We investigated temporal and spatial trends in reporting of hemorrhagic disease (HD) in the midwestern and northeastern US using a 33-yr (1980–2012) questionnaire-based data set. This data set was supported by an additional 19 yr (1994–2012) of bluetongue virus (BTV) and epizootic hemorrhagic disease virus (EHDV) isolation results from clinically affected white-tailed deer (Odocoileus virginianus) in these regions. Both the number of counties that were reported positive for HD and the northern latitudinal range of reported HD increased with time. A similar increase was observed with both the number of states annually reporting HD and the number of counties where HD was reported. Large-scale outbreaks occurred in 1988, 1996, 2007, and 2012, and the scale of these individual outbreaks also increased with time. The predominant virus isolated from these regions was EHDV-2, but the prevalence of EHDV-6, which was first detected in 2006, appears to be increasing. Temporally, the extent of regional HD reporting was correlated with regional drought conditions. The significance of increases in reported HD and the incursions and establishment of new BTV and EHDV in the US currently are unknown.
Hemorrhagic disease (HD) is caused by related orbiviruses in the bluetongue virus (BTV) and epizootic hemorrhagic disease virus (EHDV) serogroups of the family Reoviridae (Howerth et al. 2001). Both viruses are transmitted by hematophagous midges of the genus Culicoides, and for BTV, its geographic range is determined by the distribution of competent Culicoides species and environmental conditions affecting vectorial capacity and virogenesis in these vectors (Mullens et al. 2004; Schmidtmann et al. 2011). In North America, competent vectors for BTV include Culicoides sonorensis (Holbrook et al. 2000) and Culicoides insignis (Tanya et al. 1992), but a potential for additional Culicoides vectors, especially for EHDV, may exist (Smith et al. 1996).
The epidemiology of HD is complex. In white-tailed deer (Odocoileus virginianus), which is the species most affected by BTV and EHDV in North America, clinical disease and potential population impacts are highly variable throughout their range (Howerth et al. 2001). In general, the disease is most severe in more northern latitudes, where HD occurs sporadically and where population immunity is minimal (Howerth et al. 2001; Park et al. 2013). In contrast, areas of enzootic stability exist in portions of the southwestern US, where high infection rates occur annually, but clinical disease is rarely reported (Stallknecht et al. 1996; Flacke et al. 2004). In white-tailed deer, this variation in clinical response is believed to be associated with variation in both acquired (Gaydos et al. 2002b) and innate immunity (Gaydos et al. 2002a). In fawns, passive immunity also may be protective, but this has not been sufficiently evaluated (Gaydos et al. 2002c).
There is recent evidence that the epidemiology of bluetongue and epizootic HD is changing globally. The mechanisms behind this change are not clearly understood and most likely involve multiple and interrelated climatic factors affecting vectorial capacity. In Europe, numerous recent outbreaks involving different BTV serotypes have been reported, and extensive outbreaks involving BTV-8 have occurred in areas where these viruses previously did not exist (Saegerman et al. 2008). Likewise, EHD and related disease in cattle was detected in Israel (Yadin et al. 2008), Turkey (Temizel et al. 2009), and Réunion Island (Bréard et al. 2009). In the US, multiple exotic serotypes of both BTV and EHDV have been detected since the late 1990s (Gibbs et al. 2008), and in the case of EHDV-6, which was first detected in the US in 2006, genetic reassortment with indigenous EHDV-2 also has occurred (Allison et al. 2010). Deer mortality in the US has been associated with EHDV-6, and other EHDV serotypes that are not currently present, such as EHDV-7, can cause severe disease and mortality in experimentally infected white-tailed deer (Ruder et al. 2012).
Despite the changing patterns reported from Europe and the detection of new BTV and EHDV in the US, a northern expansion of HD in North America has not been thoroughly investigated. Such an expansion could be characterized not only by an increase in latitudinal range but also by an increase in disease frequency and intensity. We used a 33-yr data set to determine if such changes are evident. We also attempted to temporally correlate these potentially changing patterns with the presence of regional drought conditions.
MATERIALS AND METHODS
The reports of HD used in this study were derived from a long-term (1980–2012) HD survey initially described by Nettles et al. (1992). As part of this survey, all state wildlife agencies in the US annually report HD occurrence on a county basis. Survey and reporting formats have not changed between years. During 1980 and 1981, data were limited to the southeastern US, but all states were included during and after 1982. The four case criteria for HD reporting have remained constant from 1980 to 2012 and consist of 1) sudden and unexplained deer mortality that occurs in late summer and early autumn; 2) necropsy-based diagnosis of HD based on gross lesions; 3) isolation or molecular-based detection of EHDV or BTV; and 4) detection of deer with sloughing hooves, oral ulcers, or scars on the rumen mucosa. Criteria 1–3 can be associated with direct mortality, while reports based on criterion 4 indicate detection of common HD sequelae that can be associated with morbidity and survival or indirect mortality (Howerth et al. 2001).
