Introduction
Rabies is a highly fatal zoonotic disease with a near global distribution, spanning latitudes and animal host species, including wild carnivores. Despite the existence of post-exposure prophylaxis for humans [1] and vaccines for domestic dogs and wildlife, rabies still causes approximately 60,000 human deaths per year worldwide [2]. An estimated 50,000 human exposures annually in the United States (US) are linked to wildlife reservoir species such as raccoons (Procyon lotor) [3–5].
Historically, the variant of Lyssavirus rabies (RABV) adapted to raccoons in the United States was isolated to the southeastern portion of the country (i.e., Florida and Georgia) [6]. However, translocation of raccoons to West Virginia during the 1970s unknowingly introduced infected animals to a susceptible raccoon population, leading to epizootic spread of the disease across the eastern United States [7]. Intense management efforts, including the use of oral rabies vaccine (ORV) for raccoons, have been successful in containing the westward and northward spread of raccoon rabies virus [8,9].
Though raccoon RABV now spans the entire—more than 20°—latitudinal range of the eastern United States, including being found across a three order-of-magnitude span of county-level human population densities, the distribution of confirmed cases is far from uniform [10,11]. Some ecological contexts across a landscape are more prone to hotspots of disease emergence/spillover [12], while others facilitate sustained transmission within a wildlife host reservoir [13]. Disease dynamics are also heterogeneous in time; some pathogens are maintained at a constant prevalence (enzootic or endemic dynamics), while others produce sporadic outbreaks (epizootics or epidemic dynamics, where outbreaks of infection are followed by inter-epidemic troughs in case numbers). Epidemic dynamics can be governed by seasonal changes to host population demography or contact rates [14,15] that may differ across latitudinal climes. Raccoon rabies dynamics can be characterized in many ways. One method that reduces biases inherent to the passive surveillance utilized in the case of rabies in the United States is a quantification of “presence,” i.e., whether or not there is at least one infected animal within a prescribed area. By incorporating a better understanding of the role of environmental factors such as seasonality and urbanness in raccoon rabies presence at a landscape scale, limited resources can be better prioritized to support enhanced surveillance efforts and refine management strategies across different ecoclimes.
Latitude and seasonality are linked. In lower latitudes, rainfall can vary seasonally (e.g., rainy vs. dry seasons), whereas in higher latitudes, there can be extreme variation in temperature (e.g., summer vs. winter). Seasonal effects can drive animal demography (e.g., births, deaths) and behavior (e.g., mating, birthing, denning, feeding, migration), which can in turn impact disease dynamics via changes in aggregations, contacts, and interactions with infected individuals [14,16]. Accordingly, the number of raccoon rabies cases shows variation in seasonal trends with peaks in rabies occurrence ranging from: March (Florida) [6]; “spring” (March through June) with more sporadic peaks in early fall (October and November: Mid-Atlantic states) [17,18]; late winter (February and March: Virginia) [19]; spring and fall (11 states in the Eastern U.S.) [20]; or just in the fall-winter (Pennsylvania, Ohio, and West Virginia) [21]. Theoretical models show that a combination of wave-like spread of infection and pulsed births in northern climes could create asynchronous dynamics [22]. Additionally, in areas where individuals congregate for warmth in the winter months, pathogen transmission might increase with the change in social contacts and increase in contact duration [14,16,23]. Alternatively, in southern climes where raccoons congregate less and where they breed throughout the year, there could be more consistent transmission throughout the year.
Urbanization can result in the proliferation of urban-adapted species such as raccoons [24], potentially increasing risk of zoonotic infection [25]. For example, in urban areas there are extremely high densities of raccoons [26–31], which can lead to high mixing and contact rates, both between raccoons and with humans [32,33]. This population and contact structure has consequences for transmission: areas with a higher human population density tend to have larger and longer epizootics [34]. Yet, these high-density urban areas may also be fragmented by highways and rivers, which could potentially lead to reduced contact and transmission across the landscape, in contrast to rural areas, where raccoon densities are lower, but potentially less fragmented [35].
