chlorination, and piped water), oral cholera vaccine (OCV), or a combination of both. We allowed indirect effects and non-linear relationships between effect and population coverage. Because there are limited incidence data for endemic cholera in Haiti, we estimated the incidence of cholera over 20 years in Haiti by using data from Malawi. Over the next two decades, scalable WASH interventions could avert 57,949-78,567 cholera cases, OCV could avert 38,569-77,636 cases, and interventions that combined WASH and OCV could avert 71,586-88,974 cases. Rate of implementation is the most influential variable, and combined approaches maximized the effect.