Variation in outbreak size during the transition to endemicity: West Nile virus in New York City

Background/Question/Methods

Understanding of the dynamics of emerging infectious diseases during the transition to endemicity is an important public health concern and a challenging problem for epidemic theory. Key quantities of interest include the size of primary and secondary outbreaks and the level of infection at the endemic equilibrium. The size of annual outbreaks in seasonally forced multi-host disease transmission systems is particularly poorly understood. We studied contributing factors to the six-fold variation in the number of human cases of West Nile virus (WNV) in New York City in the years 2000-2008. We developed a multi-species dynamical model in which both host and vector species varied in competence. We parameterized the model using techniques from statistical meta-analysis and data on laboratory experiments. Intrinsic noise (demographic stochasticity) and observation error (sampling variance) were shown to explain roughly half (44.3%) of the observed variation. To explain the remaining variation, we estimated the monthly force-of-infection from data on the distribution and abundance of mosquitoes, WNV prevalence, vector competence, and mammal biting rate at two spatial scales.

Results/Conclusions

Through dynamical systems modeling and statistical meta-analysis we partitioned WNV transmission among vector species. Unsurprisingly, Culex spp and Aedes albopictus were found to be highly competent. Cx. pipiens was less competent than Cx. salinarius and Cx. restuans, however, due to a lower rate of virus dissemination. Mammal biting rate was highest among species not especially competent for WNV, e.g., Ae. Vexans. Thus, multiple species contributed to transmission, but for different reasons. At two spatial scales, the estimated force-of-infection showed a strong and consistent seasonal periodicity. This result implies that there is limited scope for intrinsic nonlinear mechanisms (e.g. multi-annual cycles) or fluctuations in extrinsic environmental variables (e.g., precipitation and temperature) to explain the variation in human cases. Indeed, correlation among seasonal anomalies in the force of infection (from all sources, whether intrinsic or extrinsic) explained only 15% of the variance in outbreak size among humans. A spatial decomposition of the force-of-infection into five spatial units (boroughs) shows greater variability, particularly with respect to the timing of the epidemic maximum and first record in each year. Thus, we propose that effects of fine scale spatial heterogeneity are key to understanding the epidemiology of WNV in New York City