| Dengue virus (DENV) is currently the most rapidly spreading vector-borne disease, with an estimated 50-100 million infections per year and 2.5 billion people---40% of the world's population---at risk of infection. Since 1990, the city of Ribeirao Preto (pop. 563,000), Brazil, has experienced DENV epidemics of increasing severity, including the largest epidemic to date in 2010 (≈13,000 cases as of April 22). However, there are no vaccines or treatments for DENV, and the only method for reducing morbidity and mortality is through control of the principal mosquito vector, Aedes aegypti. To inform surveillance and control efforts in Ribeirao Preto, a geographic information system (GIS) was created along with an address locator for geocoding dengue cases. The primary objective of this study was to investigate the utility of modifying a West Nile virus early warning system, used successfully in California to predict and prevent human cases, that models viral amplification using a localized Knox test and Monte Carlo simulation approach based on parameters of vector and host biology. Results from this study, which represents the first spatially explicit model that uses human cases to predict future dengue risk, indicate that the modified DYCAST system provided early and accurate identification of high-risk areas in Ribeirao Preto, including detection of what appears to be the cryptic interepidemic focus of transmission that later developed into the severe 2006 epidemic. During the study period, DYCAST predicted up to 90.3% (4,234/4,690) of cases, at a maximum mean of 66.3 days prior to onset of illness. Maximum sensitivity and specificity was 83.8% and 78.8%, respectively, and relative risk of DENV infection was >10x higher in cells identified as high risk. Additionally, model efficacy was retained and even enhanced by including unconfirmed dengue cases in the analysis, which has important implications for increasing the utility and applicability of the model. These findings suggest that the DYCAST system could be utilized prospectively and in real-time to identify areas at high risk of DENV transmission, in order to target mosquito control, surveillance, and public education campaigns in a timely, efficient, and cost-effective manner. Furthermore, and in a departure from previous studies, this risk model was implemented using free, open-source, and cross-platform software that could provide an inexpensive and scalable GIS solution for the surveillance and control of DENV---and potentially other infectious diseases---by Ribeirao Preto and other public health agencies in the future. Protocols for generating the spatial datasets and installing the various software components are also provided. |
