Background
Salmonellosis is a major foodborne bacterial infection that continuously poses a significant human health burden worldwide [
1]. In Canada, salmonellosis is the main cause of hospitalization and death among domestically acquired foodborne infections [
2], causing an estimated 87,510 illnesses annually [
3]. In the last decade,
Salmonella enterica serotype Enteritidis (
S. Enteritidis) became the top serovar among the non-typhoidal salmonellae in Canada [
4], the United States of America (
US) [
5,
6], and the European Union [
7].
Currently in Canada, the predominant
S. Enteritidis phage types (
PTs) among human cases are PT 8, 13a, 13, 1, 4, and 5b [
4]. Between 2006 and 2010, Canadian integrated surveillance systems identified the emergence of PT 13a and an increase in the number of cases of PT 8 [
4].
Several research studies conducted in North America have evaluated phage type-specific risk factors for
S. Enteritidis infections in humans. In Ontario, Canada, researchers demonstrated that cases with PT 8 were more likely to have had contact with dogs compared to cases with other phage types [
8]. In British Columbia, Canada, a concurrent increase in the incidence of
S. Enteritidis infections with PT 8 in humans and the prevalence of PT 8 in poultry was observed between 2007 and 2010 [
9]. The researchers demonstrated increased odds of infection with PT 8 in human cases who consumed illegally-sourced ungraded eggs compared to controls [
9]. In Alberta, Canada, an outbreak of PT 8, 13, and atypical PTs was linked to the consumption of food products purchased from mobile lunch trucks that were contaminated by illegally-obtained eggs and/or by infected food handlers [
10]. In the US, PT 8 cases were more likely to have consumed chicken or be the owner of a lizard than controls, whereas PT 13 cases were more likely to have eaten undercooked eggs in their home than controls [
11].
In Ontario and Canada, an increase in the reported number of human
S. Enteritidis cases was observed during the last decade [
12,
13]. Current Ontario studies revealed that the majority of
S. Enteritidis cases with PT 1, 4, or 6a were international travel-related, whereas cases with PT 8, 13, or 13a were mainly acquired domestically [
14,
15]. These studies provided valuable information on the seasonality and exposure locations of
S. Enteritidis cases, although they lacked information on cases’ geographical distribution and spatial-temporal clustering. Identifying areas with high rates of reported
S. Enteritidis cases can be useful for targeting prevention and control programs [
12,
16].
There have been a limited number of studies that evaluated foodborne disease surveillance data by incorporating geographical information system (GIS) data, spatial-temporal scan statistic results, exposure setting information, and clinical syndrome history. Scan statistics have been effectively used to evaluate clustering and transmission dynamics of pandemic influenza A (H1N1) in Hong Kong, China [
17], to detect
Escherichia coli O157:H7 outbreaks involving common molecular subtypes in Alberta, Canada [
18], to identify the location of high and low rate areas of campylobacteriosis incidence in Manitoba, Canada [
19], to identify high incidence clusters of tuberculosis in Linyi City, China [
20], and to find childhood cancer clusters in Alberta, Canada [
21].
This study assesses the spatial and temporal epidemiology of the phage types of S. Enteritidis that predominate in Ontario health regions by: 1) estimating phage type-specific health region-level incidence rates (IRs); 2) estimating phage type-specific monthly IRs; 3) describing phage type-specific exposure settings and clinical symptoms; 4) detecting phage-type-specific spatial, temporal, and space-time clusters of cases; and 5) examining the exposure settings of cases identified within space-time clusters. The results of this study are expected to assist public health officials with the development of disease prevention programs within the province.
Discussion
Our study enhanced the current knowledge on the spatial and temporal epidemiology of the phage types of S. Enteritidis that predominate in Ontario health regions. We used a step-wise approach, starting with a general exploratory analysis followed by a more specific statistical analysis. A number of phage type-specific high rate areas and time periods were identified during the exploratory analysis that were confirmed by the statistical analysis as significant spatial, temporal, or space-time clusters of cases.
Foodborne disease clusters are generally defined as the occurrence of a higher than expected number of cases for a given location and/or time period. These clusters may or may not meet the definition of an outbreak [
32,
33]. Subtype-based surveillance systems frequently use the term “cluster” to describe a group of cases infected with identical microbial strains [
32]. Subtyping is useful for differentiating between endemic and outbreak cases, especially for common
Salmonella serotypes, such as Enteritidis, that occur sporadically throughout the year [
34]. Differences in reservoirs and exposure settings might exist for different
S. Enteritidis phage types, and molecular differentiation can help to understand potential sources of the different phage types [
8,
34]. We defined a cluster as a health region, time period, or a health region during a particular time period with a statistically significant higher than expected phage type-specific
S. Enteritidis infection rate. Thus, we demonstrated the effectiveness of using cluster detection tests, in conjunction with subtyping methods to understand the epidemiology of a foodborne pathogen.
