Poultry Farm Vulnerability and Risk of Avian Influenza Re-Emergence in Thailand

Highly pathogenic avian influenza (HPAI) remains of concern as a major potential global threat. This article evaluates and discusses the level of vulnerability of medium and small-scale commercial poultry production systems in Thailand related to avian influenza virus re-emergence. We developed a survey on 173 farms in Nakhon Pathom province to identify the global level of vulnerability of farms, and to determine which type of farms appears to be more vulnerable. We used official regulations (the Good Agricultural Practices and Livestock Farm Standards regulations) as a reference to check whether these regulations are respected. The results show that numerous vulnerability factors subsist and could represent, in case of HPAI re-emergence, a significant risk for a large spread of the disease. Bio-security, farm management and agro-commercial practices are particularly significant on that matter: results show that these practices still need a thorough improvement on a majority of farms. Farms producing eggs (especially duck eggs) are more vulnerable than farms producing meat. Those results are consistent with the type of farms that were mostly affected during the 2004–2008 outbreaks in Thailand.

and LFS comprise eight main sections (for broiler and meat ducks farms) or nine (for layer farms): general condition of the farm, poultry feed, water quality, farm management, animal health, poultry welfare management, environmental management, data recording and productivity management for layers farms. Immediately after the first wave of HPAI outbreaks, in 2004, DLD was immensely concerned in increasing farms bio-security and preparing new policy measures [12]. For example, farmers were required to improve open poultry housing to a closed evaporative system. However these measures were not fully implemented, and still, in certified poultry farm standards, both (open and closed-evap) houses are allowed [13]. No study evaluating the actual level of vulnerability of poultry farms in Thailand using official GAP and LFS regulations was published recently. A study was performed in 2011 in three provinces, but focused on the main pathway of entry of influenza viruses onto farms with different production methods [14].
This article describes and discusses the actual level of vulnerability of medium and small-scale poultry commercial production system (sector 2 and 3) related to HPAI risk in Thailand. We use official GAP and LFS regulations as the references to identify the overall level of vulnerability of poultry farms, checking on farms whether these regulations are respected. The aim of this article is to evaluate the global level of vulnerability, to highlight the most common factors of vulnerability, and to determine which type of farms appears to be more vulnerable. Also, spatial analysis is performed in order to find any geographical pattern of vulnerability susceptible to increase the risk of HPAI farm-to-farm transmission.

Field Survey in Nakhon Pathom Province
We conducted a field survey supported by a comprehensive semi-structured questionnaire in the 173 registered commercial poultry farms in the Nakhon Pathom province, Thailand ( Figure 1) from March to September 2011. The Nakhon Pathom province, in the central region of Thailand, has been selected for our study because of its extensive agriculture activities and for logistic reasons. The province (2,168 km² ) consists of seven districts, 106 sub-districts and 870 villages, with nearly 1,000,000 inhabitants in 2011. Nakhon Pathom is nationally ranked 7th and 2nd for its broiler and duck meat production, 6th and 8th for chicken and duck layer farm (DLD, 2008). It has 212 registered poultry farms belonging to sector 1 to 3 (DLD, 2010). Farms belonging to sector 1 were not included in the study as they are considered to have limited vulnerability because they implement high level operating procedures to reduce the risk of viral incursions [6]. The sector 4 farms (backyard production with minimal bio-security and birds/products consumed locally) were also not included in this study as we considered that outdoor backyard production with minimal or no bio-security measures cannot be equally compared to farms [9,15].
The present survey included all 173 registered commercial poultry farms of the province belonging to sector 2 and sector 3 in 2011: 96 broiler farms, 33 farms of laying hens, 28 ducks meat farms and, 16 farms of laying ducks. The location of each farm entrance was recorded by GPS. An environmental description was conducted from the farm entrance. A questionnaire was used to describe agricultural practices from owners or workers; all interviews were conducted in Thai language. Questionnaires were addressed directly, without an appointment, to owners or workers, at the entrance of the farms.

