Spatial-temporal analysis of the risk of Rift Valley fever in Kenya
MetadataShow full item record
Bett, B., Omolo, A., Notenbaert, A. and Kemp, S. 2012. Spatial-temporal analysis of the risk of Rift Valley fever in Kenya. Presentation at the 13th Conference of the International Society of Veterinary Epidemiology and Economics, Maastricht, The Netherlands, 20-24 August 2012. Nairobi, Kenya: ILRI.
Permanent link to cite or share this item: https://hdl.handle.net/10568/21772
External link to download this item: http://www.slideshare.net/ILRI/spatialtemporal-analysis-of-the-risk-of-rift-valley-fever-in-kenya-14093492
Historical data on Rift Valley Fever (RVF) outbreaks in Kenya covering the period 1951-2010 were analyzed using a logistic regression model to identify factors associated with RVF occurrence. The analysis used a division as the unit of analysis. The infection status of each division was defined on a monthly time scale and used as a dependent variable. Predictors investigated include: monthly precipitation (minimum, maximum and total), normalized difference vegetation index, altitude, agro-ecological zone, presence of game, livestock and human population densities, the number of times a division has had an outbreak before and time interval in months between successive outbreaks (used as a proxy for immunity). Both univariable and multivariable analyses were conducted. The models used incorporated an autoregressive correlation matrix to account for clustering of observations in time, while spatial smoothing was done using Bayesian techniques. Functional relationships between the continuous and the outcome variables were assessed to ensure that the linearity assumption was met. Descriptive analyses indicate that a total of 91 divisions in 42 districts (of the original 69 districts in place by 1999) reported RVF outbreaks at least once over the period. The mean interval between outbreaks was determined to be about 43 months. Factors that were positively associated with RVF occurrence include increased precipitation, high outbreak interval and the number of times a division has been infected or reported an outbreak. The model will be validated and used for developing an RVF forecasting system which can then be used with the existing regional RVF prediction tools such as GLEWS to downscale RVF risk predictions to country-specific scales and subsequently link them with decision support systems. The ultimate aim is to increase the capacity of the national institutions to formulate appropriate RVF mitigation measures.