This study analyzed data from Rift Valley fever (RVF) outbreaks and serological surveys in Uganda to map the risk of RVF transmission. Surveillance records, serological surveys of cattle, sheep and goats, and spatial data on climate/ecology were analyzed. Observed outbreaks were mapped to identify occurrence patterns. Serological data were modeled using logistic regression to identify animal-level exposure factors. A geostatistical model predicted spatial patterns of endemic infection. The final serological model was used to predict areas of exposure risk based on ecological variables. Results identified areas at high transmission risk, especially during wet seasons, and pointed to the need for active surveillance in predicted risk areas.
Quantifying Artificial Intelligence and What Comes Next!
Mapping the risk of Rift Valley fever in Uganda
1. Methodology
Data: The study utilized data from three sources: (i) surveillance records on epidemics kept by the Ministry of
Agriculture, Animal Industry and Fisheries (MAAIF), (ii) de novo serological surveys involving cattle, sheep and goats, and
(iii) online spatial data on climate and other ecological variables.
Analyses: Observed epidemics were mapped to obtain occurrence maps.
Serological data were also analyzed using random effects logistic regression model to identify animal-level factors
associated with RVF virus exposure.
Furthermore, a geostatistical model, implemented using RINLA was fitted to serological data to predict spatial patterns
of endemic infections.
Predictions: The final model developed from the analysis of serological data was used to predict the spatial range of
exposure. Predictors used comprised ecological variables only.
Conclusion
Results point to the need for
more active surveillance for
the disease in areas predicted
to have high transmission risk
based on serological data,
especially during wet seasons.
Results
o Initial epidemic was observed in
Kabale, S.W. Uganda in March 2016
after periods of heavy rains.
o A total of 3962 animals were sampled
from 198 households and 150 villages
for serosurvey.
o Random effects model fitted to the
data suggested that adjusted intra-
herd correlation coefficient (ICC) was
0.35, while the ICC estimate between
animals from different herds was 0.15.
o Prediction from serological data
suggested that endemic transmissions
occur in many parts of the country
including where no outbreaks have
been observed.
Introduction
o Rift Valley fever (RVF) is an
arboviral zoonosis associated
with climate anomalies
o First RVF outbreak in Uganda
recorded in 2016 since 1968
o Multiple outbreaks at shorter
intervals since 2016; mainly in
southwestern and central parts
o This study compares spatial
distribution of epidemic and
endemic infections using
serological data to understand
the epidemiology of the disease
o Uganda is yet to adopt RVF
vaccination as a control strategy
Figure 1: Districts that have had RVF outbreaks in Uganda (A) and predicted
distribution of risk based on serological data (B). Variables used to predict this risk
included temperature, precipitation and altitude.
Mapping the risk of Rift Valley fever in Uganda
Dan Tumusiime, 1, 2 Bernard Bett,2 Simon Kihu, 2 Edna Mutua, 2 Rose Ademun 1
1 Ministry of Agriculture, Animal Industry and Fisheries
2 International Livestock Research Institute
Contact: drdantumusiime@gmail.com
Acknowledgments: This work was
funded by the Office of U.S. Foreign
Disaster Assistance, USAID.
The serosurvey was conducted by staff
in the Ministry of Agriculture, Animal
Industry and Fisheries, Uganda.
The U.S. Defence Threat Reduction
Agency supported the presentation of
this work at the virtual 6th World One
Health Congress 2020, Edinburgh.This document is licensed for use under the Creative Commons
Attribution 4.0 International Licence.
October 2020