For our analyses, we determined if the distribution, northern range, intensity, and frequency of reported HD in white-tailed deer in the northeastern (NE) and midwestern (MW) US (Fig. 1) has increased during the last 33 yr. To visualize a potential increase in HD distribution, we mapped HD reports by county and decade (1980–89, 1990–99, and 2000–09). For this analysis, the northern range was estimated based on the northernmost county where HD was reported during the 10-yr period. To avoid potential outliers, the selected county had to be adjacent to or in proximity (separated by not more than one negative county) to other counties reported as HD positive during that decade. To provide quantitative evidence of a latitudinal range expansion of reported HD, we also defined the annual northern range of HD in the NE US to be the average latitude of the top 20% of HD-positive counties ranked by latitude. The HD-positive counties were those that confirmed HD mortality or morbidity in white-tailed deer in a given year according to the four criteria outlined previously. To test the hypothesis that the northern range of HD was extending, we fitted a linear model and used this to additionally quantify the rate of increase.
As an index to intensity, we used the number of states and counties reporting HD within the combined regions annually. The frequency of reported HD activity was based on the probability of a report (number of years HD was reported/10 yr) within the decade. For 1980–89, a denominator of 8 yr was used for those states that were initially included in the survey in 1982. To account for latitudinal variation, states were stratified into three tiers by latitude. The upper tier included Minnesota, Wisconsin, Michigan, and New York (above 41°43′N). The middle tier included Iowa, Indiana, Illinois, Ohio, Pennsylvania, and New Jersey (between 41°43′N and 39°43′N). Lower-tier states included Missouri, Kentucky, West Virginia, Maryland, Delaware, and Virginia (below 39°43′N). To provide a basis for comparison, we also determined the number of states and counties that were reported as HD positive in the southeastern US, where this disease is endemic. States included Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, and Tennessee. These data were collected at the same time using the same survey.
Because the reported data largely were based on syndromic case recognition (criteria 1, 2, and 4) by numerous independent sources, we also attempted to validate results with virus isolation (criterion 3). Virus isolation data from white-tailed deer in this region were available from 1994–2012; these were restricted to virus isolations done by the Southeastern Cooperative Wildlife Disease Study (SCWDS) and, in some cases, viruses submitted to SCWDS from state veterinary diagnostic laboratories for confirmation and serotyping. Virus isolation results include data derived from wild deer and farmed or penned deer. Virus isolation protocols are described by Quist et al. (1997), and as of 2007, isolation attempts have been limited to a tissue-culture–based isolation protocol using cattle pulmonary endothelial cells (American Type Culture Collection, Manassas, Virginia, USA).
For the drought analysis, the total numbers of HD-positive counties per year were compared with regional drought conditions during that year. Drought conditions were estimated based on tabular data available through the US Drought Monitor, which is produced in partnership between the National Drought Mitigation Center at the University of Nebraska-Lincoln, the US Department of Agriculture, and the National Oceanic and Atmospheric Administration (National Drought Mitigation Center 2014). Based on the seasonality of HD in deer in which peak cases occur in August and September (Howerth et al. 2001), drought data were used from the last week in August. To represent drought severity, and geographic scale, a cumulative index value was calculated by summing the total percent areas within the five drought categories: D0 (abnormally dry)–D4 (exceptional drought); D1 (moderate drought)–D4; D2 (severe drought)–D4; D3 (extreme drought)–D4; and D4. As regional drought data are presented separately for the MW and NE states, our overall values represent a mean of the two regional indices.
When evaluated in 10-yr intervals from 1980 to 2009, the northern range of HD reports increased with time (Fig. 1). This northern expansion continued after 2009, with the addition of new HD-positive counties during 2010–12 (Fig. 2A). From 1980 to 2012, the mean annual northern range also increased with time (linear model, P<10−16, 0.3° per year; Fig. 3). The increase in range from 1980 to 2012 is supported by increasing annual trends in both the number of states (Fig. 4A) annually reporting HD and the number of counties within the region where HD was reported (Fig. 4B). The increasing trend in number of reports of HD by county also is characterized by several annual peaks (1988, 1996, 2007, and 2012; Fig. 4B); these are associated with large regional outbreaks as occurred in 2007 and 2012 (Fig. 5).
