Background: In contrast to most pastoral systems, the Somali livestock production system is oriented towards domestic trade and export with seasonal movement patterns of herds/flocks in search of water and pasture and towards export points. Data from a rinderpest survey and other data sources have been integrated to explore the topology of a contact network of cattle herds based on a spatial proximity criterion and other attributes related to cattle herd dynamics. The objective of the study is to integrate spatial mobility and other attributes with GIS and network approaches in order to develop a predictive spatial model of presence of rinderpest. Results: A spatial logistic regression model was fitted using data for 562 point locations. It includes three statistically significant continuous-scale variables that increase the risk of rinderpest: home range radius, herd density and clustering coefficient of the node of the network whose link was established if the sum of the home ranges of every pair of nodes was equal or greater than the shortest distance between the points. The sensitivity of the model is 85.1% and the specificity 84.6%, correctly classifying 84.7% of the observations. The spatial autocorrelation not accounted for by the model is negligible and visual assessment of a semivariogram of the residuals indicated that there was no undue amount of spatial autocorrelation. The predictive model was applied to a set of 6176 point locations covering the study area. Areas at high risk of having serological evidence of rinderpest are located mainly in the coastal districts of Lower and Middle Juba, the coastal area of Lower Shabele and in the regions of Middle Shabele and Bay. There are also isolated spots of high risk along the border with Kenya and the southern area of the border with Ethiopia. Conclusions: The identification of point locations and areas with high risk of presence of rinderpest and their spatial visualization as a risk map will be useful for informing the prioritization of disease surveillance and control activities for rinderpest in Somalia. The methodology applied here, involving spatial and network parameters, could also be applied to other diseases and/or species as part of a standardized approach for the design of risk-based surveillance activities in nomadic pastoral settings. Background Somalia's livestock production sector accounts for at least 40% of the gross domestic product (GDP) with 55% of the human population being directly involved in the rearing of livestock [1]. Pastoral movements across Somalia's borders for the purposes of grazing livestock and livestockrelated trade have occurred for centuries [2], resulting in seasonal movement patterns of herds/flocks in search of water and pasture. Moreover, in contrast to most pastoral systems, which are normally aimed at household subsistence, the Somali livestock production system is oriented towards domestic trade and export [3]. Livestock are shipped to various countries in the Arabian Peninsula, and trekked or transported to markets in Kenya, Djibouti, and Ethiopia [4]. In 2007 some 1,639,625 heads of cattle, sheep, goats and camels were exported through Bossasso * Correspondence: angortpel@yahoo.com 1 Veterinary Epidemiology & Public Health Group, Department of Veterinary Clinical Sciences, The Royal Veterinary College, University of London, Hawkshead Lane, North Mymms, Hatfield, Herts, AL9 7TA, UK Full list of author information is available at the end of the article BioMed Central © 2010 Ortiz-Pelaez et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Ortiz-Pelaez et al. BMC Veterinary Research 2010, 6:22 http://www.biomedcentral.com/1746-6148/6/22 Page 2 of 14 port and a total of 1,633,793 through Berbera port [4], the two major export markets in the country. Despite the collapse of the Somali government in 1991 and the lack of continued delivery of public services, an informal system of stateless order, social trust and an informal economy have allowed rural and urban populations to survive unfavourable economic and political circumstances [5]. But the Somali livestock industry is therefore now even more vulnerable to the introduction of bans by importing countries for two reasons. One being the stringent measures for livestock trade specified under the Sanitary and Phytosanitary (SPS) Agreement of the World Trade Organization http://www.wto.org, of which Somalia is not a member, and the other the poor standards of veterinary services and the absence of control measures for fighting trans-boundary animal diseases. The high mobility of the livestock population poses an additional challenge for the control and establishment of credible certification systems for the major transboundary diseases occurring in Somalia. An understanding of the aggregation/dispersion mechanisms and contact structure of the Somali livestock potentially could assist in setting up appropriate spatial