Campylobacter spp. remain one of the most frequent bacterial causes of foodborne gastroenteritis world-wide [1]. Poultry, and specifically consumption of undercooked poultry and mishandling raw poultry, is an important source of Campylobacter to humans [2–7]. The prevalence of broiler flocks colonized with Campylobacter spp. varies, ranging from 5% of flocks to more than 90% [8]. Once a flock is exposed, the bacteria spread rapidly throughout the flock, and most of the birds become colonized and remain so until slaughter [9–14]. Due to the difficulties in eliminating contamination of carcasses in slaughter plants, the control of Campylobacter in broiler flocks and production of birds free from colonization at slaughter, is essential for preventing human cases [5, 14–17].

Many researchers [11, 12, 14, 15, 18–28] have contested that the most likely source of Campylobacter to broiler flocks is the environment (i.e. horizontal transmission). Both Campylobacter-positive and -negative flocks can be present in different houses on the same farm during the same growing period [9, 11, 12], illustrating that it is possible to prevent Campylobacter from entering a broiler house. Thus, it might be hypothesized that certain characteristics of a broiler house, or management practices at the house-level, might influence the likelihood that a flock will be exposed to the bacteria. House-level factors associated with an increased risk of colonization include: concrete floors (compared to wood floors) [12]; feed dispenser in the anteroom (compared to in the chicken room) [12]; evidence of mice [12]; absence of, or ineffective, hygiene barrier or biosecurity measures [12, 14, 16, 29–31]; roof fans (compared to side fans) [16, 32]; a static ventilation system (versus a dynamic system) [33]; two or more persons taking care of the house [33]; non-cement floors [34]; and poor maintenance of house surroundings [34]. In Iceland, flock-level risk factors have been studied [32]; however, to date, the association between broiler flock Campylobacter status and the attributes and management practices of broiler houses in Iceland has not been studied.

The strong association between the increased incidence of human campylobacteriosis and increased consumption of fresh chicken meat in Iceland in the late 1990's prompted a longitudinal study of the poultry industry [35]. The ultimate goal of the full project was to identify the means to decrease the frequency of broiler flock colonization with Campylobacter, thereby reducing the burden of foodborne illness associated with poultry consumption. Our objective in this study was to identify house-level risk factors associated with the colonization of broiler flocks with Campylobacter spp. in Iceland.

Our study has identified four house-level risk factors associated with the colonization of broiler flocks with Campylobacter spp. in Iceland, out of a possible 30 explanatory variables. The higher ICC of a null model with a random effect at the farm level, compared to a null model with a random effect at the house level, suggests that there was more variation in flock Campylobacter status between farms than within farms. Thus, it was not unexpected that only a few house-level predictors remained statistically significant in the final model. In addition, our approach of screening variables within subsets enabled us to look at potentially complex relationships between predictors of a similar type, and identify those that were most strongly associated with flock status. For example, in some subsets (e.g. sanitation), including more than one predictor in the model led to unstable parameter estimates with high standard errors, therefore, only one predictor from the subset could be offered to the comprehensive model. Within other subsets, there was strong collinearity between variables, or between one variable and one category of another variable. For example, all six houses (contributing a total of 58 flocks) in which the roof and floor drains merged, also had roof drains that opened into a septic tank with overflow into an underground bed (Kendall's τ_b = 1.0). In these subsets, only one of the two variables could be included in the analysis; the predictor resulting in the model with a smaller Akaike's Information Criteria (AIC) value was chosen. Additionally, there were several variables in which the responses were extremely one-sided; almost all producers cleaned the ventilation system after every flock and changed boots before entering the house, and all but one broiler house in Iceland have concrete floors (the floor of the upper level of one house included wood elements). These variables were eliminated from the analysis because it was unlikely that there would be enough power to detect such small differences.

Misclassification could have been a potential source of bias in our study. For some variables, it was necessary to combine categories if there were only a small number of observations in one or more categories. However, we expect this bias was small because we grouped only those categories that were biologically-related, and only when a Wald's test indicated that the categories were not statistically different from each other. In addition, eliminating predictors due to problems with collinearity or unstable parameter estimates might have affected our final results. However, we found that predictors that were strongly associated with flock status in the subset analyses, were also statistically significant in the final model. Thus, it is likely that our screening process was effective in identifying the most important of the house-level predictors.

