We detected antibodies against CCHFV in 84 of the 2,440 (3.4%, 95% CI 2.8%–4.2%) small ruminants tested. Exposure prevalence among sheep and goats was similar: sheep 3.0% (95% CI 2.1%–4.1%; 36/1,220) and goats 3.9% (95% CI 3.0%–5.2; 48/1,220). At least 1 seropositive animal was found in 16 of 61 (26.2%, 95% CI 16.8%–38.4%) surveyed sheep farms and in 18 of 61 (29.5%, 95% CI 19.6%-41.9%) goat farms. Overall, the number of farms with >1 seropositive animal was 34 of 122 (27.8%, 95% CI 20.7%–36.4%) (Figure 1).

We excluded 2 farms (1 sheep farm and 1 goat farm) and a total of 40 (seronegative) animals from the risk factor analysis because of incorrect recording of location coordinates. The best-fitted model selected 3 of the considered predictors, including cattle/horse density, annual NDVI, and annual NDVI variance (Table 2). Cattle/horse density around small ruminant farms was significantly associated with exposure probability (Figure 2). A positive, albeit not statistically significant, relationship was also observed for the NDVI. In contrast, we observed a strong and statistically significant negative relationship between the annual NDVI variance and the risk for exposure of individual small domestic ruminants to CCHFV. The cross-validation analysis showed that the balanced accuracy of the model was 0.549, sensitivity was 10.2%, and specificity was 99.7%. Also, the model had good discriminatory power (area under the curve = 0.830). For a prior probability of infection of 0.05, the positive predictive value was 0.6415 and the negative predictive value was 0.9547. The spatial projection of the model showed lower predicted risk areas in central and eastern Andalusia and higher predicted risk areas north and south of the region (Figure 3).

Although some of the human cases of CCHF in Spain could be associated with farm animals, research on domestic species is limited (38). Most studies in Spain have focused on wild ungulates because of their relevance to CCHFV (24,25), but even so, farm animals and their ticks may pose a risk for humans through closer contact. By selecting small domestic ruminants, we aimed to identify the areas of greatest risk for transmission to persons in contact with them and thus complement the risk maps previously obtained by using red deer (13). Some of the CCHF patients in Spain were animal farmers (38). In the absence of an effective vaccine to protect humans against CCHFV infections, the only feasible measure to protect the population at risk is prevention of tick bites. If risk for infection is higher for an animal in a territory, it is also higher for a person in that territory because the virus is transmitted mainly by tick bites to animals and humans. The factors that predispose human contact with infected ticks or animals will determine the actual risk for CCHFV infection (12), although risk will be greater in areas more environmentally favorable for virus transmission. Our results provide public health authorities with information about which areas of Andalusia have the highest risk for CCHFV transmission for anyone linked to small ruminant production. That information will enable design of better infection surveillance programs in the region, optimizing the resources available for those programs and improving the cost:benefit ratio of preventive actions.

Our study is based on a representative subsample of small ruminants for the study region. However, we did not include other domestic species (cattle, equids) or wild ungulates, which are relevant for CCHFV and its vectors and could improve the predictive capabilities of the model. In the future, we recommend considering the full range of species involved in CCHFV transmission to improve the accuracy of risk maps. In this study we did not corroborate that antibodies were specific for CCHFV, which should preferably be done by comparative neutralization assays in Biosafety Level 4 facilities. However, numerous positive serum samples from wild ungulates in the ELISA used were confirmed as CCHFV positive by neutralization assays against different orthonairoviruses (I. García-Bocanegra, unpub. data). The manufacturer of the multispecies double-antigen ELISA test used confirms 98.9% sensitivity and 100% specificity after testing with a multitude of animal species (29).

In a previous study conducted in 2 areas of Andalusia (Córdoba and Cádiz Provinces), and using the IDVet double-antigen ELISA, CCHFV seroprevalence among domestic ruminants (including cattle, sheep, and goats) was 17.9% (38). The inclusion of cattle samples makes it difficult to compare reported seroprevalence with our findings because cattle host higher burdens of Hyalomma spp. ticks than do small domestic ruminants (39) and may thus be more prone to CCHFV exposure. That study also found higher seroprevalence than the seroprevalence that we report for areas of northern Spain, which are less favorable areas for Hyalomma spp. ticks (6.8%) (38). The higher antibody prevalence most likely results from including cattle in the survey. Our results also contrast markedly with the high antibody prevalence (76%–87%) observed in red deer in western Andalusia (40). Previous studies of small domestic ruminants from Africa, Asia, and Europe showed a wide range of seroprevalence, 0.4%–74% among sheep and 2.1%–66% among goats (28). Our results agree with those of some studies conducted in the Mediterranean region (41,42). The low individual seroprevalence observed for animals of both species indicates that sheep and goats are of less concern than cattle, horses, or wildlife for farmers and public health authorities.

