Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The Relationship Between Census Tract Poverty and Shiga Toxin–Producing E. coli Risk, Analysis of FoodNet Data, 2010–2014

The Relationship Between Census Tract Poverty and Shiga Toxin–Producing E. coli Risk, Analysis of... Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 Open Forum Infectious Diseases MAJOR ARTICLE e R Th elationship Between Census Tract Poverty and Shiga Toxin–Producing E. coli Risk, Analysis of FoodNet Data, 2010–2014 1 1 2 3 4 5 6 7 8 James L. Hadler, Paula Clogher, Jennifer Huang, Tanya Libby, Alicia Cronquist, Siri Wilson, Patricia Ryan, Amy Saupe, Cyndy Nicholson, 9 10 11 1 Suzanne McGuire, Beletshachew Shiferaw, John Dunn, and Sharon Hurd 1 2 Emerging Infections Program, Yale School of Public Health, New Haven, Connecticut; Division of Foodborne, Waterborne, and Environmental Diseases, Centers for Disease Control and Prevention, 3 4 5 Atlanta, Georgia; California Emerging Infections Program, Oakland, California; Emerging Infections Program, Colorado Department of Public Health and Environment, Denver, Colorado; Emerging 6 7 Infections Program, Georgia Department of Public Health, Atlanta, Georgia; Emerging Infections Program, Maryland Department of Health, Baltimore, Maryland; Emerging Infections Program, Minnesota 8 9 Department of Health, St Paul, Minnesota; Emerging Infections Program, New Mexico Department of Health, Santa Fe, New Mexico; Emerging Infections Program, New York State Department of Health, 10 11 Albany, New York; Emerging Infections Program, Oregon Health Authority, Portland, Oregon; Emerging Infections Program, Tennessee Department of Health, Nashville, Tennessee Background. e r Th elationship between socioeconomic status and Shiga toxin–producing Escherichia coli (STEC) is not well understood. However, recent studies in Connecticut and New York City found that as census tract poverty (CTP) decreased, rates of STEC increased. To explore this nationally, we analyzed surveillance data from laboratory-confirmed cases of STEC from 2010–2014 for all Foodborne Disease Active Surveillance Network (FoodNet) sites, population 47.9 million. Methods. Case residential data were geocoded and linked to CTP level (2010–2014 American Community Survey). Relative rates were calculated comparing incidence in census tracts with <20% of residents below poverty with those with ≥20%. Relative rates of age-adjusted 5-year incidence per 100 000 population were determined for all STEC, hospitalized only and hemolytic-uremic syndrome (HUS) cases overall, by demographic features, FoodNet site, and surveillance year. Results. er Th e were 5234 cases of STEC; 26.3% were hospitalized, and 5.9% had HUS. Five-year incidence was 10.9/100 000 population. Relative STEC rates for the <20% compared with the ≥20% CTP group were >1.0 for each age group, FoodNet site, sur- veillance year, and race/ethnic group except Asian. Relative hospitalization and HUS rates tended to be higher than their respective STEC relative rates. Conclusions. Persons living in lower CTP were at higher risk of STEC than those in the highest poverty census tracts. This is unlikely to be due to health care–seeking or diagnostic bias as it applies to analysis limited to hospitalized and HUS cases. Research is needed to better understand exposure differences between people living in the lower vs highest poverty-level census tracts to help direct prevention efforts. Keywords. census tract; E. coli; incidence; Shiga toxin; poverty. Infections with Shiga toxin–producing Escherichia coli (STEC), and northwestern states have the highest incidence of diag- both O157 and non-O157 serogroups, are an important public nosed infection [2, 3]. health problem, causing an estimated 40 000–570 000 infections Data to determine which demographic groups are most     per year in the United States, including 549–5585 hospital- aeff cted by a disease help to guide public health prevention izations and as many as 113 deaths [1]. Death usually results efforts. Much of our knowledge in the United States of which from intravascular complications such as hemolytic-uremic demographic groups are most aeff cted by STEC has come syndrome (HUS). Although most human exposures to STEC through mandated reporting to state and local health depart- come from contaminated food, people can be exposed from ments, ultimately to the Centers for Disease Control and contaminated water and direct contact with infected ruminants Prevention. As part of disease reporting in general, including and humans. Demographically in the United States, children STEC, there have been no systematic efforts to collect infor - <5 years of age, females, whites, and residents of north-central mation on socioeconomic measures. Although data are not systematically collected with individual socioeconomic infor- mation (eg, education level, income), area-based socioeco- nomic measures, such as census tract poverty level, can be Received 16 April 2018; editorial decision 14 June 2018; accepted 29 June 2018. used if an individual’s residential location is known. Residential Correspondence: J. L. Hadler, MD, MPH, Emerging Infections Program, Yale School of Public Health, One Church Street, 7th floor, New Haven, CT 06511 (hadlerepi@gmail.com). address can be geocoded, matched to the census tract in which Open Forum Infectious Diseases a person lives, and linked to the socioeconomic measures Published by Oxford University Press on behalf of