Pepin KM, Davis AJ, Streicker DG, Fischer JW, VerCauteren KC, Gilbert AT. Predicting spatial spread of rabies in skunk populations using surveillance data reported by the public.
PLoS Negl Trop Dis 2017;
11:e0005822. [PMID:
28759576 PMCID:
PMC5552346 DOI:
10.1371/journal.pntd.0005822]
[Citation(s) in RCA: 10] [Impact Index Per Article: 1.4] [Reference Citation Analysis] [Abstract] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/02/2017] [Revised: 08/10/2017] [Accepted: 07/20/2017] [Indexed: 11/18/2022] Open
Abstract
Background
Prevention and control of wildlife disease invasions relies on the ability to predict spatio-temporal dynamics and understand the role of factors driving spread rates, such as seasonality and transmission distance. Passive disease surveillance (i.e., case reports by public) is a common method of monitoring emergence of wildlife diseases, but can be challenging to interpret due to spatial biases and limitations in data quantity and quality.
Methodology/Principal findings
We obtained passive rabies surveillance data from dead striped skunks (Mephitis mephitis) in an epizootic in northern Colorado, USA. We developed a dynamic patch-occupancy model which predicts spatio-temporal spreading while accounting for heterogeneous sampling. We estimated the distance travelled per transmission event, direction of invasion, rate of spatial spread, and effects of infection density and season. We also estimated mean transmission distance and rates of spatial spread using a phylogeographic approach on a subsample of viral sequences from the same epizootic. Both the occupancy and phylogeographic approaches predicted similar rates of spatio-temporal spread. Estimated mean transmission distances were 2.3 km (95% Highest Posterior Density (HPD95): 0.02, 11.9; phylogeographic) and 3.9 km (95% credible intervals (CI95): 1.4, 11.3; occupancy). Estimated rates of spatial spread in km/year were: 29.8 (HPD95: 20.8, 39.8; phylogeographic, branch velocity, homogenous model), 22.6 (HPD95: 15.3, 29.7; phylogeographic, diffusion rate, homogenous model) and 21.1 (CI95: 16.7, 25.5; occupancy). Initial colonization probability was twice as high in spring relative to fall.
Conclusions/Significance
Skunk-to-skunk transmission was primarily local (< 4 km) suggesting that if interventions were needed, they could be applied at the wave front. Slower viral invasions of skunk rabies in western USA compared to a similar epizootic in raccoons in the eastern USA implies host species or landscape factors underlie the dynamics of rabies invasions. Our framework provides a straightforward method for estimating rates of spatial spread of wildlife diseases.
Rabies is a deadly zoonotic infection with a global distribution. In 2012, an epizootic of skunk rabies established in northern Colorado, USA and spread rapidly through three counties. The epizootic was documented through reports of dead skunks by the public. We examined the reports to determine how rapidly rabies was moving and which factors could explain the patterns of spread. We compared these estimates of spatial movement of rabies to those obtained from analyzing rabies genetic sequences that we obtained from some of the dead skunks reported by the public. By both methods, we found the virus was moving south at a little over 20 km/year and that most transmission between skunks occurred at short distances (< 4 km). Rabies was most likely to spread to new areas during the first half of the year, when skunk populations were producing new offspring. Our genetic model suggested that roads and rivers in the study landscape did not affect movement speed of rabies. We developed a framework that used the spatial data in the public reports to predict where and when skunk rabies would occur next. This framework could be used on public health surveillance data for other diseases or countries.
Collapse