HPAI in the Nakhon Pathom Province, 2004-2008
Forty one H5N1 HPAI laboratory confirmed outbreaks were reported in Nakhon Pathom from 2004 to 2005 ( Figure 2). The province has suffered strong economic effects, like the neighboring province of Suphanburi where H5N1 HPAI emerged in late 2003 [2]. Of the 173 farms from sector 2 and 3 of Nakhon Pathom inspected in 2011, 43 were established after 2004 ( Figure 3).

Vulnerability Factors, Vulnerability Scores, Statistical and Spatial Analysis
Vulnerability factors in the present study are limited to factors described in DLD, GAP and LFS regulations and are listed in Table 1. All factors are Boolean (yes/no answers). They were grouped into five categories, as per DLD regulations: Cat. 1: general condition of the farm, environmental and geographical description, environmental and farm condition, animal health and poultry welfare management; Cat. 2: agricultural and agro-commercial practice, slaughtering and selling bird on the farm, to who birds are sold, who organize transports, frequency of transport, "all-in-all-out" practice, restocking time; Cat. 3: bio-security measures and farm management, including transports, restocking practice, manure utilization, water management, rodent and insect control, frequency of disinfection, workers routine and hygiene practice, contact between farms; Cat. 4: agro-commercial practice of layers farms: egg transportation, disinfection of egg trays, percentages of sale direct to contractors, sale through aggregators; Cat. 5: knowledge of the local farmer's regarding HPAI and verification of their commitment to HPAI risk management; We assigned a weight to each vulnerability factor in order to calculate global vulnerability scores as additive model of vulnerability. Scores are then unique values affected to each farm and make vulnerability evaluation more synthetic. A vulnerability score for a farm is calculated as a linear weighted sum, by adding the weight of a factor when this factor is positive (yes answer) for the farm. The scores are scaled (divided by the sum of weights) to fit between 0 (lowest) and 1 (highest): (k) (k) The weights correspond to our appreciation of the relative level of importance between factors in term of vulnerability, as experts in this field. As always when using scores based on weights, we assume that our appreciation is subjective and that significant variation in the weighting may occurs if given by different experts.
For each farm, several scores were calculated:  A score for each category (management and environmental conditions, agricultural practices, bio-security measures, security measures related to egg production, knowledge of the farmer);  An overall score of vulnerability considering all variables but eliminating variables related to egg production in layer and meat farms;  An overall score of vulnerability considering all variables including variables related to egg production. Although variables associated with the eggs production are vulnerability factors for the farms producing eggs and are not relevant for farms producing birds, they were included in the calculation of the overall indicator of vulnerability: egg production increases the potential vulnerability with more potentially "at risk" practices and behaviors.
Statistical results by factor were calculated (proportion of positive answers among farms and standard deviation). Multivariate analysis was performed to find relationships between factors (correlations analysis, principal component analysis). Exposure analysis was also performed to find significative differences of vulnerability related to poultry type (duck or chicken) or poultry production (meat or egg) (null hypothesis: no vulnerability difference can be detected between groups; alternative hypothesis: a difference can be detected between groups). We class the scores in two categories, not vulnerable (Score < 0.5) and vulnerable (Score ≥ 0.5) to perform exposure analysis, considering the type (poultry type, production type) as risk factor. We calculated p-values for Student t-test, odds-ratios and adjusted odds-ratios (confidence intervals calculated at 5% type I error), and p-values for Chi-square, Mantel-Haenszel Chi-square and Fisher-Snedecor tests [16].
Spatial analysis of scores was also performed in order to highlight any relationship of vulnerability with environmental condition or geographic organization of space, and to find any spatial pattern of vulnerability susceptible to increase the risk of HPAI farm-to-farm transmission. We performed global spatial autocorrelation using Moran's I index (relationship between value and distance between farms), local spatial association analysis using Getis-Ord LISA index (relationship between the value of a farm and the value of neighbor farms) and search of spatial clusters using spatial scan method [17][18][19]. Data were mapped, statistically and spatially analyzed with the SavGIS geographic information system software [20].

Ethics Statement
In this study, no samples (biological or not) were collected. Farm environmental description and interviews were always conducted anonymously outside the farms. No participant names were recorded, and data were analyzed anonymously. Ethics approval was therefore deemed unnecessary by the Mahidol University IRB.