Infectious Diseases Society of America 2018. of that census tract as determined by the US Census and the This work is written by (a) US Government employee(s) and is in the public domain in the US. DOI: 10.1093/ofid/ofy148 American Community Survey [4]. Increasingly in the past Census Tract Poverty and STEC Risk • OFID • 1 Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 decade, area-based socioeconomic measures, particularly cen- whether the patient had traveled internationally in the 7  days sus tract–level poverty, have been used and found to shed a new before onset of illness. perspective on the epidemiology of a number of diseases under For this analysis, each FoodNet site geocode the residential public health surveillance [5–12]. address of all STEC and HUS cases for the years 2010–2014 Analyses of combined data from multiple years of sur- inclusive. Geocoded addresses were assigned to census tracts. veillance in Connecticut and in New York City using census Census tract poverty level, defined as the percentage of house- tract–level poverty found that STEC and HUS incidence were holds in the census tract living below the federal poverty level, consistently higher among those living in census tracts with was determined from the 2010–2014 American Community lower levels of poverty (ie, high socioeconomic status [SES]) Survey 5-Year Estimates [15]. Census tracts were categorized than in those with the highest levels of poverty [6, 13]. To deter- by their percentage of households living below the poverty mine whether these findings were more generalizable to the level (<5%, 5%–9%, 10%–19%, ≥20%), as recommended by the United States, we analyzed geocoded data linked to census tract Public Health Disparities Geocoding Project [4, 16] and as used poverty level from all reported cases of STEC and HUS occur- for other multisite EIP data analysis projects [9, 10]. Census ring in the Foodborne Disease Active Surveillance Network tract–specific denominators were determined from the 2010 (FoodNet) from 2010 to 2014. US Census. Data Analysis METHODS Age-adjusted (2000 US standard population) incidence rates FoodNet is the principal foodborne disease surveillance com- per 100 000 person-years overall and for each of the 4 poverty ponent of the Centers for Disease Control and Prevention’s categories were calculated for all STEC combined, for the O157 (CDC’s) Emerging Infections Program (EIP), a collaboration serotype only, and for all non-O157 serotypes combined. Age of the CDC, the US Department of Agriculture’s Food Safety standardization was done using 5 age categories: 0–4  years, and Inspection Services (USDA-FSIS), the Food and Drug 5–17 years, 18–49 years, 50–64 years, and ≥65 years. These cate- Administration (FDA), and 10 state health departments. gories were based on overall age group–specific incidence rates, FoodNet includes the states of Connecticut, Georgia, Maryland, combining age-specific rates that were similar into the same age Minnesota, New Mexico, Oregon, and Tennessee and selected groups. counties in California, Colorado, and New York. FoodNet staff At this stage of analysis, we made 2 decisions to guide fur- conduct active population-based surveillance for laborato- ther analyses. First, we decided to combine STEC O157 with ry-confirmed cases of STEC, including O157 and non-O157 STEC non-O157 as they had a similar relationship to poverty serogroups, and hemolytic-uremic syndrome (HUS). Enhanced (Figure  1) and to conduct all further analyses using all STEC surveillance methods have been previously described [14]. Data combined only. Second, we decided to combine the 3 lowest collected on each laboratory-confirmed case of STEC and each poverty groups (0%–<5%, 5%–<10% and 10%–20%) into a case of HUS meeting a specific HUS clinical case definition single “lower poverty” group as age-adjusted incidence did not include demographic information (age, sex, race/ethnicity, vary significantly between them (Figure  1). Further analyses street address of residence), whether a person was hospitalized by poverty compared this lower poverty group (<20% below or died, whether the infection was part of an outbreak, and poverty) with the high poverty group (≥20% below poverty). <5% 5%-<10% 10%-<20% 20% and higher 14 ** ** ** 8 ** ** ** ** ** 6 ** O157 Non-O157 All STEC ** P < .001, vs ≥20% group Figure 1. Age-adjusted 5-year incidence of O157, non-O157, and all Shiga toxin–producing Escherichia coli (STEC) by census tract poverty category, Foodborne Disease Active Surveillance Network, 2010–2014. 2 • OFID • Hadler et al 5-year incidence per 100 000 population Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 To account for possible bias introduced by this post hoc deci- Islanders. When examined by FoodNet site, the same was true sion, a P value of <.01 instead of P <.05 was considered statisti- for each site except for California and Colorado where IRRs of cally significant in comparing the newly defined lower poverty >1.0 had 95% confidence limits that overlapped with 1 (Figure 2). group with the high poverty group, and 99% confidence limits For hospitalized and HUS cases, the IRR was higher than for were calculated. all STEC cases overall and for most demographic subgroups We then calculated incidence rate ratios (IRRs) comparing (Table  2). There were 4 states with IRRs for hospitalized cases age-adjusted incidence in the lower vs high poverty categories that were statistically significantly greater