In addition to an increase in the range and numbers of reported HD-positive counties and states, the frequency of these reports also increased with time (Table 1). For the combined regions, the probability for reported HD in a given year increased from 0.26 in 1980–89 to 0.53 for 2000–09. The increase was consistent throughout the entire study area, with the exception of Vermont, New Hampshire, Rhode Island, Massachusetts, and Maine, all of which have remained negative, and Connecticut for which there is a single report from 1992. In the HD-positive states, the probability of HD in a given year decreased with increased latitude (Table 1). Based on the last 3 yr of data (2010–12), again excluding the six states that have remained negative, this trend has continued, with a region-wide probability of an annual HD report of 0.79. For the three latitudinal tiers, these probabilities during 2010–12 were 0.55, 0.80, and 0.94 for the upper-, middle-, and lower-tier states, respectively. Within the middle-tier states, a west to east difference in the historic frequency of reporting also is apparent with more overall annual reports from Illinois, Indiana, and Iowa (41/84 survey years [number of states×number of years reporting], 0.49), than observed in New Jersey, Ohio, and Pennsylvania (13/84, 0.15).
Results from southeastern states are not consistent with trends observed in the combined data from the MW and NE states (Fig. 6A). The probability of annually reporting HD was 0.79, 0.83, and 0.83 for 1980–89, 1990–99, and 2000–09, respectively. For the period 2010–12, the probability of an annual report was 1.0. The number of counties reported positive in the Southeast increased from 1980 to 2012, but based on linear regression results (Fig. 6B), the estimated rate of increase (2.00 counties per year) was lower than observed in the MW and NE data (7.11 counties per year).
Most (61%) of the reports of HD were associated with criteria 1–3, indicating a direct mortality event. Overall, 99% of criterion 4-only reports were from five states: Virginia (806; 80% of total reports from Virginia), Kentucky (42; 14% of total), Indiana (37; 22% of total), Maryland (37; 34% of total), and West Virginia (18; 22% of total). In the remaining states of Delaware, Illinois, Iowa, Michigan, Missouri, New Jersey, Ohio, Pennsylvania, and Wisconsin, criterion 4-only reports represented 2% of the 784 reports from these states.
Many of these HD reports included confirmation by virus isolation (criterion 3; Fig. 2B); HD was confirmed by virus isolation of EHDV or BTV in all states that reported positive results during this study. During 1994–2012, virus isolations included EHDV-1, -2, and -6 and BTV-10, -11, -13, and -17 (Table 2). Of these, EHDV-2 has been the predominant virus (452/529), representing 85% of isolates. The detection of BTV was limited to Missouri.
With the combined NW and NE data (all states), the number of counties reported positive for HD and the drought index followed similar temporal patterns (Fig. 7A); however, these trends were more similar after 2006. For the 2000–12 period, the number of counties reported positive for HD was positively correlated with the drought index (Fig. 7B), and the degree of correlation increased if the data set was limited to the 2006–12 period (Fig. 7C).
Time series data are consistent with a northern range expansion for reported (Figs. 1, 2A, and 3) and confirmed (Fig. 2B) HD, an increase in the number states and counties reporting HD within a given year (Fig. 4A, B), and an increase in the frequency of these positive reports (Table 1). The case-based data reflecting these trends, however, are vulnerable to potential reporting biases; data were obtained from independent sources over extended periods, and the specific methods of data collection could not be standardized between states. Overall, we do not believe that these potential biases impacted our results. Although data collection methods varied among states, the four criteria in our case definition did not change among years. The specificity of our case definition is supported by a lack of outliers, spatial and temporal consistency of reporting across state lines, a lack of reports from New England states, reporting trends that differed from the endemic southeastern states, and finally, through confirmation of these reports through virus isolation.
