Estimates of the burden of illness for eight enteric pathogens associated with animal contact in Canada

SUMMARY Enteric pathogens are commonly known to be transmitted through food or water; however, contact with animals is another important transmission route. This study estimated the annual burden of illness attributable to animal contact for eight enteric pathogens in Canada. Using data from a Canadian expert elicitation on transmission routes, the proportion of enteric illnesses attributable to animal contact was estimated for each pathogen to estimate the annual number of illnesses, hospitalizations and deaths in Canada. For each estimate, a mean and probability intervals were generated. Of all illnesses caused by these eight pathogens, 16% were estimated attributable to animal contact. This estimate translates to 86 000 (31 000–166 000) illnesses, 488 (186–890) hospitalizations and 12 (2–28) deaths annually for the eight pathogens combined. Campylobacter spp. is the leading cause of illnesses annually, with an estimated 38 000 (14 000–71 000) illnesses occurring each year, followed by non-typhoidal Salmonella spp. (17 000, 6000–32 000). The majority of hospitalizations were attributable to non-typhoidal Salmonella spp. (36%) and Campylobacter spp. (31%). Non-typhoidal Salmonella spp. (28%) and Listeria monocytogenes (31%) were responsible for the majority of the estimated deaths. These results identify farm animal and pet/pet food exposure as key pathways of transmission for several pathogens. The estimated burden of illness associated with animal contact is substantial.


INTRODUCTION
Enteric zoonotic pathogens are commonly known to be transmitted through food or water; direct or indirect contact with an animal is another key route of transmission [1][2][3]. The burden of enteric (acute gastrointestinal) illness associated with contact with farm animals, wildlife, domestic pets and their environment is not well known in Canada. Several outbreaks have highlighted pets as an important route of transmission. Salmonellosis outbreaks linked to pet food and treats [4,5], live poultry, reptiles and amphibians, and rodents have been reported in Canada [6][7][8] and the United States (USA) [9][10][11][12][13][14][15] in recent years. Outbreaks of verotoxigenic Escherichia coli (VTEC E. coli) associated with petting zoos have been identified several times [16][17][18][19]. In particular children are at a higher risk from this transmission route, given their behaviors (close proximity with pets, petting zoos, more likely to put hands in their mouth, etc.) and developing immune systems [16,[20][21][22][23].
Enteric illnesses are underascertained by public health surveillance systems because of underdiagnosis and under-reporting [28]. In order to more accurately estimate the burden of illness associated with animal contact, it is necessary to account for underascertainment and to estimate what proportion of illnesses result from animal transmission, as pathogens rarely are exclusive to one transmission route. To account for the underascertainment at each level of the public health surveillance system (i.e. case seeking medical care, submitting a sample, the sample being tested and found positive, and the positive test result being reported), pathogen-specific multipliers have been used. Estimates related to foodborne illness, and the associated hospitalizations and deaths in Canada and the USA, have been developed using this method of developing underascertainment multipliers and incorporating pathogen-specific source attribution estimates [29,30].
The USA has estimated the burden of illness (using underascertainment multipliers as described above) associated with animal contact for seven key enteric pathogens and found that 14% of illnesses are attributed to contact with animals and their environments [31]. Reported case data from the province of Ontario, Canada, identified contact with animals as the primary source of exposure for nearly 20% of the reported cases due to 14 pathogens [32]. Estimating transmission routes for enteric pathogens is difficult due to the lack of data. Expert elicitations have been conducted in Canada and internationally to estimate the proportion of enteric illnesses attributed to different transmission routes including foodborne and animal contact [33][34][35][36].
The objective of this study is to estimate the number of illnesses, hospitalizations and deaths in Canada related to animal contact, using estimates of enteric illness in Canada and the proportion of cases attributed to animal-related contact based on an expert elicitation [37]. The expert elicitation asked 31 Canadian experts to estimate the proportion of cases of 28 pathogens transmitted by five main transmission routes (food, water, animal contact, person-to-person and other) and select subcategories of the food, water and animal contact routes [35,37]. The current study builds upon the elicitation results by estimating the number of cases related to the animal contact route and subcategories of domestic pets, farm animals and wildlife specifically. These burden estimates can be used to better describe and increase awareness of this public health problem, and inform advocacy, education and further research activities, to prevent and reduce enteric illnesses associated with animal contact.

