Presentation by Joseph Nkamwesiga, Fedor Korennoy, Paul Lumu, Peninah Nsamba,
Frank Nobert Mwiine, Kristina Roesel, Barbara Wieland, Andres Perez, Henry Kiara and Dennis Muhanguzi at the 16th International Symposium of Veterinary Epidemiology and Economics, Halifax, Canada, 12 August 2022.
Spatio-temporal analysis of peste des petits ruminants in Uganda
1. 1
Spatio-temporal analysis of Peste des Petits Ruminants
(PPR) in Uganda
Joseph Nkamwesiga1,2, Fedor Korennoy3, Paul Lumu4, Peninah Nsamba5, Frank Nobert Mwiine5,
Kristina Roesel1, Barbara Wieland6,7, Andres Perez8, Henry Kiara1, Dennis Muhanguzi5
1International Livestock Research Institute
2Freie Universität Berlin
3Federal Center for Animal Health, Vladimir, Russia
4Ministry of Agriculture Animal Industry and Fisheries, Uganda
5Makerere University
6Institute of Virology and Immunology, Mittelhäusern, Switzerland
7University of Bern
8University of Minnesota
22nd International Symposium of Veterinary Epidemiology and Economics
Halifax, Canada, 12 August 2022
Spatio-temporal analysis of peste des petits ruminants in
Uganda
2. 2
Introduction
• Peste des petits ruminants (PPR) is a
disease of sheep and goats
• Caused by PPR virus, a negative
sense morbillivirus
• Associated with >90% morbidity
& over 80% mortality in naïve
flocks
Source:BUILDPPRoutbreakinvestigation
The disease
3. 3
Introduction
• PPR currently affects >70
countries in Africa & Asia
• With a prevalence of ~40%1
• Causes losses of about USD 1.5 – 2.1
billion per year globally
Geographical distribution
Source:Gaoetal./SmallRuminantResearch162(2018)12–16
1MdAhaduzzaman,2020/VetMedSci.2020Nov;6(4):813-833
4. 4
Introduction
• Spreads majorly through human socio-
economic activities:
• Management practices
– nomadism
– transhumance
– communal grazing
• Animal movement
– trade purposes
– social events
PPR transmission drivers
Source:Open-sourceinternetphotos
5. 5
Main Objective
To identify the high-level spatial conditions associated with places in which PPR tends to be
present using a purely spatial model
• and characterize those places in which the disease is frequent
6. 6
Materials and methods
Source:Gaoetal./SmallRuminantResearch162(2018)12–16
Requisite datasets
Variable description Source
1 Outbreak reports compiled from MAAIF (2007-2020) MAAIF (NADDEC)
2 Animal Movement data Movement permits (MAAIF)
3 Livestock population density (goat, sheep, pigs , cattle etc) ILRI/ FAO Livestock Geo-Wiki
4 Bioclimatic variables (temperature, precipitation etc) http://www.worldclim.org/bioclim
5 Land cover type Copernicus Global Land service
6 Road network (density, length) Uganda Bureau of Statistics (UBoS)
7 Human population density Uganda Bureau of Statistics (UBoS)
8 Wildlife Protected areas and wetlands Uganda Bureau of Statistics (UBoS)
9 Solar radiation and air temperature Copernicus Global Land service
10 Wind speed and water vapour pressure Copernicus Global Land service
12 Distance from major towns (proxy for markets, slaughterhouses) Uganda Bureau of Statistics (UBoS)
13 Soil water index Copernicus Global Land service
7. 7
Materials and methods
• Descriptive analyses
• Data summaries
• Regression models
• Logistic model
• Negative binomial model
Methods • Space–time analysis
• Space-time cube
• Getis-Ord Gi* statistics
• Emerging Hot Spot analysis
• Mann–Kendall statistics
Internet images: Wikimedia Commonscommons.wikimedia.org
8. 8
•A total of 221 passive reports recorded
(2007–2020)
•172 reports were confirmed as PPR
outbreaks based on ELISA and/or PCR test
•Covered about 40% (55/134) of districts in
Uganda
Results and discussion
Descriptive
9. 9
Results and discussion
Significance levels: ***p<0.001, ** p<0.01 and * p<0.05
Negative binomial regression model
PPR outbreak predictor Coefficient Standardized
Coefficient
Standard Error z value Pr(>|z|)
(Intercept) 0.572 1.135 0.504 0.61431
Annual rainfall -0.003 -0.283 0.001 -4.469 7.85e-06***
Digital elevation 0.001 0.106 0.000 1.899 0.05755
Road length 0.005 0.200 0.001 3.048 0.00230 **
Small ruminant density 0.005 0.107 0.002 2.134 0.03286 *
Soil Water Index 0.013 0.279 0.003 4.087 4.37e-05 ***
Protected area per district -0.001 -0.148 0.000 -2.189 0.02859 *
10. 10
Results and discussion
Significance levels: ** p<0.01 and * p<0.05
Logistic regression model
Variable Coefficient Adjusted coefficient Standard Error z value Pr (>|z|)
(Intercept) 6.408 2.514 2.549 0.01081 *
Annual rainfall -0.002 -0.852 0.001 -1.735 0.08269
Road length 0.007 1.231 0.003 2.591 0.00957 **
Small ruminant density 0.012 1.057 0.006 2.043 0.04103 *
Soil Water Index 0.011 1.055 0.005 2.091 0.03653 *
Median annual wind speed -3.609 -1.385 1.175 -3.071 0.00213 **
11. 11
• Down Trend
• Districts in the Karamoja subregion
• Up Trend
• Around the Lake Victoria crescent
(central Uganda) and southwestern
Uganda
• No obvious pattern
• West Nile region and around the Lake
Kyoga plains
Results and discussion
Hot spot trends
12. 12
• New hot spots
• Four districts (Masaka, Mubende,
Gomba & Rwampara)
• Consecutive hot spots
• Eight districts (Ibanda, Mbarara,
Lwengo, Lyantonde, Ssembabule,
Kiruhura, Isingiro and Kazo)
• Sporadic hot spots
• Rakai district
Results and discussion
Emerging Hot spot trends
13. 13
Conclusions
• Different hotspot trend categories were identified across Uganda
• High small ruminant density
• Longer road length
• Reduced annual rainfall
• High soil water index
• A basis for more robust timing and prioritization of control measures to
contribute to the global goal of control and eradication by 2030
Setting-characterisation
High-level spatial conditions associated with these settings
Potential implications of the findings
14. 14
Study limitations
• Data used based on clinical observations or outbreak reports
• Such reports constitute just a fraction of the true PPR incidence over the study period
• No precise PPR vaccination data for all the districts in Uganda over time
• However, vaccines were applied in those places in which disease was prevalent, so, in a purely
spatial model, that would come up as an association between vaccine and disease
• Key underlying assumption is that the parameters used serve as a proxy for true
value of the variables
• However, it is unclear whether those associations are influenced by other factors
15. 15
Acknowledgements
• Kristina Roesel and the #BuildUganda team
This work was funded by
Implemented in partnership with
Supervisors [Dr Henry Kiara, Univ.-Prof. Dr. Klaus Osterrieder & Dennis Muhanguzi]