Using spatial tools for high impact zoonotic agent surveillance design in backyard production systems in central Chile

Distributed under Creative Commons CC-BY 4.0 Abstract Specific locations of backyard production systems (BPSs) in Chile remain unclear, creating dificulties for designing surveillance activities for promptly detecting zoonotic agents with high impacts on health, such as avian influenza and Salmonella spp. This study aims to prove the use of spatial tools for improving the surveillance of BPSs in central Chile. A stratified and proportional random sampling was performed in 15 provinces of the Valparaiso, Libertador General Bernardo O’Higgins and Metropolitana regions. In this sampling, 329 BPSs were detected. In the first stage, 329 random sample points were allocated within the study area that searched for BPSs with poultry or swine breeding. Then, these random points were validated with remote sensing and in the field by searching for the presence of rural or semi-rural areas, nearby crops and houses or small towns within a 5 km radius around each point, while points allocated over hills or water sources (lakes or rivers) were discarded. Over 70 % of the sampling points were correctly allocated. In Los Andes, Cordillera and Chacabuco, less than 50 % of the points were allocated within feasible sampling areas. From the total BPSs sampled, 89 % met the 5 km radius criteria, and in the provinces of Valparaiso, Cordillera and Cachapoal, over 20 % of the sampling points were outside the radius criteria. This study is the first in Chile to explore the locations and sanitary statuses of BPSs. Given the lack of knowledge about the specific locations of BPSs, their identification during field activities represents a high cost for the surveillance of pathogens. We argue that using spatial tools in BPS surveillance design is an important support for healthcare management.


Introduction
Poultry and swine bred in backyard production systems (BPSs) represent an important percentage of animal production activity, especially in developing countries. 1dditionally, BPSs are recognized as a support economic activity, mainly in rural areas, stimulated by the increase in demand for organic and/or clean production systems. 2 In general terms, BPSs can be characterized as possessing poor biosecurity conditions and low technological development for both the handling of animals and the optimal distribution of BPSs that improves animal husbandry. 3This situation generates a close contact between the backyard farmers and their families and the domestic animal species maintained at the location, such as poultry (hens, chickens, ducks and geese) and swine, pets, and wildlife.][9] Characteristics of BPSs represent a permanent risk to the national health status.][12][13] Backyard poultry and swine are considered the main carriers of a number of strains or subtypes of priority zoonotic agents, such as avian influenza and Salmonella spp, 11,14 becoming a zoonotic risk due to various factors including the direct contact between humans and sick animals and the risk of producing contaminated food (public health risk).Furthermore, a BPS may also undergo economic losses due to the high mortality of the affected species (e.g., highly pathogenic avian influenza). 15Despite its importance, there are few studies focused on this high-risk population stratum, and information about the sanitary statuses of BPSs in relation to the prevalence of infection with avian influenza, Salmonella spp. or any other zoonotic pathogen is scarce.The case of Salmonella spp.also presents a high level of underreporting due to the clinical signs and low severity associated with infections in both animals and humans.
Central Chile has the highest percentage of intensive productive poultry and swine establishments and number of animals.According to the last agricultural and forestry census in Chile, by 2007, there were more than 43.5 million poultry and 2.6 million pigs in the central zone, representing 83 % and 81 % of the total abundance of each species in the country, respectively. 16The number of BPS registered by the same date corresponded to 16,289 breeding birds and 2,282 raising pigs.Hence, with the lack of geolocation of these production systems, this study aims to prove the use of a spatial approach in the design and implementation of surveillance of zoonotic agents in cases where the location and distribution of the target population is unknown and to suggest a way to study these populations or those with similar characteristics.

Target population
The present study was performed in three regions of central Chile: Valparaíso, Libertador General Bernardo O'Higgins (LGB O'Higgins), and Metropolitana de San-

Short communication
tiago regions (Figure 1).According to the last Animal and Forestry census, in Chile, in the year 2007, 16 16,289 BPSs breeding birds and 2,282 BPSs breeding pigs were reported in the study area (Table 1).The sample unit was defined as the BPSs that breed poultry and/or pigs that keep up to 100 birds or 50 pigs to survey for zoonotic agents.

