table of contents table of contents

The objective in this study was to identify risk factors for flock …

Home » Biology Articles » Agriculture » Animal Production » A farm-level study of risk factors associated with the colonization of broiler flocks with Campylobacter spp. in Iceland, 2001 – 2004 » Methods

- A farm-level study of risk factors associated with the colonization of broiler flocks with Campylobacter spp. in Iceland, 2001 – 2004

Target and study populations

The target population for our study was commercial broiler flocks raised in Iceland between May 2001 and September 2004. Our initial plan was a total-population census sampling of all flocks from commercial broiler production farms in a three-year longitudinal study beginning in May 2001; sampling was later extended by five months to include a fourth peak summer season. Sampling was carried out in the three commercial abattoirs in the South of Iceland, where commercial production was located. This level of full cooperation by broiler producers was in part due to public and media attention to the link between broiler chickens and the campylobacteriosis epidemic in 1999, and due to a price penalty to producers for positive flocks. Producers and processors were keenly interested in finding solutions. When the study began in May 2001, only one farm was excluded, due to its remote location and small production. The excluded farm slaughtered its own small flocks in an on-farm facility and sold its products locally. It ceased production in April 2003 and as best we can determine by hatchery records, it raised 12 flocks during the study period with a maximum flock size of 3,000 birds and total production of less than 36,000 broilers. During the course of the study several new farms joined the study with their first flocks. Two of these new farms were excluded due to their remote location. One was located on a coastal island, distant from the study area. It raised only two flocks with a total of 14,900 broilers before closing operations. The other excluded farm was located on the North coast of Iceland, with production exceeding initial expectations. It produced 147 flocks with a total of 1,241,026 broilers during the course of the study and an average flock size of 8,442 birds. Including the estimated production from the first excluded farm, only 161 flocks (contributing 11% of the total broiler production in Iceland during the study period) were excluded from the study. Of the 1,425 flocks included in the study (total production of 10,387,169 broilers), the maximum flock size was 23,470 birds, the mean flock size was 7,289, and the median flock size was 6,142 birds. There were 47 flocks with less than 2,000 broilers, and 36 flocks with over 20,000 broilers.

Characteristics of the farms

Commercial broiler flock production technology in Iceland is essentially the same as that in North America and Europe. Breeder production of hatching eggs, hatchery technology, broiler ventilation, feeding and water delivery systems are the same. Scale of production is smaller on average, with newer broiler barns being state of the art and more comparable in size to large scale production elsewhere. Icelandic broiler houses have concrete floors and floor drainage systems as the standard, which is a notable difference from broiler barns in North America, especially the US. A number of farms also use geothermal water to heat the broiler houses and to wash out the pens. Farms can be a mixture of newer, larger broiler houses, with original houses being smaller, whereas newer farms will have only newer-style, large broiler houses.

Data collection

Pilot sampling of broiler flocks, including pooled caecal samples, were conducted for three months of production in 2000. General data on the characteristics of each farm were gathered at the beginning of the study through a combination of phone interviews and site visits by the Veterinary Officer for Poultry Diseases of the Agricultural Agency of Iceland (Reiersen). Collection of epidemiological data began in May 2001. Since no problems were encountered during the initial collection of data, all sampling data were included in the dataset for analysis. Questionnaire designs were the collective effort of five veterinarians (including four epidemiologists) and a biostatistician. Included in the design group was the Veterinary Officer for Poultry Diseases, who had an in-depth knowledge of each farm as a result of working with the producers to eradicate Salmonella from poultry. There were several questionnaires, the main one designed to record independent variables acting at the various levels of broiler production (i.e. at the flock, house and farm levels). During the interval between flocks in each broiler house, a field technician employed by the Veterinary Officer for Poultry Diseases visited each farm to record responses from face-to-face interviews with the person most closely associated with the hands-on management of the broiler flocks and houses, and to record observations of cleaning and disinfection procedures between flocks. The design team reviewed all questions and the method of recording with the field technician to ensure clear understanding. The Veterinary Officer for Poultry Diseases accompanied the field technician on all farm visits and questionnaire recording for the first full month of sampling. During the course of the study, two university-educated field technicians were employed. The first technician was employed for two years, and trained the second technician for one month prior to leaving the project. Interview times varied from 10 to 15 minutes per questionnaire, depending on whether the producer needed to verify records. To ensure consistency in responses, data collected at the previous visit were reviewed with the producer. All questions pertaining to our analysis were closed. Although other factors potentially relevant to the complex epidemiology of Campylobacter were included in the questionnaires, it was our intent in this study to specifically identify risk factors operating at the farm level. The set of factors chosen for this analysis were deemed both sensible and comprehensive to satisfy the objectives of this study and were in keeping with farm-level factors identified in the literature.