Despite this work on the role of latitude and urbanness, we lack a comprehensive framework for understanding how these factors interact, limiting our ability to anticipate the distribution of raccoon rabies across the landscape where it is enzootic. Here we leverage extensive long-term datasets on raccoon rabies surveillance collected as part of enhanced rabies surveillance by the United States Department of Agriculture (USDA) and public health surveillance data collected by states and reported annually to the Centers for Disease Control and Prevention (CDC) to better understand the role of latitude, seasonality, and urbanness in raccoon rabies transmission ecology. The combined data consist of mostly passive surveillance across the entire range of raccoon RABV in the eastern US during 2006–2018. Our study objectives are to quantify the relationship between latitude, seasonality, and “urbanness” (as measured by human population density) on raccoon rabies presence, quantified as the probability of at least one detected case of rabies in raccoons of raccoon rabies in a given county in a given month. We predict that this probability may be higher in northern latitudes (as seasonality can increase viral spread through co-denning) and in urban areas (due to congregations around feeding locations and high densities of raccoons). We predict that the number of rabies cases in northern ecoregions will cluster more in time (i.e., seasonal winter peaks) than in southern ecoregions, and that seasonal peaks might be more pronounced in the high-density urban areas. Identification of spatial and temporal trends in where cases are most likely to occur can help with a better understanding of the factors driving endemic zoonotic disease dynamics, and can help focus education, surveillance, and management efforts on higher-risk time periods and areas.
Discussion
Here we use surveillance data to evaluate whether latitude and urbanness impact the probability of a raccoon rabies case report (“presence”). We find that county-level raccoon rabies dynamics are strongly driven by temporal (the number of rabies cases during the previous three months in that county) and spatial (caseloads in surrounding counties in the current and previous two months) context, as well as testing intensity (total number of tests conducted in a given month), supporting the development of criteria for sample prioritization such as the epizootiological importance [38,42,60]. Additionally, however, we also find evidence for the role of latitude and urbanness (as measured by human population density), especially through interactions between latitude and season (“month”) and between latitude and population density. These results suggest that there is an additional benefit to enhancing raccoon rabies surveillance, management, and education efforts in those particular counties and/or months for which additional surveillance is thought to be most informative. Areas of predicted high risk would additionally be candidates for follow-up epidemiological investigations.
In the absence of a county-level density map of raccoons, we used the proportion of a county’s land cover consisting of habitat that has previously been associated with raccoon preference [61–63] and, consequently, RABV prevalence [21] as a county-level proxy of locations we are likely to see infected raccoons. In agreement with prior literature [21], we found that in counties with less than 16% land cover of deciduous or mixed forest, there were fewer raccoon rabies cases. Thus, while we did not see a trend across the whole range of favorable habitat coverage, wildlife managers could view counties with ‘unfavorable’ habitat (meaning less than 16% of land-cover being favorable to raccoons) as being most amenable to limiting rabies spread in the absence of ORV control. Likewise, in the face of land-use change, changes to raccoon favorable habitat coverage are unlikely to have an effect unless it is already quite limited. The habitat relationships explored here, however, should be interpreted with some caution as our binary classification of the suitability of land cover is an oversimplification both of raccoon space use [30,63,64] and of the importance of raccoon density on RABV dynamics [30].
We predicted that rabies cases in northern ecoregions would be more clustered in time than in southern ecoregions; indeed, in the north we found seasonal trends with more raccoon rabies cases detected in the winter than in the summer. However, in the north, more raccoons were submitted for rabies surveillance in the summer compared to the winter; therefore, inferring case trends from looking only at the number of positive samples through time could be misleading. In the north, more raccoons might be submitted for testing in the summer than in the winter because humans are more likely to be outside in the summer (a consequence of passive surveillance that relies on public reporting) [65] when raccoons are more likely to be active with their weaned young [66]. Our results suggest that rabies might be circulating more in the winter in the north (consistent with previous studies) [21], albeit with the caveat that the month of rabies detection does not necessarily match the month in which the animal was infected. While the National Rabies Surveillance System (NRSS) is responsive to human exposure risks as outlined by the CSTE [38,42,60], there may be a reduced need for enhanced raccoon rabies surveillance (e.g., by USDA) during the summer months in the north [8].
We predicted that the probability of a raccoon rabies case report would be higher in the northern latitudes. We found this to be broadly true, as there were significant interactions between latitude and month leading to higher raccoon rabies presence as latitude increases in the fall, winter, and spring. The surprise here was the notable drop in raccoon rabies presence in the summer in the far north, despite increased sampling. This could be due to synchronous cases occurring during the winter months, leading to a reduction in potentially infectious interactions in the summer, as might be expected for the dynamics of an acute disease with almost-always fatal outcomes in a spatially structured population [67]. There were also significant interactions between latitude and population density (urbanness), with generally higher raccoon rabies presence in northern climes than southern climes for suburban and urban counties. This nuance could guide the distribution of educational resources in the north (more urban/suburban) vs. the south (more rural) and highlights the importance of adaptive surveillance [68–70]. In our analysis, we aggregated across years, possibly obscuring an effect of the time-since-introduction, which varies across the spatial extent of our data. In particular, one might expect that the relatively recent introduction of rabies to the northern states (i.e., within the past 50 years) might be consequential to disease dynamics. We do not see any evidence for this in our data. While the overall number of submitted samples and positive cases shows a slight decrease over the thirteen years of our data, the slope of this trend is independent of latitude (Fig C in S1 Text).