Results
All factors were recorded for the 173 farms in the survey, without any missing data. The consent to participate in the door-to-door interview was 100%, and no specific problems occur in the process of data collection (a pilot survey was first carried out in 2010 to adjust the questionnaire used to describe agricultural practices).
Statistical results by factors (proportion of yes answers, standard deviation) are given in Table 2. Scores of vulnerability are presented in Table 3. Category 3 (bio-security measures and farm management) is the one with the worst average score.   Analysis of vulnerability by type of production (meat or layer) revealed that excessive vulnerability correlates with farms producing eggs ( Table 4). The type of production is very significant for vulnerability since factors associated with eggs production may strongly contribute to increasing vulnerability. Furthermore, the same results are found whenever we take into account the vulnerability factors directly related to egg production. Analysis of vulnerability by poultry type (duck or chicken) shows no significant differences (Table 5). With the overall score classified into two classes (not vulnerable, s < 0.5 and vulnerable, s ≥ 0.5), the odd ratio adjusted by poultry production doesn't show a significant difference between classes. Finally, the study of scores using all possible combinations between production type (meat or layer) and poultry type (duck or chicken) shows that farms producing duck eggs are the most vulnerable in all categories, with a difference between groups always significant at a 5% type I error (Table 6). Spatial analysis was performed to help in identifying further contributing factors in environment or geographic context or organization of space. Figure 4 and Figure 5 show the spatial distribution of the vulnerability score for layer farms and for meat farms. Significant global spatial autocorrelation of global vulnerability score (Z-score > 1.96) is found within a short radius for the spatial weight used in the spatial autocorrelation index (search radius up to 10 km). With automatic cluster detection using a spatial statistic scan, we found a most significant spatial cluster for the vulnerability score in our study, and this cluster is spatially close to the most significant cluster of the H5N1 HPAI cases that occurred during the epidemics from 2004 to 2008, and automatically detected with the same technique ( Figure 6). Local spatial association is also found (Figure 7), showing again a clustered pattern, especially for high level of vulnerability.    Spatial analysis performed by separating farms into two groups by poultry production (meat or layer) showed no significant spatial clustering in each group separately.