A number of patterns were observed when assessing the geographical heterogeneity of health region-level IRs of S. Enteritidis infections for the most frequent phage types. The Central West region had the highest IRs for PTs 1, 4, and 5b, whereas the Toronto region had the highest IRs for PTs 13 and 13a. Several of these regions were later confirmed by the spatial scan statistic as regions with significant high rate clusters (e.g., cases of PT 5b significantly clustered in the Central West region and cases of PTs 13 and 13a significantly clustered in the Toronto region).
We used a smoothing method for our time-series graph to reduce the month-to-month random variation of infection rates and make the overall trends clearer. The observed trends were relatively consistent with the results of the purely temporal scan statistic, albeit not as definitive. With the exception of PT 5b, all temporal clusters occurred during 2008. Further, most clusters occurred during a distinct season. Cases of PTs 1 and 5b clustered during the winter months, cases of PT 13 clustered during the spring months, and cases of PT 13a clustered during the summer and fall months. Differences in the duration of the temporal clusters were also observed. The majority of clusters (PTs 1, 5b, 13, and 13a) were of relatively short duration (2–4 months), whereas the PT 4 cluster was of long duration (11 months). Of note, the most commonly reported phage type (PT 8) did not cluster temporally, suggesting a fairly even distribution of PT 8 cases over time throughout Ontario. A study conducted in Alberta, Canada, examining
Salmonella serotypes rather than phage types, detected several serotype-specific temporal clusters during the 11-year study period (January 1990 to January 2002) [
35]; for
S. Enteritidis, the clusters were of short duration and occurred during the winter and spring months.
The exposure setting information is rarely confirmed by data obtained through environmental health investigations or statistical associations obtained through case-control or cohort studies [
36]; however, it is considered to be useful epidemiological data for foodborne illness source attribution [
37]. Knowing when, where, and why clusters occurred can aid in the development of effective outbreak detection, prevention, and control programs. Our study identified differences between phage types with respect to the time and duration of the space-time clusters, even for clusters occurring in the same region. For example, the PT 13 and 13a clusters both occurred in the Toronto region, but during different time periods (the cluster of cases with PT 13 occurred in 2008, whereas the cluster of cases with PT 13a occurred in 2009). Moreover, the cluster of cases with PT 13 was of long duration (7 months), whereas the cluster of cases with PT 13a was of short duration (3 months). Short duration clusters might signify that cases were exposed to a single infection source (e.g., point source outbreak). Long duration clusters might signify that cases were exposed to a single source (e.g., contaminated food) over a longer time period (e.g., continuous common source outbreak) [
32,
33,
35], to multiple sources (e.g., continuous multiple source outbreak) [
32,
33], to the occurrence of secondary infections [
35], to poor food preparation practices over a prolonged period, or that the typing method used was not of high enough resolution to differentiate between different strains.
Many of the cases with PT 13 or 13a that were part of a space-time cluster reported food premises (e.g., restaurant, grocery store, bakery, deli, caterer, mobile food premise) as their main exposure setting. In North America, restaurants have been shown to be an important exposure setting for
S. Enteritidis infections [
38‐
41]. A number of predisposing factors for food contamination with
S. Enteritidis in restaurants were identified, including cross contamination from raw chicken meat to food server’s hands or cutting boards due to high food volumes and food handler’s improper food safety practices during food preparation [
36,
38], inadequate heat treatment of foods [
38], inappropriate food storage [
38], and direct contamination of food served by infected food handlers [
10,
39‐
41]. In Ontario,
S. Enteritidis accounted for only 10.1 % of the
Salmonella isolates collected at pre-harvest from conventionally-raised broiler chicken flocks between July 2010 and April 2012; 65 % of the isolates were PT 13a (Tara Roberts, 2014, personal communication).
A few of the cases that were part of a PT 8, 13, or 13a space-time cluster reported private homes as their exposure setting. Previous studies identified private homes as an important exposure setting for sporadic, home-based foodborne infections [
42‐
44]. Several predisposing factors of home-based infections have been identified, including inappropriate food handling, storage, and food preparation [
42,
43]; consumption of contaminated raw and undercooked foods [
42]; and person-to-person [
44] and animal-to-person [
45,
46] transmission.