METHODS
This study estimated the annual number of animalrelated illnesses, hospitalizations and deaths related to eight key pathogens (Campylobacter spp., non-typhoidal Salmonella spp., Giardia sp., Cryptosporidium spp., VTEC O157, VTEC non-O157, Yersinia enterocolitica and Listeria monocytogenes). Pathogens selection for this study was, in part, based on the seven included in the US study [31] with the addition of Giardia sp., which was selected based on the Canadian expert elicitation attributing 14% of giardiasis to animal contact. Other zoonotic enteric pathogens were not included due to the lack of available data.
Data focused on the 2000-2010 time period for the underascertainment multipliers and laboratoryconfirmed cases, hospitalizations and deaths and was based on the approximate Canadian 2006 census population (32 500 000), to follow the methods published for the burden of foodborne illness in Canada [28,30]. Details on specific data sources for the illness estimates and proportion of illness caused by animal contact are provided below. The reported and estimated values used in this analysis are provided as reference in Table 1. All estimates reflect the 2000-2010 time period.

Estimating total illnesses, hospitalizations and deaths
The methods for estimating the number of domestically acquired foodborne illnesses are described in detail elsewhere [28]. In brief, using this approach of estimating the number of domestically acquired cases by pathogen data on the number of laboratory confirmed cases for seven of the eight pathogens were obtained from national surveillance systems for the time period 2000-2010. These values were then adjusted for under-reporting (i.e. laboratory confirmed but not reported to local/provincial/territorial public health and national surveillance systems) and underdiagnosis (i.e. those who do not seek medical care, sample is not submitted, tested or found positive for causative pathogen) based on data from Canadian National Studies on Acute Gastrointestinal Illness (NSAGI) population, laboratory and public health reporting surveys [38][39][40][41][42]. An alternative approach was used for estimating VTEC non-O157 cases, which is not routinely identified and reported in Canada. For VTEC non-O157 cases, a ratio (1 VTEC O157:1·6 VTEC non-O157, based on literature [43]), relative to the estimate of VTEC O157 cases was used [28,30]. The methods for estimating the number of domestically acquired foodborne hospitalizations and deaths are described in detail elsewhere [30]. In brief, the methods for estimating hospitalizations and deaths for these eight enteric pathogens relied on the number of hospitalizations and deaths for each pathogen reported in the Canadian Institute for Health Information Hospital Morbidity Database (during the 2000-2010 time period, for certain pathogens only a subset of this time period was available, Table 1) [30,44]. These values were then adjusted for under-reporting (in the hospital database) and underdiagnosis (i.e. sample is not submitted, tested or found positive for causative pathogen). International travel-related illnesses, hospitalizations and deaths were excluded by subtracting the pathogen-specific proportion that is travel-related.

Proportion of illnesses caused by animal contact
The pathogen-specific proportion of illness attributable to animal contact was estimated using findings from a recently conducted Canadian enteric illness transmission expert elicitation [35]. As reported in the Methods section of the Canadian expert elicitation study, 31 experts estimated the proportion of illnesses caused by 28 pathogens transmitted via major transmission routes (foodborne, waterborne, animal contact, person-to-person and other) and for select subcategories of those major routes, at the point of pathogen consumption. The elicitation consisted of a snowball expert recruitment and collection of background information on experts, an initial online elicitation survey followed by a results discussion and a second online elicitation as an opportunity for experts to modify their initial responses based on the discussion. Animal transmission was defined as an illness transmitted by exposure to animals, i.e. personal contact (hand or mouth) with animal/pet feed, animal/pet fur/coats, saliva or feces ( Table 2). The proportion of illnesses from contact with domestic pets, farm animals and wildlife were estimated for non-typhoidal Salmonella spp., Campylobacter spp., Giardia sp., Table 1. VTEC O157 and Y. enterocolitica based on the definitions in Table 2 [37]. As previously described, triangular probability distributions were built from the expert estimates for each transmission route and pathogen, using @Risk software (Version 6·1·2; Palisade Corporation, Newfield, NY, USA) from best estimate (most likely) and 5th and 95th percentile values. These were then combined into cumulative distributions, using Monte Carlo simulation with 10 000 iterations [35,37].