Discussion
For more than 20 years, the poultry production in Thailand has radically shifted to a more industrial structure, and in 2002 more than 70% of national poultry production was oriented to the export market [13]. Industrial poultry production system in Thailand consists of vertically integrated companies controlling every stage of the production chain, from breeding to marketing processed poultry [21]. Those companies offer contractual agreements to private farms and purchase poultry products from them. The firm holds considerable control over raw material production without ownership of the production units. Contract farmers have an assured market for their poultry, access to the firm's services, and easier access to credit from banks [22]. Independent farmers mostly sell their larger surpluses to aggregators who drive pickups from village to village, purchasing birds from numerous different farms. A small amount of eggs and live or slaughter birds are sold to "End Users" (neighbors, restaurants or village, market vendors). Due to this production system, 98% of commercial poultry producers (sector 1-3) belongs to sector 2 and 3 in Thailand [3]. Sector 4 still exists as a source of suppliers for local markets.
Our goal was therefore to determine the actual level of vulnerability of this medium and small scale commercial poultry production system (sector 2 and 3) and to determine the most hazardous poultry farming practices related to the risk of avian influenza contamination and spread.
Our study clearly showed that vulnerability among farms in the province of Nakhon Pathom is still very high. It appears that the vulnerabilities related to poor farm management are predominant. Major deficiencies in bio-security measures were found in the areas of farm management, including transport, restocking practice, manure utilization, water management, rodent and insect control, frequency of disinfection, worker routines, hygiene practices and contact between farms. This is undoubtedly the result of a lack of awareness among farmers regarding the potential risks of epidemics. They are relatively complying with current regulations, but not all are entirely convinced that the risk of infections is still real, after a long period without warning since the last outbreak (in 2008).
Easy access to farm through entrances was found in 76% of farms, as the doors are open with no locks or warning sign. This percentage is very high and this is a major issue in vulnerability regarding farm to farm contamination.
75% of farms are located next to a pond or a pond is present at the farm. 19% of farms practice an integrated poultry and fish farming. Droppings and food spilled by chickens fall directly into the water and feed the fish. If birds are housed without proof housing this practice is considered as high risk to disseminate avian influenza virus [6,[23][24][25][26]. Furthermore, constant presence of wild birds or ducks attracted by the pond can contribute to the spreading of the disease. Finally, disinfection of poultry houses is not practiced in integrated poultry-fish farms since farmers are afraid of killing the fish. All farmers declare the regular presence of wild birds on the farm. 82% have no proper food storage or keep an open door at poultry food storage.
Although 85% of farmers replied positively to the use of rodent control, 83% of them use cats as the unique means of rodent control. The use of insecticides is very uncommon, and 94% think insects are not an issue. Even if insects themselves do not represent a known risk for the spread of avian influenza, insect infestation is a marker of poor hygiene and poor maintenance of the farm.
The transport of poultry, food, or fomite is an important risk factor for the transmission and dispersion of avian influenza [27,28]. On contractor farms, the company tracks and collects poultry or delivers food directly to the farm, unlike 39% of independent farms who delivers their products to various aggregators. These aggregators move from one farm to another, thus increasing the risk of avian influenza virus entry [29]. This practice represents a significant risk of spreading pathogens between farms if the disinfections of trucks are not applied at the entries and exits of the farm [30]. Similarly, the transport of poultry to the slaughterhouse several times per production cycle and restocking with non-compliance the rule "all in, all out" is also a risk factor [6]. In 93% of visited farms, disinfectant barriers at the farm entrance do not exist. 55% of farms do not practice trucks disinfection. Moreover, at all farms, truck drivers leave vehicles and walk around in their own clothes. These results show a very high vulnerability issue.
Employers of 93% of the farms do not provide or request work clothing, and at 38% of farms the same employees work on several different poultry units. Most farms do not provide food for the workers, and they frequently go out (more than two times a day) to buy a meal. Moreover, 37% of farms permit street food vendors to enter the farms and sell food to workers. Again, these results show a high vulnerability issue regarding virus dissemination.
Regular cleaning of poultry rearing equipment is practiced on only 26% of farms. Farm disinfection is applied at 59%, and no one in a poultry farm above a fishpond uses disinfectant. Disinfection of poultry houses at the end of the poultry production cycle is widely applied but still not applicable on poultry rearing equipment and egg trays. As animal food, fresh manure is used at 57% of farms, and 50% use dead fowl of farms as animal food.
Farms producing eggs are more vulnerable than farms producing birds and this result is consistent with the type of farms mostly affected during the 2004-2008 outbreaks [2]. Layer farms are more vulnerable than meat farms: factors associated with egg production strongly contributes to the vulnerability, and many farms producing eggs have low results regarding vulnerability factors linked to egg production.
Although almost all farmers knew about avian influenza, 63% of them do not think it is a dangerous disease and therefore are not concerned. Farmers adhere only to what is requested by the DLD GAP regulations. The long period without warning since the last outbreak of HPAI could also be an element of a relative lessening of farmers' vigilance.
Spatial clustering of vulnerability is also an important bio-security issue; infection in a farm may spread to neighboring farms if vulnerability of neighbors is high. Compartmentalization (defined primarily by management and husbandry practices related to biosecurity) is recommended by FAO [6,23] and has been promoted to limit disease spread between farms. Nevertheless, our results show that many vulnerability issues are related to contacts (i.e., easy access to farms, behaviors of workers, low disinfection at entrance) and are clustered. These spatial clusters of vulnerability need to receive special attention from authorities to respect the purpose of compartmentalization.

Conclusions
The results of our survey in the Nakhon Pathom province has shown that numerous vulnerabilities still exist and could represent a significant risk for large dissemination in the event of HPAI re-emergence. These results strongly suggest that bio-security practices on the majority of farms still need improvement. Particular attention should be conducted when vulnerability is spatially clustered.
It was also noted that category 3 (bio-security measures and farm management) is the one with the worst average score. In order to improve on-farm bio-security, it is necessary to enhance the capacity of poultry workers and farm owners to recognize the importance of these security measures. This can only be achieved through careful communication with an engagement of farmers and communities. DLD GAP regulations should be sharpened; stricter LFS regulations and GAP formulation with the implication of details is highly recommended for bio-security measures and farm management with a special attention for egg production management and practices.