Study design and sample size
A stratified and proportional random sampling was performed, based on the 15 provinces included in the study area.Sample size was calculated following equation 1 adjusted by equation 2, both extracted from Dohoo et al., 2010 17 : where n = simple size; Z α = the value of Z α required for confidence = 1α, where α corresponds to the level of confidence; Z α is the percentile of a standard normal distribution (1α/2); p = the expected prevalence of the pathogen (e.g., avian influenza, Salmonella spp.); q = (1 -p), and L = the precision of the estimation, also known as 'allowable error' or 'margin of error'.
where n' = the adjusted simple size; n = the previous calculated simple size (eq.1); N = the number of BPS in central Chile. 16Assuming a lack of knowledge about the prevalence of priority zoonotic agents present in BPSs located in central Chile, sample size was calculated based on a prevalence of 50 %, ensuring the maximum sample size possible when using a sample size approach for estimating a proportion, 17 a confidence level of 95 %, and a precision of 5 %.BPSs raising pigs were used for the sample size calculation, supposing that they also breed poultry.Giving the final sample size of 329 BPSs distributed accordingly to each province, as detailed in table 1.

Random points allocation and validation
Random allocation of sampling points was performed using ArcGis 10 (ESRI 2011.ArcGIS Desktop: Release 10.Redlands, CA: Environmental Systems Research Institute) according to the sample size established for each province (Figure 1).Sample points where checked for feasibility using Google Earth and Google Maps to guarantee the achievability of sampling at those points.Variables considered to identify a given point as a possible BPS were the presences of rural or semi-rural areas (small towns), nearby crops and houses or small towns.Points that did not fulfil these criteria were considered poorly positioned and were relocated.A radius of five kilometres was considered from the random sampling point, allowing for the possibility of sampling around points allocated on hills or in places with no livestock activity.Field validation was then performed to check for BPSs within the established radius.

Results and discussion
After the allocation of random sampling points for each province, the checking process showed that 251 (76 %) of the random points were well positioned in relation to the feasibility of finding a BPS, given the proximity to land with agricultural use, and 78 (24 %) sampling points were poorly positioned (Table 2).These poorly positioned points were allocated over lakes, within residential areas or within the Andes Mountains, and there was no chance of finding a BPS even within a radius of 5 kilometres.Details of the random sampling points and the distributions of well and poorly positioned points can be observed in table 2. Poorly positioned points were re-allocated, removing the possibility of allocating points in the Andes mountains or those with the other conditions previously mentioned.
Provinces that presented poorly positioned points correspond mainly to those presenting a significant percentage of surface associated with the Andes mountains (Los Andes, Cordillera, Colchagua, Petorca, Cachapoal and San Felipe).In Chacabuco and Santiago, the number of poorly positioned points could be related to the use of land, which presents a trend to urban and agribusiness uses. 18Normally, land use consists of vegetable and fruit cultures, which increased the time spent finding a BPS to sample. 19This scenario should be quite different in the two other regions to be sampled, considering that the main activities in those regions are related to animal and agricultural farming activities. 16fter adjusting the positions of the random points, it was observed that 89 % presented a BPS within a radius of 5 kilometres, and only for 11 % was it hard to sample the 5 km radius.The most distant BPS was found approximately 13 kilometres away from the random point, a situation observed in the province of Talagante.Whereas, the closest candidate BPS was detected within less than 1 kilometre of the random point, also in the province of Talagante.Given that there is a lack of knowledge about the specific location of each BPS that raises pigs or poultry or any other animals, there is a big loss of time resources and equivalent funds just on the detection of a feasible point. 20,21This increases the duration of the sampling activities and often results in requiring over an hour just to find one BPS feasible to be sampled, affecting the efficiency of field activities and the quality of the samples. 21Delays in field activities could affect the integrity of the sample, depending on the type of the sample and the agent involved. 22,23revious work in our research group spent almost one day just to reach one allocated random point and then search for the closest BPS to sample (Di Pillo, F. data not published).According to the characteristics of these production systems, most birds or pigs are raised in partially enclosed systems, where animals are confined at night and released to an extensive rearing during the day, increasing the difficulty of sampling. 24he last agricultural and forestry census from 2007 in Chile underestimated the actual number of existing BPSs in the area, since it only registered those BPSs who perform tax payments associated with agricultural and livestock activities, leaving out of its records those who are without formal "agro" activities and maintain a low number of animals on their land.This scenario generates deficiencies in the design of surveillance programs, preventing an early warning of sanitary events occurring in this population stratum. 25,26Likewise, in the event of an emergency or a real-time outbreak, the time spent identifying and monitoring these populations becomes greater, affecting the health status and an early resolution of the event. 20lobal tendency leads to the incorporation of spatial tools for the design and implementation of surveillance programmes and the storage of geo-coding data for animal health research. 27In addition, this may help to establish risk based

Table 1 .
Demographic distribution of BPSs and sample size by province and region.

Table 2 .
Verification process for random sampling points and 5 kilometres criteria for each Province.