09/11/2018 - Infodengue: A nowcasting system for the surveillance of arboviruses in Brazil - Palestrante: Leonardo Soares Bastos (Fundação Oswaldo Cruz)
Seminário Conjunto UFSCar/ICMC
Data e horário: 09/11/2018 às 14 horas
Local: ICMC-USP Sala 4-001
Palestrante: Leonardo Soares Bastos (Fundação Oswaldo Cruz, Rio de Janeiro).
Título: Infodengue: A nowcasting system for the surveillance of arboviruses in Brazil
Resumo: In cities dengue is endemic, various protocols exist for monitoring the disease activity as well as its risk factors, such as vector density, climate favorable for transmission and case counts. However, the complexity and cost of the data collection, together with the delays caused by the flow of data from data collections systems, through central health authorities and back to the health agents in charge of controlling the disease, makes timely responses to dengue outbreaks very hard to achieve. In this presentation we describe a system put in place through a joint initiative of the Brazilian National Health Ministry, city level health authorities, and multiple research institutions, bringing together epidemiologists, statisticians, entomologists, computer scientists among other specialties, to tackle the problem of bringing up-to-date epidemiological information to health agents, local decision-makers and the population as a whole. This system is operating in 788 cities in Brazil since 2015 (http://info.dengue.mat.br). One of the main design goals of the system was to include all data streams deemed relevant which were available on a regular time-frame. These are data streams selected based not only on their epidemiological relevance, but also on their continuous availability: mention about dengue on Twitter, filtered according to content and aggregated as a city level time series. Climate data (temperature, humidity, and atmospheric pressure measure); epidemiological data (these are the official case data, clinically confirmed cases reported by medical professionals through official channels). A pipeline was developed to clean, filter, and integrate the datasets. Statistical techniques were developed to: (i) correct for reporting delay using a efficient Bayesian framework; (ii) estimate the effective reproductive number of dengue taking into account the variation in temperature; and (iii) detect sustained transmission. The pipeline delivers a classification of alert every week: green (poor conditions for transmission), yellow (favorable transmission conditions), orange (sustained transmission), red (high incidence). At last, we discuss how the developed pipeline can be extended to other climate-sensitive diseases, exemplifying with Zika and chikungunya.