Outliers over the entire period were limited to reported disease in one county each in northern Wisconsin and Connecticut, and in both cases, these were reported as criterion 4 observations (hoof lesions). With the exception of Virginia, hoof lesions (criterion 4) are not commonly reported in any of the states included in this study, and increased reporting of hoof lesions from Virginia can be explained by two factors. Virginia has a very robust sampling scheme involving large numbers of examined animals (10% of the harvest, 16,658±2,881 deer/year) and the ability to detect criterion 4 HD at a prevalence below 5% (Sleeman et al. 2009). However more importantly, the detection of hoof lesions in Virginia is not unexpected, as this is a recognized sequela to BTV and EHDV infection that is commonly reported in the Southeast; this is especially common in white-tailed deer from the coastal plain that extends into Virginia and Maryland (Howerth et al. 2001). To determine if this more robust sampling scheme or the inclusion of positive reports related to hoof lesions biased our observed reporting trends, we removed the Virginia data (86% of the criterion 4 reports were from this state) from the combined NE and MW data set. The estimated annual increase in the number of states reporting HD from 1980–2012 increased slightly from 0.248 (linear regression for entire data set; y = −487.51+ 0.248*x; r2 = 0.536) to 0.260 (linear regression minus the Virginia data; y = −512.98+0.260*x; r2 = 0.548). The number of counties annually reported positive for HD were reduced without the Virginia data, but regional trends did not change (Fig. 6C).
Spatial and temporal consistency in reporting across state lines is apparent in the HD distributions presented for the overall reported distribution of HD (Fig. 2A) and in the outbreaks during 2007 and 2012 (Fig. 5). There also is consistency between states and within subregions related to frequency trends of HD reporting (Table 1).
Although we cannot discount the possibility of false-positive reports, the lack of any positive reports and virus isolations from the New England states (Maine, Massachusetts, New Hampshire, Rhode Island, and Vermont) suggests that these were minimal. New England is outside of the known distribution of C. sonorensis (Tabachnick 1996), and consistent with our results, there were no BTV or EHDV isolations reported from livestock tested from these states from 1991 to 2002 (Ostlund et al. 2004).
The observed increases in both the number of states reporting HD and the number of counties reported positive annually for the combined NE and MW data set differed from the Southeast, where HD is endemic (Fig. 6A, B). In the NE and MW states, the probability of a state reporting HD increased from 0.26 (1980–89) to 0.53 (2000–09), representing a more than a twofold increase. In contrast, the probability of a state reporting HD in the Southeast increased from 0.79 (1980–89) to 0.83 (2000–09), representing a 1.05-fold increase (Fig. 6A). Likewise the number of counties reported HD positive annually in the combined NE and MW regions increased (based on linear regression models) at an estimated rate of 7.11 counties per year as compared with 2.00 counties per year for the southeastern states. As all of these regions were surveyed at the same time using the same methods and subject to the same potential reporting biases, it is unlikely that the overall spatial and temporal trends that we report can be attributed to reporting artifacts. We cannot measure or discount possible effects related to increased disease reporting due to better regional awareness of HD over time. Such increased awareness could be more pronounced after an initial HD report from a state or region or following large outbreaks that occurred during 2007 and 2012. However, given the importance of the white-tailed deer resource to the agencies that have participated in this survey since 1980, it is unlikely that large scale mortality as seen in recent outbreaks would have gone undetected and unreported. It is also unlikely that regional awareness would be synchronized between states within a given year or that such awareness would gradually expand in a northerly direction during this period without a cause.
Based on virus isolation results, most of the reported HD in white-tailed deer in the NE and MW states included in this study was associated with EHDV-2 (Table 2). During the large-scale regional outbreaks during 2007 and 2012, EHDV-2 represented the predominant virus isolated from white-tailed deer. The over representation of EHDV-2 is consistent with previous studies in the southeastern US (Stallknecht et al. 1995). Although EHDV-6 was initially detected during 2006 from white-tailed deer in Illinois and Indiana (Allison et al. 2010), the numerous isolations during 2012 were unexpected. During 2012, EHDV-6 represented the predominant virus isolated in Michigan and Wisconsin, and it was documented for the first time from a NE state (Maryland). From 2006 to 2011, EHDV-6 was consistently isolated from deer in the US at low levels, but it was never associated with either regional or large-scale outbreaks. Although the current evidence supports a successful introduction, the long-term significance of this introduction remains unknown. The few BTV isolations reported in this study are consistent with historic data for these regions. With the exception of Missouri and Kentucky, a very low prevalence of BTV antibodies (market cattle surveys) and BTV isolates (National Veterinary Services Laboratory, Veterinary Services, Animal and Plant Health Inspection Service clinical submissions) were detected from these states from 1991–2002 (Ostlund et al. 2004). Additional test results from cattle in Illinois (2000–02) support these results with a low reported prevalence (1–2%) of BTV antibodies (Boyer et al. 2007). Consistent with our virus isolation results, Boyer et al. (2008) reported that prevalence of EHDV antibodies in these same 2000–02 cattle sera ranged from 5% to 15%. The apparent variation in the prevalence of EHDV and BTV may relate to differences in the distribution of Culicoides vectors associated with these viruses. The distribution of C. sonorensis, for example, may be defined by both climate and soil characteristics affecting larval habitats, and these factors could limit the distribution of BTV in upper MW and NE states (Schmidtmann et al. 1998, 2000, 2011). In contrast, with EHDV, field evidence suggests that other Culicoides species may be involved in EHDV transmission in the eastern US (Smith et al. 1996).
