Spatial analysis of livestock disease data in sub-Saharan Africa : a scoping review

Abstract

BACKGROUND : In livestock disease surveillance, spatial analysis methods play a major role in the identification of areas where the risk of disease could be higher. Though widely used in human health, their extent and depth of use are not well known in livestock health in sub-Saharan Africa and this has hindered their update in livestock disease modeling. This study set out to provide a comprehensive review of spatial analysis methods and their application in livestock disease data analysis in sub-Saharan Africa. METHODS : Articles were searched using keywords related to spatial and spatio-temporal analysis of livestock diseases in sub-Saharan Africa in PubMed, Web of Science, Embase, and Scopus. Articles were reviewed in terms of name of author, country of study area, study design, livestock species, livestock diseases, research tasks, and spatial epidemiological methods in terms of spatial statistics and models among others. RESULTS : A total of 56 articles were selected for review. Descriptive approaches such as simple maps of incidence and prevalence (n = 22) have been commonly used. Spatial scan statistics of the Kulldorff (n = 15) have also been the common spatial statistics employed. Model based spatial analysis has also been used (n = 14). Key research tasks that have been performed include investigating disease distribution, risk factors, space and time interaction and spatial risk prediction. The shortfalls of the reviewed studies include lack of exploration of irregularly shaped cluster scan statistics in case the actual disease clusters are irregular. There is also lack of use of multivariate scan and joint spatial models in case of multiple groups or diseases to show comorbidity. Model based spatial analysis has not accounted for space and time interaction. Machine learning niche models have failed to account for spatial autocorrelation in the data. Model based spatial risk prediction has mainly been retrospective as opposed to prospective for early warning. CONCLUSION : Future research may consider the application of multivariate scan statistics and joint spatial models for disease comorbidity analysis. It may also explore the use of irregularly shaped cluster scan statistics to enable detection of irregular disease clusters. Research opportunities may also include the use of machine learning models that account for spatial autocorrelation. Future spatial prediction is another area worth exploring to show future disease risk trends for early warning.