risk-based surveillance activities and control measures that may lead to the establishment of an internationally accepted certification system for this nomadic pastoral livestock production system. Data on the geographical patterns of human and animal population distributions are now commonly considered in epidemiological studies. High density of susceptible populations has been shown to be a key factor in the transmission of infectious diseases in humans [6]. In the case of animal populations, spatial proximity is closely linked to the transmission of many infectious diseases [7]. The geo-referencing of animals kept in herds through the point location of the farm is relatively simple and costeffective [8]. However in settings where mobility is a major feature of the husbandry systems the geographical location of animals is no longer a discrete entity hence alternative approaches are required for the collection and analysis of their spatial data. Network data are not usually linked to geographical data, and social network analysis rarely considers the spatial configuration of the links [9] apart from the pure visualization of the networks. Definitions of links within networks based on spatial criteria such as the distance between pairs of nodes, usually estimated using the Euclidean distance ("as the crow flies") are rarely available in the literature. For example, Webb [10] made the assumption in her analysis that there was a link between two farms if their postcodes were less than 25 km apart, as a proxy for the catchment area of farm animal veterinary practices. Dent et al. [11] considered poultry premises to be linked if they were within 3 km distance from each other. The purpose of epidemiological modelling of spatial data is to explain or predict the occurrence of disease [1218]. The production of risk maps can guide decision makers provided that the underlying assumptions of uncertainty and variability are exposed [19]. Using the data from a previous rinderpest survey conducted in the Central and Southern regions of Somalia, it was possible to explore the topology of a contact network of cattle herds based on a spatial proximity criterion and other attributes related to cattle herd dynamics. The objective of the current study is to integrate spatial mobility and other attributes with GIS and network approaches in order to investigate their effect on disease presence and develop a predictive spatial model of presence of rinderpest. The risk map generated by the predictive model can be used to inform the design of risk-based surveillance activities of rinderpest by identifying high-risk point locations where to prioritize disease surveillance and control activities. Surveillance efforts could be targeted at high-risk areas which would lead to more effective use of the scarce veterinary resources in Somalia, and ultimately, contribute to the establishment of an internationally credible certification system for livestock export by nomadic pastoral systems. The standardisation of the proposed methodology for cattle will serve as a model for other livestock species (e.g. small ruminants and camelids) and other diseases. Methods Rinderpest data A cross-sectional survey was carried out in Central and Southern Somalia to estimate the prevalence of rinderpest (RP) in 2002-2003. The study area covered ten administrative regions of Central and Southern Somalia: Mudug, Galgadud, Hiran, Middle Shabele, Lower Shabele, Bay, Bakool, Lower Juba, Middle Juba and Gedo (Figure 1). Ninety percent of Somalia's cattle population are kept in these regions [20]. Details of the design and implementation of the survey are described elsewhere [21]. At selected sampling sites, a minimum number of 15 eligible animals were sampled per herd and questionnaires were administered to the livestock owners of the sampled herds in order to collect data on species, herd/ flock size and location during each climatic season over the two years prior to the date the survey was conducted in that point and livestock markets (primary and secondary) where the pastoralists sold their animals during the same period. A total of 9216 serum samples were collected from cattle aged 1 to 3 years at 562 sampling sites. In addition, 1071 sera were collected from cattle older Ortiz-Pelaez et al. BMC Veterinary Research 2010, 6:22 http://www.biomedcentral.com/1746-6148/6/22 Page 3 of 14 than 3 years at 58 sampling sites. All serum samples collected were tested for the presence of RP antibodies using a RP Competitive Enzyme Linked Immunosorbent Assay (C-ELISA) directed against the H protein of the virus [22]. All samples having a percentage of inhibition (PI) above 50% were considered positive in accordance with the OIE recommendation for testing large
[1]
M. A. Oliver,et al.
Kriging the Local Risk of a Rare Disease from a Register of Diagnoses
,
2010
.
[2]
D. Rogers,et al.
Spatial Analysis in Epidemiology
,
2008
.