Space-time clusters of cases with PT 1 or 4 included several overlapping health regions, occurred during nearly identical winter and spring months, and were of short duration (2-3 months). The majority of these cases reported international travel as their exposure setting. International travel was demonstrated by a number of studies as an important risk factor for
S. Enteritidis infections in North America [
15,
47,
48]. In the US, among all salmonellosis cases between 2004 and 2008, 11 % reported international travel as their exposure setting, and among those, the most commonly reported serotype was Enteritidis (22 % of travel cases) [
47]. In the region of Waterloo, Ontario, Canada, between June 2005 and May 2009, 48.7 % of
S. Enteritidis cases were international travel-related [
48]. In Ontario, Canada, between July 2010 and June 2011, 51.9 % of
S. Enteritidis infections were international travel-related, and certain phage types (e.g., 1, 4, and 5b) were isolated from cases who visited all-inclusive resorts in the Caribbean or Mexico during the winter and spring months [
15]. The seasonal spike of PT 1 and 4 cases in late winter and early spring, when people often travel to warmer destinations, warrants creating advisories to inform travelers about the risks of eating abroad and how they can protect themselves against
S. Enteritidis infections.
A number of limitations should be recognized before interpreting our study results. Surveillance programs underestimate the true burden of infections due to under-diagnosis and under-reporting of cases [
3]. In Canada, it was estimated that for every reported salmonellosis case there were 26.1 unreported cases in the general population [
3]. Under-reporting and under-diagnosis can be influenced by differences in populations’ medical care seeking behaviour and access to medical care [
49], physicians’ specimen request and diagnosis practices [
50], and laboratories testing protocols and reporting standards [
50]. Regional differences in successful case follow-up should also be considered. Loss to follow-up of cases might be greater in low population density regions of the province due to difficulties encountered by public health staff in contacting cases. A large number of cases had missing or unknown exposure setting information, which might have biased our study results. The proportion and accuracy of known exposure setting information reported by investigators can depend on several factors [
36], including time passed from exposure to case interview and the related recall bias, difficulty and the effort made by the investigator to contact a case, follow-up protocol and questionnaire used by the investigator (e.g., face to face interview vs. phone interview vs. questionnaire sent through the mail), a case’s willingness to be interviewed, and possible survival bias. In our study, differences in unknown exposure setting among phage types were noted. The proportion of unknown exposure setting information was higher for cases with PT 8, 13, or 13a (64–66 %) compared to cases with PT 1, 4, or 5b (22–28 %), suggesting that international travel cases had more readily available exposure history; therefore in our study, the overall proportion of cases who reported international travel as their major exposure setting was likely slightly over-estimated. Lastly, misclassification of international travel-related cases might have also occurred, especially for cases for which the incubation period was short, and for cases with a longer disease incubation period who became infected before departure [
48].
Obtaining exposure setting information is a first step toward developing effective prevention and control programs; however, the location and the primary source of contamination of food products that lead to infections are not always identical [
51-
52]. Therefore, future research studies are needed to identify the primary source of contamination, and the type of food products that cause infections.
This study demonstrated the utility of retrospective spatial and temporal analysis of subtype-based surveillance data using exploratory and statistical methods to detect clusters of cases. Phage type-specific spatial and spatial-temporal clusters should be followed up by public health authorities to identify novel local individual-level risk factors. Increased enforcement (e.g., restaurant inspections) and education (e.g., food safety training for restaurant employees and the general public) in health regions with identified spatial or spatial-temporal clusters have the potential to decrease the incidence of PTs 8, 13, and 13a. Further, prevention programs (e.g., travel advisories) that are targeted during the winter and spring months have the potential to decrease the incidence of PTs 1, 4, and 5b. During the study period no outbreaks were reported in Ontario; thus, the evaluation of current outbreak detection methods used by public health staff at various PHUs is warranted. Future studies are needed to evaluate the frequency of false positive clusters, to assess the effectiveness of cluster detection using statistical methods, to compare the more traditional outbreak investigation procedures to scan statistic cluster detection techniques, and to measure the feasibility of statistical methods for identifying infection clusters. Purely spatial or purely temporal clusters might be the result of a space-time cluster, which should be considered when evaluating our study results. There is a need also for prospective research studies to identify clusters of S. Enteritidis infections in real-time (e.g., weeks, months), and to assess and evaluate individual-level risk factors for infections included in these clusters. Moreover, there is a need for high resolution molecular subtyping methods (e.g., multiple locus variable-number tandem repeat analysis or whole genome sequencing) to better understand relationships between cases in a cluster.
Authors’ contributions
CV developed the study design, analysed the data, interpreted results, wrote the first draft of the manuscript, responded to editorial comments, and prepared the final manuscript for submission. MTG was consulted on data analysis, study design, interpretation of results, and reviewed and commented on manuscript drafts. DLP, SAM, FP, and JMS provided advice on the data analysis, interpretation of results, and reviewed and commented on manuscript drafts. All authors read and approved the final manuscript.