 Hello everyone. My name is Valentina Dolo. I'm a PhD student from the University of Cambridge studying at the Department of Veterinary Medicine and today I'm going to present to you a special predictive model of anthrox disease across Kenya using a Bayesian approach implemented by R. Inla. So to begin with, anthrox is a zoonotic disease caused by bacillus and thrases and prior to infection the bacteria exists in the form of dominant spores in the soil that are resistant to extreme environmental pressure often surviving for decades and contributing to the persistence of anthrax outbreaks. Now animals can get anthrax when they graze on contaminated soils and upon ingestion of the spores they change into the vegetative form multiply within the host and cost disease or sometimes death in susceptible animals humans can get anthrax from direct contact with infected animals they exhibit four forms of the disease cutaneous form gastrointestinal form and the inhalational form a fourth and rare form of human disease injection anthrax was recently added to account for the infections observed among heroin drug users. Now anthrax causes environmental contamination massive economic losses in the agricultural sector and it's also a public health burden in several countries and threatens biodiversity. Although anthrax cases have been documented globally Africa has been reported to have the highest prevalence of anthrax disease in livestock now anthrax remains a huge burden in Kenya with a sharp increase in cases reported from the year 2005 and this could most likely be due to improved surveillance following the inception of frameworks such as the zoonotic diseases unit in Kenya. Now some of the ecological drivers of anthrax are well known and include precipitation, temperature, soil amongst others and although these drivers have been used to model the geographic extent of anthrax risk recent studies have applied classical algorithms that cannot capture the underlying spatial dependencies observed in the anthrax surveillance data. Here we apply a Bayesian approach to analyze a long-term spatial dataset spanning 30 years of livestock anthrax case data to investigate the drivers of the geographical distribution of anthrax risk across Kenya. We recorded 582 cases of livestock anthrax from 1991 to 2020 and we obtained this data from the Kenya directorate of veterinary services in Nairobi and five regional veterinary investigation laboratories in Karatina, Nakuru, Eldoret, Kerrich and Marriakani. We also recorded 20 wildlife outbreaks from the Kenya wildlife service and these cases were confirmed through clinical and laboratory diagnosis. Statistical analysis was done using R in LA package which stands for integrated nested Laplace approximation. The map on the extreme left shows the livestock cases in red and the wildlife cases in yellow and what we did was for the livestock cases we designed a 10 kilometer by 10 kilometer grid cell structure as shown in the middle and calculated the number of anthrax cases within each grid cell. So in the end we had a total of 184 grid cells measuring 10 kilometer by 10 kilometer with at least one case of anthrax within each cell after grouping all the 582 case locations. So the image on the right is showing an example of this where you have several cases group them into grid cells and then you obtain the centroid coordinate of each grid cell which now represents the number of case counts per location so these were our new presence locations. We used a zero truncated Poisson likelihood to model the counts of anthrax outbreaks per location because we only had presence data without zeros. The equation for the model is as shown above where CI represents the anthrax case counts per location so these are the 184 grid cells which had at least one anthrax case. The ECI represents the posterior mean of the expected values alpha here is the intercept x is a matrix of the covariates. Betas here represents the linear coefficients while the delta k represents the nonlinear effects and the u represents the spatial random effects and the spatial random effects were obtained using a mesh as shown on the on the right and this mesh was used to calculate the spatial random field and the effects were then added to the model. The covariates used were bio2, mean denot temperature range, bio18, precipitation of the warmest water, elevation, distance to water, soil calcium and soil water. The images above are showing the results of the model. The gray graphs on the right are showing the nonlinear effects of four covariates, bio18 and bio2 on top and soil calcium and soil water below. These were obtained using cubic regression splines with five knots. The red arabas on the right are showing the intercept on top, the fixed effect for elevation in the middle and the fixed effect for distance to water in the bottom. The results of the Bayesian model showed that distance to water bodies were significantly associated with the reduced incidence of anthrax outbreaks. Past studies have demonstrated a significant negative link between distance to water bodies and the suitability of an area for the occurrence of anthrax. This is most likely linked to the fact that most animals use communal watering points thus there's an increased likelihood of observing anthrax outbreaks close to water bodies than further away. Elevation had a positive effect on the incidence of anthrax outbreaks. The remaining four variables had nonlinear effects. Possibly better explained by looking at the effect on the fitted values. The graph on the right shows the covariate values against the fitted values. The table on the left shows the DIC values of the full model and the various versions of the model where a single covariate was removed each time. We also calculated the difference in DIC between the full model and the models missing various covariates. The mean general temperature range Bio2 had the strongest effect on the model and removing it increased the DIC by a magnitude of 83. Bio18 precipitation of the warmest quarter also had a strong effect and the incidence of anthrax cases increased with increasing precipitation up to about 400 mils then reduced. Soil calcium also had a positive effect on anthrax incidence initially but this effect wore off as the values increased beyond 10 and soil water had mostly a negative effect on the incidence of anthrax. These are the results of the spatial random field showing the spatially correlated random effects and these were added to the final model to calculate the fitted values. The maps above show the mean predicted anthrax risk in the middle and the lower credible interval on the left and the upper credible interval on the right. Blue areas are those with lower risk while the warmer colors moving towards red are those with increasing risk. The model prediction showed that most parts of central western and coastal Kenya were at risk of anthrax. However the small pockets of anthrax risk areas in the northern parts of the country specifically in Turkana County were alarming. Now Turkana is classified as an arid and semi arid land with mostly pastoralist communities who rely on mobility to get access to water and grazing resources. These pastoralists are often economically and politically marginalized lacking access to both veterinary and public health services usually available to the rest of the population. As such they're at greater risk of zoonotic diseases like anthrax and the sparsity of recorded outbreaks in this region could reflect the limited surveillance practices and not necessarily the absence of livestock cases. Thus more effort could be put in place to improve anthrax surveillance across this region. By accounting for spatial dynamics we demonstrate an approach that is easy to interpret and replicate for other diseases and this approach is particularly useful for studies that have patchy surveillance data and underlying structural dependencies. This risk model can support the planning of surveillance and prevention campaigns particularly in marginalized pastoralist communities which are disproportionately affected. With that I would like to acknowledge the following my supervisors, the Gates Cambridge Trust, the Royal Geographical Society and the Kenya One Health Online Conference for organizing this amazing event. Thank you. Thank you very much. We can give her a clap and I'm not sure whether we can take questions but I can see there was a question asking why there are no cases of anthrax in northern Kenya but as you want to say something those I saw you trying to answer. She did make a reference to precipitation as a factor and northern Kenya being usually very dry most of the time that can be an explanation. Okay, thank you.