Analysis
The median and 90% credible intervals for the proportion of illnesses due to animal contact and transmission subcategories, by pathogen, as defined by the Canadian expert elicitation, were entered into individual pathogen models as a PERT distribution, where the median and 90% credible interval values from the expert elicitation were used as the inputs for mean and the upper and lower bounds of the PERT distribution. The @RISK add-in for Microsoft Excel was used, with 100 000 iterations to generate a mean and 90% probability intervals (PIs) for estimates.

RESULTS
These eight pathogens account for 528 279 domestically acquired enteric (acute gastrointestinal) illnesses each year in Canada of which an estimate of 84 751 (90% PI 52 952-123 985 were related to animal contact ( Contact with farm animals was the most common cause of animal-associated illnesses, representing almost 42 608 (90% PI 24 219-66 780) illnesses and over half of the illnesses for the five pathogens for which subcategories were estimated (Campylobacter, non-typhoidal Salmonella spp., Giardia sp., VTEC O157 and Y. enterocolitica) (

DISCUSSION
These are the first Canadian estimates of illness attributed to animal contact, accounting for under-reporting and underdiagnosis of illnesses. This study contributes to the development of an understanding of the overall burden of enteric illness in Canada. To date, estimates of foodborne illness [28] and acute gastrointestinal illness associated with drinking water [45,46] have been completed. This is the first study to estimate illness specifically associated with subcategories of animal contact for farm animals, pets and wildlife. These estimates provide a relative comparison of transmission pathways and dominant sources of enteric illness, which is critical when identifying public health priorities, designing effective interventions, and providing evidence to inform policy and regulatory decisionmaking at the local, provincial and federal level in Canada. In addition, burden studies help to identify priority pathogens of concern domestically, and knowledge gaps for further research.
Comparing the same seven pathogens included in the US study [31], more illnesses were associated with animal contact in Canada than in the USA, estimating 209 cases per 100 000 Canadians vs. an estimated 149 cases per 100 000 in the USA [47]. This is likely due to the differences in total illness estimates as well as generally lower proportions of illness estimated to be via animal transmission in the USA, which relied primarily on case-control studies and outbreak summaries to inform the US inputs. The values used for the proportion of illness attributed to animal contact in Canada from the expert elicitation are within the range of values reported in international studies from the USA, Australia and the Netherlands [31,33,34,36] as well as Canadian studies, using surveillance data and reported exposures [32,48,49] (Table 5).
This study highlights farm and animal/pet food exposure as an important pathway for illness transmission. Illness associated with farm animals may occur from occupational exposures, such as Campylobacter spp. infection among workers at poultry-processing plants [50,51], non-typhoidal Salmonella spp. infection following contact with baby chicks [52] or visiting a petting zoo [2]. Reptiles and related feeder rodents may be responsible for a substantial portion of the petrelated Salmonella spp. illnesses estimated [1,7,22,53]. It was estimated in the USA that 6% of all sporadic Salmonella infections may be attributable to reptiles or amphibians [54]. While younger puppies may contribute to Campylobacter infections [55], pet food has also been found to be a concern for exposure to enteric pathogens [56,57].
Canadians' exposure to various animals and pathogen prevalence is an important context to understanding public health risk related to the animal contact transmission route. A Canadian 2015 population study, estimated in the past 7 days that 63·4% of Canadians have 'any contact with animals, animal waste, habitat or food', 6·9% visit a farm or barn, and 1·1% and 1·3% visit any petting zoo or an agricultural fair, respectively [58]. The likelihood of enteric pathogen transmission from household pets may be lower compared with an encounter with farm animals; the higher frequency of pet contact [59] would suggest this as an important potential route of transmission.
The prevalence of Campylobacter spp. has been reported to be approximately 6·5% for petting zoo animals and 24·7% for household pets [60]. Both of these animal sources have a generally lower prevalence compared with FoodNet Canada data from farms for swine (85%), beef cattle (78%) and dairy cattle (79%) [61]. The prevalence of non-typhoidal Salmonella spp. detected in animals on farms (swine, broiler chickens, beef and dairy cattle) by the FoodNet Canada surveillance in sentinel sites across Canada was generally lower than Campylobacter spp. [61].