[3]
M. Gilbert,et al.
Mapping H5N1 highly pathogenic avian influenza risk in Southeast Asia
,
2008,
Proceedings of the National Academy of Sciences.
[4]
P. Durr,et al.
How best to geo-reference farms? A case study from Cornwall, England.
,
2002,
Preventive veterinary medicine.
[5]
H. Croze,et al.
Geographical information systems for studying the epidemiology of cattle diseases caused by Theileria parva
,
1990,
Veterinary Record.
[6]
M. Puotinen,et al.
A geographic information systems model for mapping risk of fasciolosis in cattle and buffaloes in Cambodia.
,
2004,
Veterinary parasitology.
[7]
D. Pfeiffer,et al.
An analysis of the spatial and temporal patterns of highly pathogenic avian influenza occurrence in Vietnam using national surveillance data.
,
2007,
Veterinary journal.
[8]
Jennifer E. Dent,et al.
Contact structures in the poultry industry in Great Britain: Exploring transmission routes for a potential avian influenza virus epidemic
,
2008,
BMC veterinary research.
[9]
A. M. Abdullahi.
Pastoral production systems in Africa: a study of nomadic household economy and livestock marketing in Central Somalia.
,
1990
.
[10]
G. Glass,et al.
The geography of sexual partnerships in Baltimore: applications of core theory dynamics using a geographic information system.
,
1999,
Sexually transmitted diseases.
[11]
Roger Woodard,et al.
Interpolation of Spatial Data: Some Theory for Kriging
,
1999,
Technometrics.
[12]
Stephen J. Walsh,et al.
Spatial arrangement of social and economic networks among villages in Nang Rong District, Thailand
,
2000,
Soc. Networks.
[13]
R. Webster,et al.
Kriging: a method of interpolation for geographical information systems
,
1990,
Int. J. Geogr. Inf. Sci..
[14]
F. Carrat,et al.
Epidemiologic mapping using the "kriging" method: application to an influenza-like illness epidemic in France.
,
1992,
American journal of epidemiology.
[15]
A. Omore,et al.
Improvement and diversification of Somali livestock trade and marketing: Towards a formalized grading system for export quality livestock in Somalia
,
2012
.
[16]
Luc Anselin,et al.
Specification tests on the structure of interaction in spatial econometric models
,
1984
.
[17]
J. Hardin,et al.
Generalized Linear Models and Extensions
,
2001
.
[18]
D. U. Pfeiffer.
Geographical information science and spatial analysis in animal health.
,
2004
.
[19]
D. Rogers,et al.
Prediction of areas around the Mediterranean at risk of bluetongue by modelling the distribution of its vector using satellite imaging
,
2001,
Veterinary Record.
[20]
Peter D. Little,et al.
Somalia: Economy without State
,
2003
.
[21]
M. D. de Jong,et al.
Local spread of classical swine fever upon virus introduction into The Netherlands: Mapping of areas at high risk
,
2008,
BMC veterinary research.
[22]
Dirk U. Pfeiffer,et al.
A spatially predictive logistic regression model for occurrence of theileriosis outbreaks in Zimbabwe
,
1997
.
[23]
C. Webb,et al.
Farm animal networks: unraveling the contact structure of the British sheep population.
,
2005,
Preventive veterinary medicine.
[24]
Roger Th. A. J. Leenders,et al.
Modeling social influence through network autocorrelation: constructing the weight matrix
,
2002,
Soc. Networks.
[25]
Gary C. White,et al.
Analysis of Wildlife Radio-Tracking Data
,
1990
.
[26]
Anna Lindley,et al.
Somalia Country Study
,
2005
.
[27]
R. S. Morris,et al.
Application of Remote Sensing to Enhance the Control of Wildlife-Associated Mycobacterium bovis Infection
,
2002
.
[28]
Richard Rothenberg,et al.
Social and Geographic Distance in HIV Risk
,
2004,
Sexually transmitted diseases.