| dc.description.abstract | Determination and quantification of variation in disease responses allows estimation of variations in risk at different levels of population organization. In the context of livestock disease control, levels with large unexplained variation can be identified and targeted with the expectation that interventions targeted at that level will have the greatest impact on the response. Sources of variation of East coast fever (ECF) seropositivity in cattle were evaluated using data collected in Mbeere district, Kenya. The data exhibited hierarchical structure as it was collected by multistage random sampling. In the study, serostatus of ECF in cattle was determined by enzyme-linked immunosorbent assay (ELISA). The overall ECF seroprevalence was 19.3% (range: 3.9% to 48% across divisions) [95%CI: 13.7%, 24.9%] indicating that ECF occurrence is endemically unstable in the district. Multilevel models, used for statistical analyses of data that have a hierarchical structure, were used to fit the models using Schall’s algorithm. Two types of models were fit: one incorporating both significant fixed effects and random effects (mixed models) and the other incorporating random effects only (random models). In the mixed models analyses, substantial variation (64%) resided at the herd-level while 43% and 40% of the variation resided at the herd- and division- levels respectively in the random models analyses. The large herd variation in the disease exposure indicated that observations within a herd had a common cause and/or strong dependency in terms of causative parasite exposure, a potential risk factor for within-herd disease transmission of ECF. Disease variation across divisions was thought to be due to differences in vector environmental suitability habitats. These findings suggest that effective ECF interventions in Mbeere district need to be targeted at divisional- and herd-levels. | en |