Less is understood about wildlife-associated transmission; based on our study findings, the burden associated with wildlife for all pathogens is notable and the estimated illness associated with Campylobacter spp. and non-typhoidal Salmonella spp. is considerable. Wild birds may be a primary source of these estimated illnesses as identified as a source of transmission in the United Kingdom (UK) and France [62][63][64]. It is estimated in the UK that about 10 000 illnesses may be associated with wild birds each year [62]. Garden birds, playgrounds and beach sand activities may be environments where wild birds and human behavior intersect causing illness [65,66]. Rural wildlife exposure through hunting is also a potential route of transmission and there is evidence that exposure to deer and wild boars may be a source for shiga toxin-producing E. coli and Y. enterocolitica [67][68][69].
Limitations of these enteric illness estimate models and the expert elicitation study have been discussed elsewhere [28,30,35,37]. General limitations concerning uncertainty of illness estimate models and potential bias of expert elicitations apply (recruitment, elicitation tool, question framing, methodology and analysis) and have also been discussed elsewhere [70][71][72]. These estimates reflect illnesses for the time  period of 2000-2010 and recent changes in rates of illness for pathogens, such as the reported decrease in E. coli O157 cases in Canada in recent years [73] are not reflected in these results and should be considered when interpreting. Sources of information related to the attribution of illness to the animal/pet contact transmission route are limited. Selection of pathogens to be included may have some limitations as transmission of VTEC non-O157, Y. enterocolitica, L. monocytogenes and Giardia sp. via animal contact is less well known. Contact with farm and domestic animals has been identified as a risk factor for VTEC non-O157 [74,75] and Y. enterocolitica, respectively [76]. While the evidence is less clear for L. monocytogenes, transmission from animal to human is plausible, as Listeria has been identified in pet food [77], urban poultry flocks [78] and at least one study identified living on a cattle farm as an increased risk of listeriosis [79]. Other countries have also estimated that a small proportion of listeriosis cases may be transmitted via animal contact [31,33,34] (Table 5). Animal contact transmission of Giardia sp. may be relatively uncommon as current molecular epidemiological data suggests that animals are more often infected with species-specific assemblages that do not cause disease in humans [60,[80][81][82]. Molecular characterization of Giardia sp. in patients in Northern Canada found suggestive zoonotic transmission [83]. Furthermore, animal contact has been implicated in three reported giardiasis outbreaks in a review of Giardia sp. outbreaks in the USA (1971-2011), associated with rabbits at a petting zoo, cattle at a farm and a pet reptile at a long-term care facility [84].
The approach used in this study, which is similar to the US approach [31], assumes that the disease severity and frequency with which cases are underdiagnosed are independent of the mode of transmission. In addition, we estimated the overall pathogen-specific proportion of illnesses attributable to animal contact; the proportion of illnesses attributable to animal contact may vary by age because of the differences in exposures and behaviors [31]. The hospitalizations and death estimates may therefore be conservative, as they do not reflect a potential increased representation of illness in children due to this transmission route. The role of sick vs. healthy animals and immunity are not explored in these estimates but are worth considering in future studies when more data become available to differentiate risks at the individual level.
To further understand transmission dynamics and the burden of illness to specific animals, case-control studies for key pathogens to identify specific higher risk animals/settings and risk behaviors facilitating transmission would be beneficial. Additionally, assessment of exposure frequency among Canadians, studies to collect data on pathogen prevalence, concentration and subtyping in relevant animal populations and mechanisms for pathogen reduction for some pets (e.g. reptiles) would further support the understanding of the role animal contact plays in enteric illness transmission and burden.
The burden of enteric infection associated with animal contact is considerable and emphasizes the need for prevention activities. Enhanced awareness and education for the public, farm/occupational workers and pet owners about the potential risk of illness associated with animal contact is key to preventing animal contact-associated illnesses. This includes highlighting the importance of preventative behaviors through consistent messaging at veterinarian offices, pet stores, petting zoos and other venues and reinforcing their role in preventing illness (e.g. recommendations identified in Reducing the Risk of Pet-Associated Zoonotic Infections [3]) and broader communication through websites promoting safe pet ownership, such as the Worms and Germs Blog (http://www.wormsandgermsblog.com) [85] and Healthy Animals, Healthy People [86]. Public health guidance (e.g. petting zoo infection prevention guidelines, recommendations for high-risk populations) and implementation of public health interventions (e.g. enhanced awareness, hand-washing stations) will also contribute to reducing the burden of enteric illness associated with animals. The evaluation of best practices and identification of the most effective prevention activities to reduce disease are required [3].