Challenges in modeling emergent vector-borne diseases

Disease models are challenged with prediction for the purposes of prevention or mitigation of future outbreaks. External factors, e.g., lack of funding, often lead to interruptions of data collection and research on vector-borne pathogens that only sporadically affect the human population. This confounds our ability to anticipate outbreaks of disease. Looking at the 2012 West Nile epidemic, we asked 1) Can mosquito surveillance for WNV provide a reliable indication of human risk? And 2) Can models assess and correct for sampling bias in active versus inactive seasons? Using detailed datasets from New Orleans, St. Tammany, Livingston, and West Baton Rouge Parishes, we have developed mathematical models and applied statistical techniques to answer these critical questions. We present here a framework for the implementation of a predictive risk model that addresses the multiannual effects of dynamic avian reservoir demography and weather upon arbovirus activity indicators; we further provide uncertainty and sensitivity analyses to quantify bias in this model; and provide insight into the importance of and minimum requirements for baseline surveillance of emergent but quiescent diseases. A discussion of the generalizability of these results to other emergent vector-borne diseases such as dengue and chikungunya will follow.