Simulation of control scenarios of porcine reproductive and respiratory syndrome in Nghe An Province in Vietnam

Abstract The main objective of this study was to develop various models using North American Animal Disease Spread Model (NAADSM) to simulate the transmission of Porcine reproductive and respiratory syndrome (PRRS) virus between farms in Nghe An Province in Vietnam in order to inform the prevention and control of this important disease. Using real data from the household survey, credible parameters for direct/indirect mean contact rates between different farms were estimated. A total of eleven models were developed, including immunization scenarios. In addition, we conducted sensitive analysis on how the mean contact rates influenced the results. The immunization scenarios showed that a high proportion of pigs in medium size farms needs to be vaccinated in order to reduce the transmission to pigs in small farms under the Vietnamese pig production system. In order to promote the use of vaccinations, incentives (such as a vaccine subsidy) for medium size farms may be needed. It could be the most cost‐effective control and prevention strategy for pig diseases in Vietnam. Our study provides insights on how pig diseases can be spread between pig farms via direct and indirect contact in Nghe An under the various hypothetical scenarios. Our results suggest that medium/large farms may play an important role in the transmission of pig diseases.


| INTRODUC TI ON
Simulation models for evaluating the spread of contagious animal diseases are an important decision supporting tool for disease control (Keeling, 2005;Keeling et al., 2001;Morris, Sanson, Stern, Stevenson, & Wilesmith, 2002;Woolhouse, 2003), as demonstrated for example, during the Foot and mouth disease (FMD) outbreak in 2001 in the UK (Keeling et al., 2001;Taylor, 2003). Various modelling platforms have been developed, such as the North American Animal Disease Spread Model (NAADSM) a stochastic, spatial, farm-level state-transition modelling framework that was developed to simulate the spread of highly contagious diseases in animals (Harvey et al., 2007). The user-established parameters define the disease transmission between farms, determined mainly by rates of direct contact, indirect contact and distances between farms.
In Vietnam, pork accounts for about 70% of all livestock products (Lich, 2001), with majority of pork produced by small-scale farmers (GSO, 2016). Medium size pig farms (accounting for 20%-25%) are the main suppliers for piglets and weaners to small farms. The linkages between these different production systems and the large number of actors involved in pork production have implications for disease transmission. Intense contact between different farm movements of live animals between farms are considered the main route of disease transmission, which may lead to continuous emergence of epidemics (Gilbert et al., 2005).
To better understand the influence of contact patterns between farms on disease transmission and how this affects disease control, a simulation model was developed using porcine reproductive and respiratory syndrome (PRRS) as the disease example. In Vietnam, PRRS is endemic and has a major impact on production. It is a viral disease in pigs caused by a single-stranded and small enveloped RNA virus of the family Arteriviridae, in the order Nidovirales (Dea, Gagnon, Mardassi, Pirzadeh, & Rogan, 2000;Meulenberg et al., 1993;Thiel et al., 1993). Genotyping studies identified two types, distinguished as the European and the North American genotype (Murtaugh, Elam, & Kakach, 1995;Nelsen, Murtaugh, & Faaberg, 1999). The disease causes reproduction disorders (such as abortions or stillbirth), respiratory disease, slow growth rates, lethargy and anorexia in all age groups (OIE, 2008;Rossow, 1998;Stadejek et al., 2002;Wensvoort, 1993;. In Vietnam, the first outbreak was reported in the late 1990s (Nguyen, Vuong, & Vo, 2015).
Since then, the PRRS virus has spread quickly across the country, contributing to serious economic losses in the pig production sector.
The main objective of this study was to develop various models using NAADSM to simulate the transmission of the PRRS virus between farms in Nghe An Province in Vietnam in order to inform the prevention and control of this important disease.