The mechanisms behind the apparent northern expansion and increased reporting of HD are unknown, but probably are related to numerous interacting habitat, climatic, and landscape variables, including temperature, precipitation, predominant wind patterns, and wetlands; all of these can affect vector distribution, population size, and vector competence (Baylis et al. 2004; Mullens et al. 2004; Berry et al. 2013). There have been few comprehensive analyses of the drivers of disease outbreaks due to EHDV and BTV in the NE and MW US. In cattle in Illinois, BTV and EHDV antibody prevalences were positively associated with temperature and with EHDV, forest patchiness (Boyer et al. 2010). Using the Virginia subset (1993–2006) of the same HD data set that we used in the present study, Sleeman et al. (2009) reported direct correlations between HD incidence and average temperature for winter, early summer, and late summer, and an inverse relationship with June precipitation. In a subsequent spatial-temporal analysis of HD data from Alabama, Georgia, North Carolina, and South Carolina using a generalized linear mixed model framework, wind speed, dew point, rainfall, and land surface temperature all were related to HD occurrence (Xu et al. 2012).
Based on the identification of a potential negative correlation between HD and rainfall by Sleeman et al. (2009) and Xu et al. (2012) and the ease of retrieving drought-related data and maps through the US Drought Monitor, we concentrated on this single variable. On a region-wide basis, late August drought conditions, as represented on the US Drought Monitor, were correlated with the observed temporal HD patterns (Fig. 7A–C). The correlation was improved (Fig. 7C) if the analysis was restricted to years 2006–12. This may relate to an increase in the geographic distribution of HD during these later years and more numerous state and county reports (Fig. 4A, B). Although not specifically analyzed, spatial patterns in extreme outbreak years also appear to coincide (Fig. 5). During 2007, extreme drought conditions within the MW and NE US during the last week in August were centered in Kentucky and during 2012, in Missouri. Although it is unlikely that this single variable would be sufficient to explain HD patterns, extreme drought may be an important risk factor associated with large-scale HD outbreaks, and this possibility deserves additional attention. Drought has also been suggested as a risk factor with Schmallenberg virus in Europe, which also may be dependent on Culicoides transmission (Calzolari and Albieri 2013).
As related to wildlife population health, the significance of increases in reported HD and the incursions and establishment of new BTV and EHDV in the US are unknown. Furthermore, although the changing patterns we describe were measured over 33 yr, there is no basis to determine if they will persist. However, based on limited field and experimental data, incursions of these viruses into naïve white-tailed deer populations can result in estimated population losses of 6–16% (Missouri; Fischer et al. 1995) and 20% (West Virginia; Gaydos et al. 2004). If such outbreaks occur more commonly on a 2- to 3-yr cycle, as reported for Virginia and the southeastern US (Howerth et al. 2001; Sleeman et al. 2009) and if changing climate conditions result in more numerous large-scale regional outbreaks as observed in 2007 and 2012, potential population impacts or management implications related to increased nonhunting mortality are possible. In addition, this potential problem could be made worse with the successful introduction of new viruses. New viruses, such as EHDV-6, provide both antigenic and genetic diversity that could potentially result in increased viral fitness and a means to escape the positive effects of increasing herd immunity. At present, management for HD has been limited to isolated cases in which reductions in legal deer harvest were attempted following large-scale HD mortality events. The effectiveness of this approach has not been tested and, at present, we do not have the knowledge to predict or mitigate potential outcomes related to HD range expansions.
This study was made possible through continued financial support from the member states of the SCWDS provided by the Federal Aid to Wildlife Restoration Act (50 Stat. 917) and through long-term cooperative agreements with the US Department of Agriculture, Animal Plant Health and Inspection Service. This study was made possible by the efforts of numerous staff biologists and technicians affiliated with state and federal wildlife agencies that annually submitted case data and provided field samples for virus isolation. Additional samples and virus isolates were provided by state veterinary diagnostic laboratories in Michigan, Illinois, Pennsylvania, Indiana, and Ohio. Laboratory and data management support at SCWDS were provided by D. Mead, B. Hanson, R. Poulson, and C. McElwee.