The integration of geostatistical analysis with social network improve active disease surveillance
-
1
North Carolina State University, United States
-
2
Department of Population Health and Pathobiology, College of Veterinary Medicine, North Carolina State University, United States
-
3
Department of Preventive Veterinary Medicine and Animal Health, Faculty of Veterinary Medicine and Animal Science, University of Sao Paulo, Brazil
Background and specific objectives
Trade of infected, untested or false-negative animals have been associated major disease spread events (VanderWaal et al., 2018; Machado et al., 2019), which can either be at short- and long-distances (Firestone et al., 2012). Therefore, understanding of how farms are connected through animal trade is critical to 1) identify high-risk movements; 2) predict disease risk; and 3) design improved surveillance and control strategies (Guinat et al., 2016; Cárdenas et al., 2018). For most infectious diseases, where animal movement is regulated, tested negative animals are allowed to move between farms with the same disease statuses, but in most cases there is not a restriction in moving animals into areas of greater infectious risk, consequently, the use of animal movement alone aimed at disease control often neglect the underlying disease risk.
In order to develop a more effective disease response plans, the combination of a comprehensive analysis of contact network integrated with model-based geostatistical methods (Diggle et al., 1998) have the potential to uncover unknown risks which may improve active surveillance activities. The approach proposed here is especially useful where monitoring of the disease burden is carried out through passive surveillance (Jones et al., 2019). The objective of this study is to describe the complete movement patterns between horse farms in one Brazilian state to identify high-risk farms, improve Equine Infectious Anemia (EIA) control strategies by the integration of geostatistical analysis to map EIA movements between risk areas. Furthermore, we aimed to quantify susceptible movements, especially those from infected to non-infected areas, ultimately guide interventions and improve EIA surveillance and control.
Methods
Descriptive analysis
Horse movement data was obtained from the State of Rio Grande do Sul, Brazil. Altogether 14.588.68 movements, between 38.263 farms and 495 municipalities were collected. The data included farm locations, farm of origin and destination, date of movement, movement types (e.g., reproduction, sport), number of animals moved and all EIA outbreaks from 2015 until 2017. Before any movement data was analyzed, a completeness analysis removed data any incomplete information. We constructed the static contact network at municipality level for the full three years (2015 to 2017). Furthermore, the animal flows between municipalities were calculated, which included the number of movements and total of animals traded.
Geostatistical analysis and animal flow
A case-control design was used to map the EIA spatial risk. All farms with at least one movement during the study period were eligible. Horse farms with at least one EIA positive animal were case, in total 158. A case was matched to a control (one case::five controls), control farms were randomly based on municipality (Fig. 1). The final case-control locations were modelled via Bayesian binomial logistic regression with spatiotemporal random effects to account for spatial latent process between locations, which may vary over time (years). The spatiotemporal random effects were modelled using stochastic partial differential equations, which represents a Matérn spatio-temporal Gaussian field as a Gaussian Markov random field via triangulation.
Results
Among the main problems found in the data we highlight same “id” for farm of origin and destination (n = 1.318.518 movements), missing or error in the geolocation (n= 533.333 movements), or missing dates, farm of origin, farm of destination, and geolocation (n =534.317 movements). In Figure 2-A, show the municipalities’ absolute number movements with at least one of the above completeness problem. Figure. 2-B, the proportion of records removed over the total number of movements. Total movement data considered usable, was 1.235.383 movements, among 495 municipalities, in total 1.877.215 horses traded. Areas of elevated EIA risks were identified mainly in West and in East-central areas of the study area (Fig. 3-A). We identify a substantial amount of movement between areas of elevated risk, which correlates with the region’s endemic characteristics (Fig. 3-B). More importantly the majority of the heavier flows between municipalities were direct from infected into uninfected areas, therefore it indicates an exceed risk for new infections reaching uninfected regions.
Conclusions
The analysis of networks in conjunction with the Bayesian hierarchical approach captured the spatial trends for EIA distribution and at what magnitude flows between risk and non-risk areas may continue the spread of EIA. The municipalities in this region were intensely connected; therefore, the use of the network analysis allowed identifying key municipalities which play an important role in the containment and or propagation of any infectious disease. The network identified key regions with higher chances for EIA spread or more vulnerable to new infection, while in some areas the spatial risks were also higher; therefore, those common areas would benefit the most for active EIA surveillance while the main goal is to identify untested animals or reducing overall areas risks.
The authors declare that there are no conflict of interests
Acknowledgements
This project was funded by the Department of Population Health and Pathobiology, College of Veterinary Medicine, North Carolina State University, provided startup funds for GM
References
Cárdenas, N.C., J.O.A. Galvis, A.A. Farinati, J.H.H. Grisi-Filho, G.N. Diehl, and G. Machado, 2018: Burkholderia mallei: the dynamics of networks and disease transmission. Transbound. Emerg. Dis. 0, DOI: 10.1111/tbed.13071.
Diggle, P.J., P.J. Ribeiro, and O.F. Christensen, 2003: An Introduction to Model-Based Geostatistics,.
Firestone, S.M., R.M. Christley, M.P. Ward, and N.K. Dhand, 2012: Adding the spatial dimension to the social network analysis of an epidemic: Investigation of the 2007 outbreak of equine influenza in Australia. Prev. Vet. Med.DOI: 10.1016/j.prevetmed.2012.01.020.
Guinat, C., A. Relun, B. Wall, A. Morris, L. Dixon, and D.U. Pfeiffer, 2016: Exploring pig trade patterns to inform the design of risk-based disease surveillance and control strategies. Sci. Rep.DOI: 10.1038/srep28429.
Jones, A.E., J. Turner, C. Caminade, A.E. Heath, M. Wardeh, G. Kluiters, P.J. Diggle, A.P. Morse, and M. Baylis, 2019: Bluetongue risk under future climates. Nat. Clim. Chang.DOI: 10.1038/s41558-018-0376-6.
Machado, G., C. Vilalta, M. Recamonde-Mendoza, C. Corzo, M. Torremorell, A. Perez, and K. VanderWaal, 2019: Identifying outbreaks of Porcine Epidemic Diarrhea virus through animal movements and spatial neighborhoods. Sci. Rep. in press, DOI: DOI:10.1038/s41598-018-36934-8.
VanderWaal, K., A. Perez, M. Torremorrell, R.M. Morrison, and M. Craft, 2018: Role of animal movement and indirect contact among farms in transmission of porcine epidemic diarrhea virus. Epidemics 24, 67–75, DOI: 10.1016/j.epidem.2018.04.001.
Keywords:
disease spread,
Data completeness,
Risk areas mapping,
Movement flows,
Non-stationary Gaussian models
Conference:
GeoVet 2019. Novel spatio-temporal approaches in the era of Big Data, Davis, United States, 8 Oct - 10 Oct, 2019.
Presentation Type:
Senior oral presentation
Topic:
Spatial-explicit or spatio-temporal network analysis
Citation:
Machado
G,
Ardila Galvis
JO,
Grisi-Filho
JH and
Cárdenas
NC
(2019). The integration of geostatistical analysis with social network improve active disease surveillance.
Front. Vet. Sci.
Conference Abstract:
GeoVet 2019. Novel spatio-temporal approaches in the era of Big Data.
doi: 10.3389/conf.fvets.2019.05.00075
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
01 Jun 2019;
Published Online:
27 Sep 2019.
*
Correspondence:
Prof. Gustavo Machado, North Carolina State University, Raleigh, United States, gmachad@ncsu.edu