Our selected model included horse/cattle density as a predictor of CCHFV exposure risk. The lack of association between wild ungulate abundance and CCHFV exposure risk most likely indicates a low rate of interaction with the small domestic ruminant farms selected for the study. The rate of interaction between wild ungulates and their ticks and small ruminants raised on extensive farms that are also used for large game hunting is probably higher, perhaps leading to a higher risk for exposure to CCHFV. Cattle and horses are relevant hosts for H. marginatum ticks (6,23,32,41) and may be more abundant than wild ungulates in the vicinity of small ruminant farms, so their association with CCHFV was not unexpected. Previous studies already described the relevant role of farm animals in the risk for exposure to CCHFV (24,43). Among domestic animals, global CCHFV seroprevalence is second highest among cattle, after camels (44). Thus, our findings suggest that a regional strategy should perhaps be implemented to better control ticks on farm animals. For sheep and goats, increasing the frequency of acaricide application may result in more effective tick control (41).

The best-fitted model also included 2 abiotic predictors, NDVI and annual NDVI variance, which would probably define the environmental (climatic) niche for CCHFV vectors in southern Spain. The distribution of ticks is limited not only by host distribution but also by a combination of host presence/abundance and environmental favorability (17,45). Ticks occupy only a subset of their host range because they undergo a large part of their cycle on the ground, where abiotic factors determine tick development and survival rates (17,32). One of the limiting factors for tick survival and activity is moisture level, a determinant of tick abundance and a relevant driver of CCHFV transmission risk (13,17). The negative relationship observed between annual NDVI variance and CCHFV exposure risk may suggest that areas with substantial fluctuations in vegetation productivity (e.g., seasonal croplands) are unfavorable for CCHFV vectors, a possibility that agrees with the low predicted spatial risk in the agricultural lands of the Guadalquivir River basin.

The spatial distribution of the farms with seropositive animals was heterogeneous; most were distributed south and east of the study region. Because our model showed that this distribution was associated with some biotic and abiotic factors of the farm neighborhood, we were able to capture the environmental niche for small domestic ruminant exposure risk to CCHFV. The model projection identified that the areas of Andalusia with the highest abundance of wild ungulates (mainly red deer, wild boar, and Iberian ibex [Capra pyrenaica]) (46) had the highest risk for exposure to CCHFV. As previously observed for red deer (13), the predictive model not only projects the risk for small domestic ruminants but also for other hosts of CCHFV vector ticks. The model identified the areas of highest risk to be Los Alcornocales Natural Park (Cádiz and Málaga Provinces), the area surrounding the Doñana National Park (south of Huelva Province), most of the Sierra Morena mountain chain, and the Sierras de Cazorla, Segura y Las Villas Natural Park (northeastern Jaén Province). Sánchez-Seco et al. (26) found CCHFV-positive Hyalomma spp. ticks in Los Alcornocales Natural Park, whereas we detected high CCHFV prevalence among ticks and high antibody levels in the wild ungulates of Doñana National Park (24) and identified Los Alcornocales Natural Park, Doñana National Park, and the Sierra Morena mountain chain as high-risk areas (13,25). Recently, we found that ≈30% of wild boar in Sierras de Cazorla, Segura y Las Villas Natural Park have antibodies against CCHFV (47). Our risk map identifies areas of low and high risk that were identified on larger spatial resolution on a map generated for Spain from a model based on red deer (13). However, the limited sensitivity (10.2%) of our model to predict the risk for exposure of small domestic ruminants to CCHFV prevents us from detecting all areas where CCHFV may be circulating among small domestic ruminants in Andalusia. Therefore, in the future, it would be desirable to base estimates of CCHFV actual distribution in Andalusia on vector population dynamics and CCHFV prevalence among the vectors. Comparison of the findings of the small domestic ruminant-based model with existing evidence on the prevalence of CCHFV infection/exposure and the predictive outcome of wildlife-based risk models indicates that despite its limited predictive sensitivity and tendency to false negatives, our model can capture spatial foci of high and low CCHFV risk. Consequently, despite the observed limitations, it may constitute a useful tool for preventing cases of CCHF in humans. The high specificity of the model indicates that the identified low-risk hotspots are actually zones with low risk for exposure. We conclude that modeling of CCHFV exposure risk for small domestic ruminants, although at low rates of virus exposure, is a useful tool for mapping CCHFV transmission risk hotspots and preventing CCHF in humans, at least in the study area.