For flocks with multiple catch lots, more pooled samples were obtained, and we considered whether this might have increased the likelihood that the flock would be positive for Campylobacter. In a previous analysis of flock-level factors [32], the number of catch lots was not a significant risk factor for the colonization of Icelandic broiler flocks with Campylobacter, after controlling for flock size. Most of the positive catch lots in our study were positive in all four pooled caecal samples, inferring high sensitivity of the methodology for the sample type. On a flock basis, 84% of the positive flocks yielded Campylobacter in all samples collected, and of the positive flocks with more than one catch lot, a high proportion were positive on the first catch lot. Several standard management practices in Iceland might have contributed to these findings. Beginning in 1999, before the start of the study period, the Icelandic poultry industry adopted a high standard of cleaning, chemical disinfection and drying of live haul crates and trucks, under the assumption that bringing crates covered with faecal material into broiler houses for partial depopulation was an important source of contamination to the remainder of the flock. Unlike other countries, Iceland does not have commercial catching crews that travel from farm to farm. The workers on each farm catch their own birds and follow their own biosecurity rules. In addition, the reason for collecting and slaughtering a flock in multiple catch lots differs in Iceland compared to other countries. In Iceland, the practice is based more on slaughter line speed capacity in the abattoirs; only the largest flocks (over 20,000 birds) are slaughtered in three or four catch lots, typically over three or four consecutive days. For flocks slaughtered in more than one catch lot, the average interval between catch lots is quite short. The practice of thinning (i.e. slaughtering half the flock as broilers and the second half as larger roasters) is rare in Iceland, and is typically done only during the Christmas season. During the winter, the prevalence of Campylobacter in broiler flocks is very low (see Methods). Therefore, because of the relatively short time interval between shipping catch lots, the low within-flock prevalence used in our sample size calculation to detect early stages of colonization, the high sensitivity of the isolation method, and the high proportion of catch lots that were positive in all samples, we expect that there was little variation with respect to the risk of false negative classification between flocks slaughtered in single and multiple catch lots.

Wild birds captured on broiler farm premises frequently carry Campylobacter jejuni [13, 19, 28, 36–39]. Campylobacter prevalence has been shown to be higher on farms with frequent sightings of wild birds than on farms with few wild birds [12]. A previous study in the UK [16] has found that vertical ventilation systems are associated with positive flocks. In the present study, in an effort to determine if wild birds might be a source of Campylobacter to broiler flocks, we classified the ventilation system on each broiler house according to its type (vertical, horizontal, or both), and integrity (open or closed). An underlying assumption of our classification was that wild birds might be more likely to perch on vertical ventilation shafts, and, if those vents were open, the flock might be more likely to be exposed to the bacteria, either directly (e.g. wild birds entering the house or defecating into the vents), or indirectly (e.g. rain water washing faecal material in through the vents). From the subset analysis, we found that after controlling for ventilation type, the integrity of the ventilation system was not a significant risk factor for flock colonization (OR = 1.0, p = 0.975). However, the risk of colonization was predicted to increase when the flock was raised in a house with vertical, or vertical and horizontal ventilation shafts compared to horizontal vents. There are a few possible reasons for these findings. First, our classification of ventilation type might not have been refined enough (e.g. we did not differentiate between air inlets and outlets, and wild birds might preferentially perch on vertical outlets for warmth). Although researchers in Sweden [12] did not find an association between Campylobacter occurrence and air outlets (ceiling or wall) or inlets (passive ceiling, active ceiling, or active or passive wall vents), the authors stated that their conclusions might have been uncertain because confounding (univariable analysis only) and clustering were not accounted for in the analysis. Secondly, there might be some mechanism related to the type of ventilation system other than wild birds that contribute to flock colonization. Our finding of a protective effect of horizontal shafts might be related to the thoroughness of house sanitation, as it has been suggested that horizontal fans are more accessible for proper cleaning and disinfection [16]. A brief exploration of interactions between ventilation type and each sanitation variable revealed a significant positive interaction between vertical ventilation shafts and the use of bacterial disinfectant and/or fumigation, although, we did not include this interaction in the comprehensive model due to highly inflated and unreliable parameter estimates. Finally, the effect of ventilation type might be related to other potential sources of Campylobacter such as flies. Researchers in Denmark [40] have shown that flies captured within 50 m of a broiler house carried Campylobacter, and that per volume of ventilation air, 4.5 times the number of flies entered the house through a roof inlet than through wall inlets. Considering our findings, and those of Hald et al. (2004), further investigation in this area is warranted.

