Integrating transportation and economic information for the African Swine Fever Virus (ASFV) risk modelling purposes in the Greater Mekong Sub-region (GMS) using online data and grey literature

• ¹ Center for Applied One Health Research and Policy Advice, College of Veterinary Medicine and Life Sciences, City University of Hong Kong, Hong Kong, SAR China

Introduction: African Swine Fever (ASF) is an important disease in pig production as it causes significant morbidity and mortality in affected herds. In August 2018, China reported its first case of ASFV, followed by Vietnam, Cambodia, Mongolia and Hong Kong in 2019[1]. As of May 2019, Laos, Myanmar, Thailand have no reported cases of ASFV. The proliferation of trade and transportation networks across borders via land, water and air routes have substantially contributed to the dissemination of (re)emerging human and animal pathogens such as ASFV across the globe.[2] Our study focuses on the Greater Mekong Sub-region (GMS), which includes the ASFV-affected countries Vietnam, Cambodia and the relevant provinces in China (Yunnan, Guangxi Zhuang Autonomous region). Pathogens of veterinary public health importance, such as foot and mouth disease (FMD), classical swine fever (CSF) and ASF can potentially cause heavy economic losses to these countries. The GMS countries are lower and/or middle income countries (LMICs), with Thailand and China belonging to the upper middle income group.[3] The risk of ASFV transmission is increased by low biosecurity pork production systems, limited veterinary public health infrastructure and capacities of LMICs in the GMS. Transmission of ASFV may also be enhanced by the interconnectedness of the GMS via trade and economic corridors. These include the Association of the Southeast Asian Nations (ASEAN) Highway Network (AHN) and the Trans-Asian Railway (TAR), which are collaborative projects involving the United Nations Economic and Social Commission for Asia and the Pacific (ESCAP).[4] For effective spatial risk modelling of ASFV, data of the complex interacting regional transportation networks via air, sea and land for trade, short- and long-term human migration are required. Integration of transportation data has been performed in the European Union (EU) and used in ASFV risk assessments.[5, 6] Unlike the EU, data for the GMS region are sparse and limited. This study will demonstrate how economic and transportation data from a variety of sources can be merged to create a dataset suitable for ASF spatial risk modelling and to perform preliminary analyses. Methods: An online search of grey and peer reviewed literature regarding the economic corridors and transportation routes across the GMS was performed using Google Search and Google Scholar. The administrative level 1 and transportation spatial data were downloaded from Open Street Maps [7] and the Open Development Mekong (ODM) website (https://opendevelopmentmekong.net/). Administrative level 1 areas are classified as provinces by Cambodia, Vietnam, Thailand and the Lao People’s Democratic Republic (Laos). For Myanmar, administrative level 1 areas are classified either as a State, Region or Union Territory. Port and airport locations were verified via the Google Maps application. The eight transportation variables of interest in each area are: • international airports • regional/domestic airports • river (Mekong and non-Mekong) ports • Deepwater ports (international) • seaports (international/regional) • AHN road networks; • TAR rail infrastructure. The above variables were coded as 1 or 0 (1= presence of infrastructure, 0= no presence). A transportation index is the calculated score for the transportation variables only and the maximum index score is 8. To add the ASFV-related variables, the information on ASFV reported cases were downloaded on 29 May 2019 from Food and Agriculture Organization of the United Nations (FAO) EMPRES-I Global Animal Disease Information System [8] and cross referenced with the World Organization for Animal Health’s (OIE) World Animal Health Information Database (WAHID) Interface.[9] The transportation connections and adjacency to ASFV-affected administrative level 1 areas were binary coded as 1 (yes) or 0 (no). To generate the preliminary risk score, the values of the ASFV-related variables were added to the transportation index score, with double weighting given to those connected to ASFV-affected province/s via international ports, international/regional ports, cross-border trans-shipping or trans-loading routes, and/or airports. Trans-shipping refers to transfer of cargo from ship to ship at a port, while trans-loading refers to transfer of cargo from ship to either rail freight or road vehicles. The above data were stored in Microsoft Excel. For the spatial mapping, QGIS 3.4 was used to generate two maps. The first map shows the transportation index in each area. The second map shows the score which includes ASFV presence or a connection to an ASFV affected area. Results: Data was integrated for each of the 198 administrative level 1 areas across five GMS countries (Vietnam, Cambodia, Laos, Thailand and Burma). A total of 19 data sources were found from open data websites (6), intergovernmental agencies (6), government agencies (all rights reserved/grey literature (4) and commercial/unknown/all rights reserved (3)). Further details on the data sources are listed in Table 1. Hai Phong City in Vietnam had the highest transport index score of 7. Those which scored 6 included Kampong Chhnang, Kampot (Cambodia); Mon (Myanmar); Bangkok, Nakhon Si Thammarat, Rayong, Songkhla,(Thailand); as well as Da Nang, Hanoi, Khanh Hoa, and Quang Ninh (Vietnam) (Figure 1). They represent the provinces with the highest scores in the GMS countries. The map is shown in Figure 2. The preliminary risk scores were between 0 to 20. The only area which scored 0 was Naypyitaw in Myanmar. For the ASFV-affected provinces in the GMS countries, the minimum preliminary score was 5. Vietnamese provinces of Cao Bang, Dak Nong, Ha Giang had a minimum preliminary risk score of 5 each, followed by the other four Vietnamese provinces (Hau Giang, Quang Tri, Vinh Long and Yen Bai) with a preliminary risk score of 6. A total of 77 administrative level 1 areas across the five countries had a preliminary risk score of 7 or more (Figure 3). In Myanmar, Shan, Mon and Yangon each scored 8. For Cambodia, a total of 6 provinces with Kratie and ASFV-affected Ratanakiri scoring the highest at 11. For Thailand, there were a total of 136 provinces with a preliminary score of 7 or more. Bangkok scored the highest at 12, followed by Chonburi, Nakhon Si Thammarat and Rayong at 9. In Laos, 11 provinces scored 10 or above. Oudomxay in Laos had the highest score of 15. For Vietnam, 31 of the 39 provinces which scored 7 or above were ASFV affected areas. Hanoi in Vietnam had the highest preliminary risk score of 20, followed by Quang Ninh and Hai Phong with the scores of 16 each. All three areas have reported ASFV. For those which scored 15, Da Nang in Vietnam was not affected by ASFV as of May 2019 but Ha Tinh has reported ASFV. The map with the preliminary risk scores is shown in Figure 4. Discussion and conclusion: The effective integration of transportation and economic data in the spatial dataset has been demonstrated without the use of commercially available Big Data sources. This constitutes the first step of building a risk model to assess the risk of ASFV introduction associated with transport network connectivity for GMS countries. In this preliminary analysis, a simple binary coding was used to indicate presence of the transportation infrastructure via land, sea/river, and air. More complex variable weighting rules will be developed for the next phase of model development. There are limitations to the data as it only provides information on the transportation infrastructure’s presence and the land/sea/air connections to ASFV-affected provinces. Directional movement as well as the volume of movement are not included. Further work includes adding more detailed information (i.e. size based on number of port terminals, economic activity, human migration, and airport connections) to the dataset. Data on the presence of wild boar habitats, pig density and other agricultural variables may be added. The information in this study can be used for other medical and veterinary public health emergencies to identify areas at increased risk of disease incursion. This preliminary analysis can already be used as an additional information source by LMICs for prioritizing ASFV surveillance and optimizing resource utilization for risk reduction by identifying potential high risk hotspots. Further work is in progress for more detailed spatial risk modelling and mapping using the data presented here.