Discussion

In this cross-sectional observational analysis of a dataset of all STEC cases in England reported to NESSS over a 6-year period, the crude incidence of STEC was significantly lower in the most disadvantaged quintile, compared to the least disadvantaged. There were differences in reported exposure to known risk factors by SES; with lower odds of reporting exposure to: salad, fruit, vegetables or herbs; private water supplies; recreational freshwater; walking in a paddock; day trips; contact with soil; UK travel; and non-UK travel amongst the most disadvantaged group. Lower odds of reported exposure to fish/shellfish were also identified. Although this is asked in the ESQ, it is not widely regarded as a risk factor for STEC and likely to reflect incidental differences in population-level exposure by SES. There were also differences in healthcare contact by SES, with lower odds of reporting contact with GP and higher odds of reporting visiting A&E or being hospitalised amongst people living in the most disadvantaged areas.

Two previous studies, in the USA and Finland, found a lower incidence of STEC in communities with lower education levels and hypothesised that this could relate to poorer food safety practices and differential food consumption habits amongst more highly educated individuals [Reference Chang9, Reference Jalava11]. However, the Finnish study also identified an increased incidence of STEC associated with the proportion of low-income households with children [Reference Jalava11], indicating that the relationship between SES by age is important for GI infections. The results of our study support the inverse association of STEC incidence with deprivation found in the USA and Finland, rather than positive associations reported in Japan or a lack of association reported in Canada and Denmark [Reference Pearl14, Reference Simonsen, Frisch and Ethelberg15]. Our study is also consistent with the suggested difference by age reported by the Finnish study (although the positive relationship between incidence and SES in 0–4 year olds in our study was not statistically significant).

In other studies which found higher risk of GI pathogens other than STEC in socio-economically advantaged groups, the authors hypothesised that the finding could relate to: differential exposure, such as consumption of higher risk food amongst those of a higher SES [Reference Whitney10, Reference Simonsen, Frisch and Ethelberg15, Reference Patil, Cates and Morales24–Reference Younus27]; more foreign travel; greater exposure to farm animals or the rural environment. Alternatively, the higher observed risk in high SES groups could be related to differential healthcare-seeking behaviour practices [Reference Simonsen, Frisch and Ethelberg15, Reference Younus27]. In our study, there was evidence for differences in non-UK travel, rural environmental exposures and healthcare contact, which would support these assertions. However, there was no evidence to suggest differences in consumption of known foodborne risk factors or greater direct exposure to animals. This may suggest that different risk factors, such as increased opportunities of person-to-person spread, may be more important in disadvantaged groups; a suggestion that has also been postulated by others [Reference Newman8]. This hypothesis warrants further investigation in future studies.

One previous study explored the social patterning of a limited number of potential risk factors for STEC (consumption of hamburgers, type of water available for consumption, recreational water exposure and personal hygiene) in the general population in Argentina [Reference Bentancor28] as opposed to risk factors reported by cases, as in our study and results are therefore not comparable.

In our study, we observed significantly lower odds of reported non-UK travel amongst disadvantaged STEC cases. It is possible that the pattern of lower odds of reporting non-UK travel amongst more disadvantaged groups reflects the population-level distribution of non-UK travel. Since travel is a known risk factor for STEC infection [Reference Parry5, Reference Byrne17], lower exposure may confer a lower risk of STEC infection.

There is also evidence to suggest that perceptions of attractiveness and safety of local walking environments may differ by SES. Those in more disadvantaged areas may perceive that their local walking environment is less attractive and less safe compared to those in less disadvantaged areas [29]; this could partly explain the lower odds of reporting exposure to walking in a paddock or contact with soil in lower SES groups.

