Comparison of the efficiency and cost of West Nile virus surveillance methods in California

Abstract

Surveillance systems for West Nile virus (WNV) combine several methods to determine the location and timing of viral amplification. The value of each surveillance method must be measured against its efficiency and costs to optimize integrated vector management and suppress WNV transmission to the human population. Here we extend previous comparisons of WNV surveillance methods by equitably comparing the most common methods after standardization on the basis of spatial sampling density and costs, and by estimating optimal levels of sampling effort for mosquito traps and sentinel chicken flocks. In general, testing for evidence of viral RNA in mosquitoes and public-reported dead birds resulted in detection of WNV approximately 2-5 weeks earlier than serological monitoring of sentinel chickens at equal spatial sampling density. For a fixed cost, testing of dead birds reported by the public was found to be the most cost effective of the methods, yielding the highest number of positive results per $1000. Increased spatial density of mosquito trapping was associated with more precise estimates of WNV infection prevalence in mosquitoes. Our findings also suggested that the most common chicken flock size of 10 birds could be reduced to six to seven without substantial reductions in timeliness or sensitivity. We conclude that a surveillance system that uses the testing of dead birds reported by the public complemented by strategically timed mosquito and chicken sampling as agency resources allow would detect viral activity efficiently in terms of effort and costs, so long as susceptible bird species that experience a high mortality rate from infection with WNV, such as corvids, are present in the area.

Keywords: Birds; California; Cost effectiveness; Surveillance methods; West Nile virus.

FIG. 1.

The locations of the three participating agencies: (A) Sacramento–Yolo Mosquito and Vector Control District, (B) Kern Mosquito and Vector Control District, and (C) the Coachella Valley Mosquito and Vector Control District.

FIG. 2.

Seasonal patterns in the mean number of positive results (traps, flocks, or dead birds) per $1000 spent on testing for each surveillance type. Averages are the result of smoothing using locally weighted regression to reduce the noise in the data.

FIG. 3.

Uncertainty of West Nile virus (WNV) prevalence estimates in relation to trapping density. Uncertainty was calculated for each surveillance week as the width of 95% confidence intervals (CIs) for infection prevalence in female Cx. tarsalis and Cx. pipiens complex mosquitoes captured by all trap types combined, or gravid or CO₂-baited traps alone. The lines and gray areas represent the means and 95% CIs using a generalized additive model to depict overall trend in uncertainty. Color images available online at www.liebertpub.com/vbz