Findings

Epidemiologic investigations, including reports from the CDC's FoodNet, have demonstrated higher incidences of Salmonella infections in children compared to adults [1, 2]. In 2003, the incidences of laboratory-confirmed cases of Salmonella infections reported at FoodNet sites were 122.7 cases per 100,000 population for children aged < 1 year, 50.6 cases per 100,000 population for children aged 1-4 years, and 10.8 cases per 100,000 population for those aged ≥ 5 years [3]. During 2002-2006, Michigan Department of Community Health (MDCH), reported 63.96 cases of salmonellosis per 100,000 population for children aged < 1 year, 12.62 cases per 100,000 population for children aged 1-9 years, and 8.45 cases per 100,000 population for aged ≥ 10 years [4]. One of the reasons of higher incidence of salmonellosis cited in the literature is the increased case detection of Salmonella infections in children compared to the adult population. In young children with gastrointestinal symptoms, increased detection is likely due to: 1) the parents will seek medical attention and 2) the healthcare provider will more likely submit a sample for culture [5].

The majority of sporadic cases of salmonellosis in the adult population results from consumption of contaminated food and traveling to areas where Salmonella is more prevalent [6]. However, risk factors for salmonellosis in children have not been extensively evaluated in population-based epidemiologic studies. The majority of the studies in children have been conducted in response to salmonellosis outbreaks, particularly in daycare centers and nurseries. The risk factors identified in outbreaks may not be similar to those of sporadic cases of salmonellosis since exposures in sporadic cases vary widely. A few studies have identified eating undercooked eggs [7], infant formula [1], attending a day care with a child with diarrhea [1], household member with a gastrointestinal (GI) symptoms [7], and contaminated home environment [8] as risk factors for Salmonella infections in children. It has been suggested that contaminated environmental sources contribute more than contaminated food vehicles in the acquisition of Salmonella infections in children [7]. This report describes a population-based case-control study conducted to identify potential risk factors for sporadic, non-typhoidal Salmonella infections in Michigan children aged ≤ 10 years.

Methods

Study design and subjects

A case-control study was conducted between December 15, 2006 and October 15, 2007 among Michigan children aged ≤ 10 years. Salmonellosis is a notifiable disease under the Michigan Communicable Disease Rules [9]. Therefore, physicians and laboratories across Michigan are required to report cases of salmonellosis to their local health department (LHD) in either the jurisdiction where the individual with suspected salmonellosis resides or where the reporting facility is located (passive surveillance system). In addition to reporting to the MDCH, LHDs and clinical laboratories send isolates to the MDCH Bureau of Laboratories for serotyping and the results are entered into the state wide Michigan Disease Surveillance Systems (MDSS).

In this study, cases were defined as children aged ≤ 10 years with laboratory-confirmed Salmonella infections with permanent Michigan address. Case children were excluded from the study if the case 1) was reported as part of a salmonellosis outbreak investigated by Michigan Department of Community Health (MDCH) officials and 2) had a reported congenital malformation (e.g., birth defect), serious medical condition, or a concomitant infection.

Controls were children aged ≤ 10 years who were not diagnosed with any enteric infections (e.g., salmonellosis, campylobacteriosis) by a healthcare provider and did not experience any enteric disease symptoms (e.g., diarrhea, vomiting, nausea) during the 30 days prior to the interview day. Controls were excluded if the child had a congenital malformation or serious medical condition. Controls were enrolled by two methods: 1) case parent(s) were asked to identify a child of similar age to their own child and 2) using the on-line telephone directory http://www.whitepages.com. This directory has a reverse address function that allowed the compilation of lists of potential control household phone numbers. Households were called, and after explaining the study's objective, we asked if there were any children aged ≤ 10 years living in the household. Consented parents or caretakers of these children were interviewed.

Questionnaire and data collection

Data was collected using a structured questionnaire on sociodemographic characteristics (e.g., age and sex, household income, parental education), child feeding practices (e.g., breast feeding, formula milk, use of pacifier), child rearing (e.g., daycare, pre-school, or elementary school attendance), food exposures (e.g., consumption of milk, meat, poultry, eggs, produce, water) and various environmental exposures (e.g., contact with animals, contact with a person having GI symptoms). The questionnaire is presented under additional file 1.

Exposures were assessed during the 3 days preceding the illness onset date for cases, and the 3 days prior to the interview day for controls. The 3 days exposure experience encompasses the common incubation period for salmonellosis and has been used in other similar studies [10, 11]. Parents or caretakers were given the option to fill out a self-administered mail-in questionnaire or participate in a 15 to 20 minute phone interview with a trained interviewer. The questionnaire was pilot-tested on volunteer parents, issues identified in this exercise were addressed, and appropriate changes were incorporated into the final version of the questionnaire. An informed consent was obtained from parents or caretaker prior to administering the questionnaire. The study protocol was reviewed and approved by the IRBs at Michigan State University and MDCH.

Data management and statistical analysis

An age-matched (<1 year, 2-<6 years, and 6-10 years) pair univariable analysis of cases and controls using the conditional logistic regression was performed to examine associations between Salmonella infections and hypothesized risk factors. A multivariable analysis was conducted to identify factors associated with Salmonella infections, simultaneously controlling for potential confounders including race and household income, and matched odds ratio (MOR) and respective 95% confidence interval (95% CI) was computed for each predictor variable. All variables with a p values < 0.25 on univariable analyses, along with those hypothesized risk factors for Salmonella infections, were initially considered for inclusion in the multivariable model. After identifying the correlated variables, we examined the impact of collinearity by separately entering the variables (e.g., income, education, and race) into the multivariable regression model. We used the backward elimination method to obtain a parsimonious but yet plausible model. Additionally, to investigate potential effect modification, we also included several two way interaction terms in the model, 'household income and race,' 'household income and reptile ownership,' and 'and reptile ownership. A p-value < 0.05 was considered statistically significant in all analysis. SAS was used to analyze the data.

Population attributable risk (PAR) for selected variables were estimated from the final multivariable model using Levin's formula for the calculation [12].

Results

A total of 123 cases representing a participation rate of 72.78% (123/169) were enrolled. A total of 139 control children were enrolled using one of the following two methods. Additional files 2 and 3 describe the enrollment of cases and controls respectively and list the number and type of exclusion criteria that were encountered during this study.

The majority (85%) of the cases were interviewed within 20 days from the date when their records were available in the MDSS database. Control child was enrolled in parallel with cases enrollment however the cases were not individually matched with controls on the time of enrollment.

Comparison of the enrolled cases and controls is presented in Table 1. The enrolled cases and controls did not differ by socioeconomic and demographic characteristics including parental education, and annual household income except for distribution of racial composition (p < 0.01).

Table 1 Socioeconomic characteristics of children aged ≤ 10 years enrolled in a population-based case-control study to identify risk factors for Salmonella infections, Michigan, 2007.
Full size table

Tables 2 and 3 show the results of univariable and multivariable logistic regression models respectively. The final multivariable model after adjusting for race, household income, and other known risk factors for salmonellosis [contact with a person having GI symptoms and other food related exposure (eating egg/meat/poultry) in the past 3 days] revealed that Salmonella infections were significantly associated with attending a daycare center (MOR = 5.00 (95% CI: 1.51-16.58), contact with cats (MOR = 2.53 (95% CI: 1.14-5.88), and contact with reptiles (MOR = 7.90 (95% CI: 1.52-41.01). None of the two-way interactions tested (household income and race, household income and reptile ownership, and race and reptile ownership) were statistically significant (p > 0.05). Reptile exposure had the highest PAR% (35.61) among the risk factors identified for salmonellosis in our study.

Table 2 Univariable analyses of putative risk factors for Salmonella infections in children aged ≤ 10 years, assessed in a population-based case- control study, Michigan. 2007.
Full size table
Table 3 Multivariable analysis of putative risk factors for Salmonella infections in children aged ≤ 10 years, assessed in a population-based case-control study, Michigan, 2007.
Full size table

Discussion

We found that cases of laboratory-confirmed salmonellosis were associated with several potentially modifiable risk factors and thus findings of this study can be used for enhancing salmonellosis prevention efforts in Michigan. Concurring with other investigations, our data suggest that children attendance of a daycare significantly increases the odds of Salmonella infections. Infections with several enteric agents have been associated with attending daycare [8, 12, 13]. The spread of infections in daycare centers is facilitated by crowding and microbial contamination of the environment [14]. Simple measures to control and prevent infections such as washing hands with soap and water after changing diapers, after assisting children with the toilet, and before handling food would help to substantially reduce the incidence of infections related to daycare.

We have found that contact with animals including reptile was associated with significant risk for salmonellosis in Michigan children. Salmonellosis is a well-recognized bacterial zoonoses [15]. Animals are the predominant reservoirs for the bacteria, and the prevalence of Salmonella carriage varies by species [16]. An estimated 90% of all reptiles including turtles and iguanas, carry and shed Salmonella in their feces intermittently [17]. Salmonella serotypes have been isolated from most vertebrates including dogs and cats with carriage rates of up to 36% and 18%, respectively [18]. However, a much higher (up to 94%) Salmonella carriage rate has been observed in reptiles and amphibians [19]. In a recent case-control study conducted using FoodNet sites, reptile ownership was associated with more than 5-fold increased risk of salmonellosis in children aged < 1 year [1].

In contrast to other similar studies [20] that reported an association between reptile-associated Salmonella specific serotypes (e.g., S. Stanley, S. Poona, S. Litchfield) and contact with reptile, our study did not find such association. This could be because of the limited sample size when the data were stratified by serotypes. We did not find significant differences in the frequency of Salmonella serotypes isolated from case children exposed to only reptiles or only cats. A recent study on Salmonella serotypes in captive reptiles reported a high diversity of Salmonella serotypes among reptile, which did not include serotypes Stanley, Poona or Litchfield [21]. Additionally, the MDCH registry-based study classified 59 isolates of Salmonella serotypes associated with children cases to be reptile-associated. The classical reptiles-associated Salmonella s erotypes were not reported [22]. In this study, we found serotype Enteritidis and Typhimurium, which are frequent contaminants of chicken and meats, to be common among case children with reported contact with reptiles or cats. These findings are consistent with the previous report [22]. This assumed cycle of transmission may have resulted from foods contaminated with these serotypes that was fed to these animals [21].

Food-related exposures such as consumption of chicken, meat, and eggs/egg-containing product (established risk factors for salmonellosis in adults) did not show a significant association with Salmonella infections. It is conceivable the actual magnitude of food exposure related risk for Salmonella infections was not demonstrated in this study due to several factors. First, we obtained exposure information from interviewing surrogate sources (i.e., parents or caretakers). Secondly, certain exposures such as contact with animals are more likely to be accurately recalled, but the consumption of specific foods is more difficult to recall and thus prone to measurement error.

Exposure to contaminated environment was more commonly reported than exposure to contaminated food vehicles as a possible source for Salmonella infections in children [7, 23]. Results of this study are in agreement with the findings of a case-control study conducted in the UK, where none of the household-related variables showed significant associations with Salmonella infection [24]. A recent study reported that children aged < 1 year who ride in a shopping cart with meat or poultry placed next to them have a 4-fold increased risk for salmonellosis [1]. Although some studies [8, 24] have reported the isolation of Salmonella from household dust, soil samples near the home, and samples from bathrooms, we have not microbiologically evaluated contamination of the household environment.

This was the first population-based case-control study in Michigan conducted to identify risk factors for salmonellosis in children. Although not all reported cases were enrolled in the study, participants and non-participants were drawn from the same base population, and the two groups were similar with respect to demographic characteristics (p > 0.05). Our data were limited to laboratory-diagnosed cases only, and thus are biased (reporting bias) by factors that affect the probability of an illness being reported; caretakers of case children with mild to moderate disease symptoms may not seek medical care (selection bias) [5]. Further, recall bias may be present in the measurement of some exposure variables. Parents knew the disease status of their children prior to the interview and this may have influenced their responses [25]. We did not obtain specimens for culture from controls to exclude asymptomatic cases of salmonellosis. It is possible that some of our control children were asymptomatic Salmonella carriers and thus would have been misclassified. However, given the very low (1%) prevalence of chronic carriers of Salmonella in healthy populations, we expect minimal misclassification due to this phenomenon [26].

Conclusions

This study revealed that attending a daycare center and contact with cats and reptiles within three days before a child's illness onset are important risk factors for infection with Salmonella serotypes in Michigan children. Data suggest that the contribution of environmental sources plays an important role in the acquisition of Salmonella infections in children. Further efforts are needed to educate parents and caretakers about the risk of Salmonella transmission to children from cats and reptiles, in addition to individuals having GI symptoms.