In our study, we also found that the most disadvantaged individuals were less likely to visit their GP but more likely to visit A&E and be hospitalised compared to the least disadvantaged. There is some evidence to suggest that, at the population-level, people living in more disadvantaged circumstances are generally more likely to interact with all three of these healthcare services [Reference Cookson30], although this is not specific to GI infections. There is also evidence to suggest that, amongst individuals with GI infections, people living in more disadvantaged circumstances are more likely to be hospitalised [Reference Olowokure31, Reference Rose32]. The discrepancy at the GP level could suggest that accessing a GP varies by type of illness for low SES and that there are differences in healthcare-seeking behaviour amongst more disadvantaged individuals with STEC in terms of choosing which services to access; differences in likelihood of milder cases in lower SES group being requested to provide or actually submitting a stool sample (and so not being diagnosed as STEC and appearing in this dataset) or potentially differences in recognition of symptoms at a more advanced stage of illness, delaying seeking of care or differences in access to care. It also raises the possibility that milder cases of STEC infection are less likely to be identified as such in the lower SES group, a trend which has been observed for illnesses in general [Reference Cookson30]: this could contribute to our finding of a lower STEC incidence in the lower SES group, which has been identified in other studies reliant on laboratory results for their analysis [Reference Banatvala33–Reference Nichols37]. Finally, it could suggest that although the incidence is lower in the more deprived population, case severity may be higher. However, further studies would be required to robustly assess these hypotheses including a better understanding of why the use of A&E and hospitalisation appear to be higher amongst disadvantaged individuals, despite adjusting for potential severity (stx gene). Further work to explore the time between symptom onset and seeking care, as well as differences in stool sampling availability and acceptance, would be of value, as would an exploration of the barriers to seeking care at an earlier stage of illness, to understand whether delayed care is responsible for the higher hospitalisation rate.

A limitation of our study is the cross-sectional observational design. Further research is needed to assess if the associations identified in our study are likely to be causal. Furthermore, the identification of STEC cases included in this study was via laboratory reporting. Laboratory confirmation underestimates the true burden of STEC in a population and such under-reporting may vary by socio-demographic characteristics which may have biased our study population. An area-level measure of SES was used and thus ecological fallacy is a possibility. However, person-to-person spread is an important risk factor for STEC infections and therefore community or area-level risk would be an important factor in considering the individual risk of infection. It has been shown there are larger social networks amongst cases of GI infection in lower socio-economic areas compared to higher socio-economic areas [Reference Rotherham38]. It was not possible to capture all categories of risk for STEC, including some person-to-person risk factors and more specific food exposures, for example, whether pink or rare meat was served. A further limitation of our study is that no comparable data exist on the specific food and environmental risk factors at the population level by SES. Thus, the differences detected in our study may simply reflect the patterning by SES of these risk factors in the general population. There were also some missing data in our study, particularly for ethnicity, which we addressed using multiple imputation. In addition, since exposure to a private water supply is relatively rare, despite being a known STEC risk factor, it was not possible to perform a multivariable analysis for this due to very small numbers in some covariates. Finally, the lack of significant findings for risk factors in children aged <16 years is likely to be due to low study power since the findings were in the same direction as those in the whole study population.

A strength of our study is the novel analysis of data collected through a country-wide, representative surveillance system to explore the social patterning of risk factors amongst STEC cases in England. Our study thus captures extensive risk factor and exposure data within a well-characterised STEC population. To the best of our knowledge, this is the first study to explore the social patterning of risk factors for STEC in the UK. STEC is a rare but potentially very serious infection and, particularly in children and the elderly, is likely to result in interaction with healthcare and/or public health services. Furthermore, testing of stool samples is routine in UK laboratories. These factors mean it is likely that NESSS captures data on a high proportion of diagnosed STEC cases in England and, hence, this study is likely to be representative of known STEC cases nationally.

In summary, we identified a higher incidence of STEC infection in the highest, compared to the lowest, SES group. There was evidence of social patterning of healthcare contact with more disadvantaged STEC cases having higher odds of A&E presentation and hospitalisation, which may suggest differential healthcare access or consequences of STEC infection by SES. We also found evidence to suggest that some risk factors for STEC are socially patterned, with lower odds of reported exposure to certain known risk factors such as non-UK travel, exposure to salad, fruit, vegetables or herbs and walking in a paddock. This could suggest that certain population groups may be more likely to be exposed to risk factors for STEC, which may partially explain the distribution of STEC cases by SES. Educational programmes geared towards reducing the prevalence of STEC infection should include making higher SES groups aware of their higher risk and specific risk factors. Further research is required to determine the reasons for the differential healthcare contact by SES and particularly for the higher hospitalisation rates observed amongst the lowest SES group. Furthermore, research is needed to determine whether there are risk factors for STEC infection that are more common in lower SES groups.