The slaughter plants provided additional data, in the form of monthly reports summarizing records of flocks slaughtered each day.

Bacteriological sampling and processing

Depending on the size of the flock and management practices on the farm, broiler flocks were shipped to the slaughter plant in one to four catch lots, defined as a group of birds collected on one day and transported to the slaughter plant. The maximum trucking distance was 100 km. In Iceland, live haul crates and trucks are cleaned and disinfected with great care, and there are no commercial catching crews (i.e. all flocks are caught by each farm's workers). Caecal samples were chosen to ensure representation of farm-origin flock Campylobacter status, and for their higher sensitivity compared to cloacal swabs or faecal samples. At the processing plants, systematically selected caeca (including contents) were excised from 40 birds from each catch lot by veterinarians employed by the Chief Veterinary Office of Iceland and placed in sterile WhirlPac bags to create four pooled samples containing ten caeca each. The caeca were collected from the viscera pack of carcasses on the evisceration line in the abattoir, after automatic evisceration. Flock slaughter lots are well-separated in Icelandic abattoirs, which facilitates clear flock identification. The sampling protocol was to select an indicator carcass (not sampled), and then collect one caecum from each 10th or 5th subsequent carcass, whichever frequency worked best for work flow. Caeca were collected using one pair of latex examination gloves per pooled sample; gloves were changed between pooled samples. Caeca were removed by manually freeing an individual caecal loop from connective tissue, pinching it off at its base, and pulling it free. Samples were then transported and processed at the Laboratory of the Institute for Experimental Pathology, Keldur, Iceland, either the same day or after holding overnight at 4°C. The required sample size per flock was estimated to detect early stages of flock colonization or alleles with poor colonizing ability on the basis of a within-flock prevalence as low as 10%; four pooled samples would ensure 99% confidence of detecting at least one positive bird in a catch lot [28]. Serial dilutions of caecal contents were plated on Campy-Cefex agar [29] and incubated at 42°C under microaerobic conditions for 48 hours. Colonies were counted, and confirmed as Campylobacter spp. by microscopy and latex agglutination (DrySpot Campylobacter test kit, Oxoid DR0150M).

Although enumeration was not required for this epidemiological analysis, the choice of a direct plating method that enabled enumeration was important to other aspects of the large multi-disciplinary project. Campy-Cefex was chosen due to low cost, good sensitivity and enumeration on caecal samples. The method requires 24 to 48 hours for confirmed detection of Campylobacter spp. (versus at least 72 hours for the NMKL method), enabling identification of positive flock lots to obtain retail product samples prior to distribution (two cartons of ready-to-ship broiler carcasses were held pending caecal sample results for another component of the full project). During the first eight months of sampling, caecal samples were analysed by both methods [30]. Based on the results of this comparison, the Laboratory of the Institute for Experimental Pathology concluded that the Campy-Cefex method was at least as sensitive as the NMKL method for detecting Campylobacter spp. in poultry caecal samples, and began using Campy-Cefex for their official Icelandic surveillance program. Since genetic sub-typing was deemed necessary for other project analyses, we were unable to go further into the species identification of the isolates.