| Study location and population
Nghe An is located in the north central coast region of Vietnam and is the largest province by area in the country, with an estimated human population of 3.1 million (GSO, 2018). In order to simulate the spread of the PPRS virus between farms in NAADSM, the geographical locations and farm characteristics (such as herd size and production type) are required. Information on the number of pig farms with their herd size at the district level was obtained from local authority, though exact data on actual geographical locations of farms was not available. Random points corresponding to the number of farms per district were therefore created using QGIS (Quantum GIS development team 2012. QGIS version number 3.0.1), while longitude and latitude coordinates of these points were extracted as farm locations and then introduced into NAADSM (Figure 1).
Local data indicated a total of 232 farms in Nghe An (Table 1).
Data on the number of farms were available for 18 out of 21 districts. These farms were classified into three production types: small <100 pigs; medium ≥100 and <1,000 pigs; and large farms ≥1,000 pigs (Nga, Ninh, Hung, & Lapar, 2014). Based on our criteria, the proportion of small, medium and large farms was 23.18%, 68.24% and 8.15%, respectively.  with other farms. Farm contact information was used to estimate the "mean contact rates" for "direct contact" and "indirect contact".

| Model parametrization
The former is defined as the introduction of an infected animal from one farm to another. The latter includes the movement of people, vehicles, materials and equipment between farms.
For direct contact, data were collected on how often farmers introduced pigs on their farms and if and how often they share boars for breeding purposes. To gather information on indirect contact which potentially could transmit the PRRS virus between farms without pig movement, four questions were asked: (1) "How often did vehicles enter your farm over the last 6 months?"; (2) "How often did veterinarians/ animal health workers visit your farm over the last 6 months?; (3) How often did other farmers and traders visit your farms over the last 6 months?"; and (4) "How often did you share any equipment with other farms". A Poisson distribution was fitted for the different model parameters, with the average number of contacts used to define λ. The distributions for mean direct and indirect contact rates were computed on a weekly basis. This period was selected as the virus can be infective for a week at 21°C (Benfield et al., 1992).
The estimated probabilities of infection transfer of the PRRS virus by direct contact was 1, while a value of 0.1 was assigned for indirect contact as these values were used in previous studies (Neumann, Morris, & Sujau, 2007;Thakur, Revie, Hurnik, Poljak, & Sanchez, 2015). No data were available to parameterize the contact distance between farms. Based on our experience, we used the BetaPERT distribution which is defined by its minimum (0.2 km), its most likely value (10 km), and its maximum (100 km) (Table 1).
We assumed that all farms were free of PRRS at the beginning of scenarios, and none of the pigs had any immunity to a new strain of virus. Furthermore, we assumed that once a single pig became infected then the whole farm was considered infectious within a week. Each farm had an equal chance to be in contact with other farms given the distance between source and recipient farms and combinations of predefined production types. Other types of indirect contact (such as airborne and fomite) were not considered. In addition, we assumed that a continuous flow (CF) system was used in all farms in our models instead of all-in-all-out system (AIAO) which is not common in Vietnam.

| Model structure and outcome
In Vietnam, medium farms are mainly responsible for supplying piglets and weaners to small farms ( Figure 2). However, there is almost zero animal movement "from small to medium", "from small/medium farms to large farms" and between large farms. Therefore, indirect contact was only considered "from small to medium" and "from medium/large to large" farms (Table 2). We modelled 11 scenarios -scenario A1 to A3 assumed a completely naïve population: scenario A1 assumed that the PRRS virus was transmitted by direct contact only, scenario A2 assumed both direct and indirect contact and scenario A3 assumed both direct and indirect contact without movement to large farms. For A1-3 scenarios, one medium farm was randomly selected to be seeded with an infection and the same farm-initiated infection in the subsequent iterations. The remaining farms were susceptible at the beginning of the scenarios and then became infectious until the end of the simulation. In order to evaluate the impact of an initial outbreak by production type, we generated two scenarios B1 and B2 where an initial outbreak started in a randomly selected single small (B1) or large farm (B2) via both direct and indirect contacts. Scenarios C were endemic scenarios where different proportions of medium size farms were naturally immune by 30% (scenario C1), 20% (scenario C2) and 10% (scenarios C3) following previous PRRS outbreaks. Scenarios D assumed different vaccination coverages for medium size farms: scenario D1 vaccinated by 100%; scenario D2 75%; and scenario D3 50% as those were the main pig suppliers for small farms (accounting for 70% of pig production in Vietnam). It was assumed that natural immunity and vaccine Indirect contact (for all production types)  (Nodelijk et al., 2000;Nodelijk, Nielen, Jong, & Verheijden, 2003). Therefore, it was assumed that these farms followed the susceptible-infectious (S-I) transition structure, with no potential for recovery (Keeling & Rohani, 2011).

| Sensitivity analysis
The direct contact transmission probability ( Table S1).

| RE SULTS
Of the farms enrolled in the survey, none reported that boars or equipment were shared with other farms.   Table S2). However, mean indirect contact of 0.5 showed a large reduction in the number of median infected farms (−48.80%; 209->107). Of note, was a sharp decrease in the median number of infected large farms from 19 to 3. movement from medium to small farms. Therefore, no animal movement from small to medium farms was considered as the small farms are the last stage of pig's life cycle in Vietnam, which are brought to the slaughterhouses. Our survey showed that the indirect contact rate in large farms was considerably higher than small and medium farms, likely reflecting the fact that there are frequent vehicle and human movements for pig sales and farm managements in comparison with small and medium scale farms.

| D ISCUSS I ON
Our models showed that the transmission route to large farms via indirect contact could have a significant impact on our results.
Given the strict restrictions in place, however, it is unlikely that the large farms allow the sharing of vehicles and human movement with other farms. Therefore, the medium farms need to be targeted to efficiently reduce/prevent the transmission of the PRRS virus to small-scale farms, which account for 70% of the total production in Vietnam (Lapar, Binh, & Ehui, 2003). Immunization scenarios showed that a high proportion of medium size farms should be vaccinated in order to reduce the transmission to small farms under the Vietnamese pig production system. In general, medium size farmers are relatively wealthier and more likely to invest in vaccines and biosecurity rather than small farmers. In order to promote update of vaccinations, however, incentives (such as a vaccine subsidy) for medium size farmers may be needed. It could be the most cost-effective control and prevention strategy for pig diseases in Vietnam.
The model assumed that the whole farm became infectious when one pig was infected from the farm, which was reasonable as it is unlikely that the PRRS virus would die out without onward transmission of diseases to other pigs in the farms. Similarly, many studies have suggested that R 0 of PRRS virus was higher than 1 (Charpin et al., 2012;Nodelijk et al., 2000;Zhang, Kono, & Kubota, 2014). Some studies have suggested that PRRS virus has a high degree of genetic and antigenic variability (Chang et al., 2002;Tian et al., 2007), which may result in different levels of cross-protection of different morbidities. However, we assumed that there was no emergence of a new strain or multiple strains circulating during the modelled study period. Simulation models are approximate imitations of real-world, so it is necessary to validate how our models represent the dynamic of the disease in a population. However, we were not able to validate our models due to lack of real outbreak data in Nghe An.
The main limitation of our study was that our direct contact transmission probability was considered to be 1 which may have led to an over-estimation as some of farms may use AIAO production system, so virus transmission rate might be less than 1. The importance of the transmission probability was also confirmed in the sensitivity analysis. Values used were based on studies by others (Neumann et al., 2007;Thakur et al., 2015), but may need to be adjusted for Vietnam as more evidence becomes available.
This study is the first attempt to quantify the indirect contact   . It may be possible that the virus can be transmitted from poultry to pigs as mixed livestock farming systems are very common in small farms in Vietnam. These potential risk factors could therefore be added as the indirect contact. Lastly, we did not take into account of the prevention and control strategies in the current models, which were likely to have overestimated the disease transmission between farms. One study suggested that PRRS transmission was significantly associated with location of the farms, farm management, human and animal contact (Truong & Gummow, 2014).
A few simulation studies on the spread of PRRS virus have been conducted in New Zealand and Canada using similar software (Neumann et al., 2007;Thakur et al., 2015). The mean direct contact rates used in New Zealand (0.064-0.21/week) and Canada (0.01-0.51/week) were slightly higher than our study (0.072-0.073/week).
However, indirect contact rate in our study was considerably higher

ACK N OWLED G EM ENTS
The authors thank the National Institute of Veterinary Research (NIVR), Ministry of Agriculture and Rural Development (MARD). The authors also thank the staff of sub-department of animal health from Nghe An Province for the collaboration with research team in conducting this study. This study was funded by the CGIAR Research Program on Livestock and is supported by contributors to the CGIAR Trust Fund. CGIAR is a global research partnership for a food-secure future. Its science is carried out by 15 Research Centers in close collaboration with hundreds of partners across the globe. www.cgiar.
org We also would like to give special thanks to Mireille Ferrari for English editing.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.