Urban areas are important for public health and wildlife managers because not only are there more opportunities for rabid raccoons to interact with high densities of humans and associated companion animals, but urban areas are more challenging and expensive to manage with oral vaccine baiting [71,72]. Although we predicted that the probability of a raccoon rabies case report would be higher in urban areas, our results are complicated because of the interaction with latitude and important sampling limitations. In general, there was slightly higher raccoon rabies presence in counties with intermediate human population density in the mid-latitude and northern latitudes; this could potentially be explained by more continuous swathes of favorable raccoon habitat in counties with intermediate population density than in high population density, but still many humans to report raccoons. What was contrary to our expectations was the higher raccoon rabies presence in rural (low human population density) counties, compared to urban counties (with the exception of very high human density counties in the south). We expected that because urban areas have high densities of raccoons, there are more opportunities for human-raccoon interactions, leading to more opportunities for a sample to be tested (and hence for rabies to be detected) in higher density settings. One potentially contributing factor is spatial and demographic bias in suspected rabies sample submission [73,74]. For instance, people in rural areas might be more aware of when an animal is acting strangely, increasing the likelihood that a submitted sample would be found rabid, should it rise to the standard of warranting submission. In contrast, submissions from urban areas might be more independent of the likelihood of infection. Increased and targeted education could reduce this discrepancy, by levelling a priori expectations between urban and rural populations.
Our counterintuitive urbanness results might have to do with challenging sampling issues. The surveillance data were mostly passive and states and counties may vary in surveillance intensity and consistency [11,75]. For example, raccoon rabies is more recently endemic in the north, perhaps with consequently more recent education, awareness, and reporting; this could have led to the northern counties being more uniformly sampled than southern counties. Passive surveillance usually underestimates true disease prevalence [76], and can lead to reporting issues which vary by raccoon and human density. For instance, raccoons are often found in high densities close to human areas (i.e., urban areas), which could potentially account for higher rates of submission for rabies testing in urban areas [41]. Conversely, it is hard to tell if raccoon rabies is not present in a county with low human population density, or if surveillance may be inadequate [11]. We indeed see this trend in our county-level data – counties with more people had more raccoons submitted for testing, while there were fewer submissions from rural counties, especially in the south (Fig C in S1 Text).
In addition, counties are likely not the best spatial resolution for the study of disease dynamics [11]. Aggregating samples at the county-level may be too broad for estimation of some covariate relationships. For example, counties vary in size and shape (especially in the south), and information can be lost by averaging widely varied habitat types and human population densities over the county scale. Though, it is important to note that we did not see any explanatory relationships between county land area and rabies presence. Likewise, the borders of counties are delineated according to a combination of terrain, anthropogenic features, and political boundaries. While some of these might have consequences for animal movement and/or disease spread, others are more permeable. There could be some misclassification bias as the county of detection might not be the county of infection and the month of detection might be not scale precisely with the month of infection due to the variable rabies incubation period, however we expect similar misclassification across space and time which would add additional noise to our results. Future work looking into how these results might map onto a more granular view of the spread of disease is needed [11].
The distribution of “urbanness” is also heterogeneous within counties, and county boundaries and urban areas do not necessarily align. We tried to mitigate this challenge by using model selection to choose from a variety of plausible urbanness proxies, but because there is no robust metric for urbanness at the sub-county level, we should view our urbanness results with caution. Having access to exact sample locations, or at least townships, would be ideal. In 2022, the Council of State and Territorial Epidemiologists voted to add sub-county location data to the list of recommended variables for rabies reporting (to the state) and notifications (to CDC) [38]. Spatial understanding of rabies cases analyzed through passive surveillance data systems, and consequent potential for spatial targeting of management and interventions, may therefore improve in future years.
Despite the challenges studying urbanness with county-level data, ultimately, we did find a significant interaction between human population density and latitude on rabies presence, in addition to the more expected relationships with temporal lag, spatial effect, and total number of tests. Rabies is more likely to be detected in counties that had positive tests in previous months and/or in neighboring counties in the recent past, in counties with greater sampling effort. Yet, rabies dynamics also have complex relationships with latitude, urbanness, and seasonality, impeding generation of generalizable predictions across large-scale landscapes. These are complex systems with trends that are less absolute and more nuanced. Identification of temporal and spatial trends in where cases are most likely to occur can help with a better understanding of the factors driving endemic zoonotic disease dynamics, and can help focus public education, surveillance, and management efforts on higher-risk time periods and areas [77].