A broiler flock was considered positive for Campylobacter if at least one of the pooled samples from any of the catch lots was positive on culture. Data were then collapsed to the farm level, such that the number of positive flocks and the total number of flocks raised were summed for each farm.

Summer data

Since the clear majority of positive flocks in our study were detected during the warmer months, we focussed our analysis on flocks raised during this high risk period to reduce problems associated with interactions of management factors with season [31]. Thus, flocks that hatched between March 15 and September 15 of each year of the study were considered to have been raised during the summer season. This definition of summer corresponds to the periods of restrictions imposed by the Icelandic government on when manure is allowed to be spread on fields and pasture (March 15 to October 31).

Definition of farm-level variable

A farm-level variable was defined as one that was consistent for all houses on a farm during the study period. However, as can be expected over a three and a half year study, producers may have instituted changes at the farm level such that flocks raised in the early part of the study were subjected to a different management practice or circumstance than flocks raised in the latter part. In this situation, if at least 80% of the flocks from a farm were subjected to a particular management practice, then that practice was deemed to be the standard for the farm.


Table 1 lists the categorical variables that were available for analysis. Only farms with complete data for all variables (28 farms) are shown since only these farms were included in the multivariable analysis described below. Due to the small number of farms, for categorical predictors with more than two levels, categories were combined if it was biologically sensible to do so. Continuous predictors are summarized in Table 2.

Table 1. Farm-level categorical variables available for analysis of Campylobacter colonization of broilers in Iceland (n = 28a)

Table 2. Farm-level continuous variables available for analysis of Campylobacter colonization of broilers in Iceland (n = 28a)

Initial screening of categorical variables consisted of identifying those that were highly correlated with each other (Kendall's τb ≥ 0.8) (Table 1).

Multivariable modelling

Since the goal of our model-fitting process was primarily aimed at identifying the most important of the farm-level predictors, we examined a number of potentially useful models. Six logistic regression models, using the binomial distribution to adjust for the number of flocks from each farm, were fitted to the data using both automated and manual variable selection methods. The models were of the following form:

ln [pi/(1-pi)] = β0 + β1X1i + ... + βkXki

where pi is the proportion of positive flocks from farm i, β0 is the intercept, and β1X1i + ... + βkXki is the linear combination of predictor variables for the ith farm.

Four different automated model selection methods were applied to fit adequate models using any of the predictor variables listed in Tables 1 and 2. When manual selection methods were employed, only one of the two strongly correlated variables for manure spreading was included with the remaining available predictors. The model selection procedures were as follows: 1) automated forward selection; 2) automated forward stepwise; 3) automated backward selection; 4) automated backward stepwise; 5) manual backward selection using the variable "manure spread on fields in summer season"; and 6) manual backward selection using the variable "manure spread on fields in winter season". For all models, the test of a term's significance was the likelihood-ratio test; variables with p < 0.05 were eligible for addition to the model and p ≥ 0.05 were eligible for removal. The likelihood-ratio test was also used to evaluate the significance of groups of variables (i.e. farm water source). Akaike's Information Criterion (AIC) was recorded for each model. The linear relationship between each continuous predictor and the outcome was evaluated by adding a quadratic term to the regression model and assessing its significance, with p ≤ 0.05 indicating a non-linear relationship. In the manual selection methods, as each variable was removed from the model, confounding was deemed to exist, and the variable was retained in the model, if the coefficient of another significant variable changed by more than 20%. With one exception (see discussion), interactions were not assessed since there were relatively few farms. Model diagnostics included the calculation of Pearson residuals to identify outliers; observations with large residuals were further evaluated by re-fitting the model without the observation and comparing the coefficients to the full model. Potential influential observations were identified by examining large Cook's distance values. Stata software version 9 (StataCorp, College Station, TX, USA) was used for all statistical analyses.

rating: 0.00 from 0 votes | updated on: 11 Aug 2008 | views: 13089 |

Rate article: