Overall, the SES category with the most persons in Connecticut was the highest SES group, <5% below the poverty line (55.2% of the population during 2000–2005 and 48.9% during 2006–2011). The lowest SES group, >20% below the federal poverty level, was the smallest for both periods (10.5% and 12.4% of the population, respectively).

Few studies are available that examine the relationship between foodborne disease incidence and socioeconomic status. Our study showed the following key findings: 1) STEC disease, whether caused by O157 or non-O157 serotypes or whether manifested as HUS, was uniformly associated with high census tract SES; 2) salmonellosis in persons 5–9 and >10 years of age was associated with high census tract SES, whereas salmonellosis in children <5 years of age was associated with low census tract SES; and 3) different Salmonella serotypes had different associations with census tract SES, with serotypes Enteriditis, Newport, and Montevideo associated with high census-tract SES, serotype Heidelberg associated with low census-tract SES, and serotype Typhimurium having no association. These findings provide additional information about the epidemiology of these foodborne diseases and should inform efforts to reduce their incidence.

Our findings study are similar to those found in a study of campylobacteriosis in Connecticut during the same period (7): higher disease incidence among those living in higher SES census tracts. A previous study in Michigan found that Salmonella incidence increased with higher education and income levels (9). An analysis of recent FoodNet data by race/ethnicity showed that overall STEC rates were highest for whites (a surrogate for higher SES) and lowest for blacks (a surrogate for lower SES) (16). The current study results were consistent with results from these studies that examined surrogates for SES.

Several explanations have been suggested to explain why persons in higher SES census tracts might have higher incidence of Salmonella and STEC infections and HUS (and Campylobacter infection), compared with persons in lower SES census tracts. A commonly proposed reason is that those living in areas with higher census tract–level SES might have increased access to care and might be more likely to submit specimens, regardless of disease severity, whereas those in lower socioeconomic groups might seek care or diagnostic testing only when illness is serious or prolonged (8,9,17). Several lines of evidence argue against this explanation in the United States. For 2000–2003, FoodNET assessed factors associated with seeking medical care and submitting a fecal specimen among persons with acute diarrheal illness and found that ≈20% of persons with acute diarrheal diseases sought medical care, 19% of whom submitted a fecal specimen (18). The analysis found that a household income <$25,000 was associated with seeking medical care (18). This association of lower income with seeking medical care for diarrheal illness, the uniform trend of increasing incidence from lowest to highest SES group, and the opposite association for some Salmonella serotypes indicate that medical-seeking behavior is not a major explanatory factor for our results. Furthermore, HUS, a disease almost always requiring hospitalization and thus less subject to potential health-seeking bias, had the same association with higher SES as did milder forms of STEC infection.

The more likely explanation for these findings is that SES affects the prevalence of known high-risk factors, such as international travel, consumption of high-risk food items (e.g., raw fruits and vegetables and undercooked meat), and eating at restaurants (16). That is, high SES itself is not a risk factor but rather a surrogate for certain high-risk behaviors. An analysis of the Connecticut portion of 3 FoodNet population surveys during 2000–2007 found that higher-income residents were more likely than lower-income residents to have traveled internationally, eaten in restaurants, and eaten chicken in the previous 7 days (17). An analysis of population survey data from the 2006–2007 Connecticut FoodNet found that residents with higher SES ZIP codes were more likely than those with lower SES ZIP codes to have eaten fresh hamburger at home that was pink, to have consumed salad containing lettuce or greens, and to have traveled internationally in the previous 7 days (J. Wagner, unpub. data). Studies outside the Connecticut FoodNET have had similar results. Several studies have found that contaminated raw fruits and vegetables are a growing source of outbreaks in the United States and have increased in both numbers and proportions of all reported foodborne outbreaks (19,20). A recent food attribution study attributed 32% of all bacterial foodborne illnesses, of which Salmonella and STEC make up a large proportion, to produce commodities, including fruits, nuts, and vegetables of the fungi, leafy, root, sprout, and vine-stalk variety (21). High SES is associated with more fruit and vegetable consumption, which may be an explanatory factor for our findings. A study in the United States found that higher neighborhood SES was positively associated with fruit and vegetable intake and that an increase of 1 SD in neighborhood SES was associated with 2 additional servings of fruit and vegetables per week (22). Low SES communities often have access to fast food and prepackaged food but lack adequate supermarkets, which causes limited access to fresh fruits and vegetables (23,24). In addition, ≈40% of adolescents from low SES backgrounds have less than daily consumption of fruits and vegetables (25). A study addressing fruit and vegetables as vehicles for transmission of foodborne pathogens found that the association of Salmonella with fresh produce appears to be serotype-specific because of adhesion mechanisms in some serotypes (26), a finding that may partly explain why the association of higher SES with higher incidence was seen only among some serotypes of Salmonella.

In contrast to the findings in adults, children <5 years of age who live in low SES census tracts were more likely than those living in high SES census tracts to have Salmonella infection. In addition, all persons with S. enterica ser. Heidelberg infection were more likely to live in a low SES census tract. These findings are novel: the Michigan study did not look at age, and no reported studies in the United States have systematically examined the relationship between census tract–level SES and Salmonella serotype incidence. However, the previously published Connecticut campylobacteriosis study had similar findings: children living in low SES census tracts had the highest incidence (7,27). Several studies have shown that young children in low SES circumstances are more likely to be exposed to raw meat–contaminated surfaces inside and outside the home (e.g., in shopping carts in grocery stores) (16,17). Also, different Salmonella serotypes are associated with different food items (21). S. enterica ser. Heidelberg has been associated with the consumption of eggs and poultry (28–30), and a study assessing Salmonella prevalence in 6 commodities at point of processing found high prevalence of S. enterica ser. Heidelberg in chicken and turkey (31). Possibly, more eggs and poultry are consumed in lower SES groups than in higher SES groups.

Our findings have several implications for risk communication and research. Efforts could be made to increase awareness among persons in high SES groups about their relatively high risk for STEC and Salmonella infection and about what actions they can take to reduce risk. Education campaigns about high-risk foods other than meat and about the importance of properly handling produce could be run in publications with a higher SES target audience. Regarding research, several considerations need to be explored further. Whether our findings in Connecticut are representative of STEC and salmonellosis incidence nationwide is unknown. The specific reasons behind SES-related risk for foodborne illness still need to be made clear, including for each of the leading Salmonella serotypes. For young children with salmonellosis, potential intervention points to reduce exposure inside and outside the home need to be identified. It remains unclear why children living in areas of higher SES are more at risk for STEC infection and HUS than children living in lower SES areas. These unanswered questions need to be investigated so that effective consumer-level interventions can be developed.

This study has several limitations. First, it was limited to data from Connecticut, and findings might not be generalizable to other states. Second, SES information was unavailable for 7 census tracts; however, they accounted for <1% of all census tracts in Connecticut. Third, although the data came from active surveillance and are therefore complete within the context of laboratory-confirmed disease, not all persons with gastrointestinal illness seek medical care or have diagnostic testing (1,18), so these data are underestimates of the true incidence of STEC and salmonellosis. Fourth, this study used SES measured at the census tract, not individual level; the results of this analysis should be understood and interpreted within this context. Finally, we did not look at the consistency of the associations of incidence with SES for different adult age groups but assumed they were similar.

Public health infrastructure objective 7.3 of Healthy People 2020 stresses the need to “increase the proportion of population-based Healthy People 2020 objectives for which national data are available by socioeconomic status” (32). This study provides additional evidence that area-based SES measures such as census tract–level poverty, especially if other SES measures are unavailable, can be useful for describing the incidence of foodborne illnesses. Our findings show differences in STEC and Salmonella serotype-specific incidence by census tract–level SES. The findings suggest direction for risk communication efforts and for additional studies to explain the differences and to facilitate additional education and outreach activities.