The risk of Campylobacter colonization was predicted to increase when the flock was raised in a house in which the producer's boots were cleaned and disinfected prior to entering the house. Using univariable analyses, several researchers [29–31, 33] have not found an association between flock colonization and routines for cleaning and disinfecting footwear. Others [41] have noted an increasing percentage of Campylobacter-positive flocks with a decreasing hygiene score (including the disinfection of boots), although the hygiene score itself was not statistically significant. Campylobacter jejuni has been isolated from farmer's boots [13, 19, 22, 39], and from a footbath containing tap water at the broiler house entrance [20]. Several studies in the UK [14, 16, 21] have shown that the effective use of boot dips prior to entering the houses either delayed or prevented colonization. Although our results were inconsistent with other studies, researchers in Sweden [12] noted that farmers were frequently careless about boot dipping (e.g. only dipping toes or heels, passing through the disinfectant very quickly, or dipping boots when clumps of mud were present). Thus, our results might reflect the improper use of boot dips by Icelandic producers, or, ineffective disinfectant solutions. An alternative explanation for the positive association might be related to the wording on our questionnaire. Producers were asked about cleaning and disinfection before entering the house (always, sometimes, or never). In retrospect, this question might not have been precise enough and we cannot be certain about what specific practices were represented or the consistency of recording. Nevertheless, the finding of an increased risk of colonization should encourage producers to assess the use and effectiveness of disinfectant boot dips, and their general cleaning and disinfection procedures prior to entering broiler houses.

Our study has shown that the risk of Campylobacter colonization was higher when the flock was raised in a house cleaned with geothermal (high temperature) water. The isolation of Campylobacter from surface water [20] and puddles [28] adjacent to broiler houses, points to possible environmental sources of Campylobacter to broiler flocks. Although not all geothermal wells were on-farm (i.e. some farms had geothermal water piped in), potentially, farm-based geothermal wells could have warm surface water pools, which might serve as Campylobacter reservoirs for birds, flies and other insects, and for the broiler flocks. However, several researchers [42–47] have found that isolation rates and/or survival of thermophilic Campylobacter spp. from various water sources were highest when the water temperature was between 2°C and 10°C, and lowest when the temperature exceeded 15°C. Thus, the reason for the positive association between Campylobacter status and the use of geothermal water for cleaning remains unclear. In Norway, the prevalence of Campylobacter in three surface water sources (lakes and rivers) was strongly associated with the number of indicator bacteria (from effluents, farming, and waterfowl) in the water [44]. Therefore, in our study, geothermal water might be a surrogate for farm location and associated environmental pressures for Campylobacter (e.g. proximity to Campylobacter reservoirs in cattle herds, broiler breeder and egg layer flocks, sheep, migratory waterfowl, or other environmental sources). To our knowledge, this is the first time this risk factor has been identified in the literature, and it will be of interest to ascertain if this predictor is recognized in future studies and to determine the mechanism of its effect.

Our study has identified four risk factors related to broiler house attributes and house-management practices, for the occurrence of Campylobacter in broiler flocks in Iceland; vertical or vertical and horizontal ventilation shafts, cleaning and disinfecting boots prior to entering the broiler house, and the use of geothermal water for cleaning houses. We found that horizontal vents had a protective effect, therefore, extra care should be taken when cleaning vertical vents, and producers should ensure that the disinfectant used has an appropriate bactericidal effect for Campylobacter. Alternatively, the increased risk associated with vertical ventilation shafts might be related to the height of the vents and the potential for vectors such as flies to gain access to the house. This is an area that warrants further research, and further refinement in the classification of vertical and horizontal systems might be necessary for this purpose. For newly constructed houses, horizontal ventilation systems could be considered. Boot dipping procedures, including the frequency and thoroughness of dipping boots, the frequency of changing the dip, and the effectiveness of the dip against Campylobacter should be examined on farms experiencing a high prevalence of Campylobacter. Future studies are warranted to ascertain how geothermal water increases the risk of colonization, and to determine if this factor is a surrogate for micro-climatic conditions or agro-environmental pressures on the farm and surrounding region.