Digital Epidemiology and Surveillance  (DES) Session 1

Abstract: Close

Abstract: Emerging pathogens events, from the H1N1 influenza pandemic, to the current Zika invasion of the Americas, represent serious global health threats that necessitate rapid risk assessment and intervention planning. Computational and statistical models provide an invaluable assisting tool in this effort, but their design and calibration face major scientific challenges, not least the accounting for the effect of awareness and human response. In the era of global information, awareness may anticipate the epidemic spread rising in yet unaffected territories and impacting their risk of experience the epidemic. I will discuss this aspect through two examples from the recent Ebola and MERS-CoV epidemics. In the first example, spontaneous reaction to the risk of global dissemination yielded flight cancellations and borders? closure. Through accurate data collection and extensive numerical simulations we showed that the strong modification in the flight network observed delayed by only a few weeks the risk of outbreak propagation in new countries. In the second example, we measured collective and public health awareness through digital proxies (Google Trends, ProMED-mail and Disease Outbreak News of WHO) and we quantified the impact of this component on the management of imported MERS cases in unaffected countries. We found that when attention was high, the time from hospitalization to isolation was reduced from 6 to 2 days, on average, consequently halving the risk of onward transmission following importation. These two examples show how the rise of novel information sources and the large increase in data availability are making possible the accounting for the human factor on epidemic models, increasing the reliability of risk assessment analyses. Close

Abstract: Poor air quality episodes in high populated cities across the world are getting more and more common, these events are not longer sporadic, instead, they are getting recurrent and put millions of people at risk due to the high concentrations of pollutants on the air that they breath. Most of the impact of those episodes go unnoticed because they imply symptoms like headaches, fever or respiratory problems which do not become important to be diagnosed. However, those symptoms affect our daily life, our performance at work and thus, in turn, they impact our economy and/or society. To understand and quantify that effect, we have analyzed a large database of social media messages (136 Million geolocalized tweets) in Spain and, using natural language processing and machine learning techniques, we have identify 0.8 Million tweets in which users talk about suffering symptoms like ILI, common cold, fever, headache, digestive or respiratory problems and potential treatments (perceived health). We have also collected information from official sources on air quality, pollens and weather. With that information we have constructed weekly time series and studied their interdependence and predicting power. We found first ILI cases can be explained and nowcasted by perceived symptomatological data and second that perceived symptomatological data can be nowcasted from atmospheric factors such as pollutants, pollens and climate data. Our results apply both at the regional and city level at different regions in Spain, suggesting that using this kind of digital health data from users in social media could help councils and governments to construct better air quality monitoring systems that not only consider level of contaminants in the air, but also how those levels impact in real time perceived (and possibly real) health conditions of the population. Close

Abstract: Gastroenteritis is one of the most common illnesses worldwide. It is characterised by the symptoms of diarrhoea and vomiting. Although most cases of gastroenteritis in high income countries are self-limiting, there is a significant impact on healthcare services and the economy. Determining the burden of gastroenteritis is challenging. Presentation biases mean that public health datasets of gastroenteritis incidence, and of incidence of gastroenteritis causing pathogens, do not give a complete picture of the community burden. We have been exploring the possibility of establishing a comprehensive near real-time picture of the levels of activity of gastroenteritis in the UK by investigating online data for syndromic surveillance of this illness. This includes webpage view statistics and an online community cohort survey; both of which have been extensively demonstrated as suitable for surveillance of other illnesses. Incidence data from these new online sources is compared to more traditional surveillance data from public health departments. This work contributes towards an improved understanding of gastroenteritis burden, which can influence policy decisions regarding the management of this illness. Close

Abstract: The Flusurvey is an internet-based tool through which real-time surveillance of self-reported influenza like illness (ILI) in the community is undertaken. The Flusurvey collects information on vaccination status as well as the reason(s) given for either getting or not getting vaccinated. (Participants are allowed to provide more than 1 reason). We used these responses to explore the attitudes participants have about the influenza vaccine. 2,901 (34.9%) of Flusurvey participants reported as being vaccinated for the 2015/16 season (5,418 participants (65.1%) were not vaccinated). The majority of those vaccinated were vaccinated at their GP (56.5%), with 21.4% and 18.2% vaccinated at their place of work and pharmacy, respectively. 45.4% of those vaccinated said they were in a risk group and 32.6% said the vaccine was readily available to them and vaccine administration was convenient. Of those not vaccinated; 49.9% felt they did not belong to a risk group, 43.6% were not offered it by their GP, 7% said the vaccine was not free of charge, 6.4% were worried the vaccine was not safe or could cause illness or other adverse events, 7.5% doubted the effectiveness of the vaccine and 2.5% believed the vaccine could cause influenza, Encouragingly, a relatively small percentage of people had negative attitudes about the influenza vaccine (i.e. concerns around safety, side effects or efficacy of vaccine). Our results suggest that increasing vaccine availability and improving convenience of administration (e.g. in the work place) would increase vaccine uptake in the general population. Close