than 1 at a P  <  .01 for (1) all STEC cases, (2) hospitalized cases, and (3) HUS cases. level. For all 4, the IRR for hospitalized cases was higher than IRRs were determined by year, by site, by sex, and by race/eth- for all STEC cases (Figure 2). nicity to determine if the relationship was consistent by each of Of the 5234 STEC cases, 437 (8.3%) had traveled internation- these variables. ally in the 7 days before symptom onset, 4455 (85.1%) had not, Finally, STEC cases were separated into international travel– and travel status was unknown for 342 (6.5%). Table  2 shows associated cases and cases acquired in the United States to overall age-adjusted IRRs of lower to high (≥20%) poverty determine how the overall findings of the relationship of census groups for each STEC outcome by international travel status. tract poverty level to STEC and its severe outcomes were influ- For cases in each travel group (international travel and domes- enced by international travel. International travel–associated tic exposure only), infection was associated with living in lower cases were those occurring in a person who had been out of the poverty census tracts. Point estimates of the relative rates were country at any time during the 7 days before illness onset. greater than 1.0 for each age group for both domestic and inter- Statistical analyses were performed using SAS, version 9.3 nationally acquired cases, although for some groups with small (SAS Institute Inc, Cary, NC, USA). IRRs and 95% confidence numbers of cases, this was not statistically significant (data not intervals were calculated using the Statcalc function in Epi shown). Info 7. Because most cases were acquired from domestic exposure, domestic cases were examined by serogroup (O157 vs non- RESULTS O157), age, and surveillance site. The findings were consistent with those from the analyses of all STEC cases: for all groups, There were a total of 5234 cases (96% of total) that were able to sites, and serogroups, the rate of STEC was higher in the less be geocoded to the rooftop level. The characteristics of these poor census tracts than the poorest census tracts and of a simi- cases and average annual incidence of each of the 3 outcomes lar magnitude to that using all STEC cases regardless of interna- are shown in Table 1. Overall, 26.3% were hospitalized and 5.9% tional travel status (data not shown). had HUS. The majority (55%) of STEC cases were non-O157. Crude incidence rates of all 3 outcomes were highest among DISCUSSION the youngest age groups, non-Hispanic whites, residents of lower poverty census tracts, and in the later years of the 5-year This study had several important findings. First, it demon- period. Numbers of cases by site ranged from a low of 223 in strated that persons in the United States of higher SES status, New Mexico to a high of 1274 in Minnesota. Site-specific inci- including very young children, have had a consistently higher dence ranged from a low of 0.99 in Georgia to a high of 4.74 in risk of acquiring clinically consequential STEC infection than Minnesota (data not shown). those living in poverty. Second, the findings were similar for Figure 1 shows age-adjusted rates of STEC O157, STEC non- O157 and non-O157 STEC. These findings have implications O157, and all STEC by census tract poverty level. Overall and for both current intervention efforts and future research. for each group of serotypes, the highest poverty group (≥20%) Before this analysis, there have been 3 published analyses had the lowest rate, and it was statistically significantly lower of the relationship of SES to STEC in the United States, all than for each of the other poverty groups (P < .001). There was using area-based SES measures, all with similar findings. The no consistent hierarchical relationship of incidence rate within first was a national study from the National Notifiable Disease the other poverty groups. Based on these findings, we combined System covering nationally reported data from 1993–2002. O157 and non-O157 serotypes and collapsed the 3 lowest pov- Using counties as the area-based unit of analysis, it found lower erty groups into 1 group <20% below the federal poverty level county-level incidence with higher county poverty levels [17]. for subsequent analyses. However, its findings were limited. It analyzed nationally noti- Age-adjusted IRRs comparing incidence of all STEC, hos- fiable data for E. coli O157:H7 (not all STEC) from 1993–2002, pitalized cases, and HUS cases in the combined lower poverty used a very broad area-based level, and only used county-level, group with the highest poverty group are shown in Table  2. not individual, variables. It was further limited in shedding For all STEC, IRRs were consistently significantly greater than no light on the contribution of neighborhood poverty to risk 1.0 for each age group, by sex, by surveillance year, and for all in specific age groups or on whether the risk associated with major race/ethnic groups except for non-Hispanic Asian/Pacific neighborhood poverty changed over time. The other 2 were the Census Tract Poverty and STEC Risk • OFID • 3 Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 Table 1. STEC Case Characteristics and Incidence by Severity, Serotype, and Selected Demographic Characteristics, FoodNet 2010–2014 a b c c c Characteristic No (%) of STEC Cases Population STEC Incidence Incidence of Hospitalized Cases Incidence of HUS All cases 5234 47 898 7 45 2.19 0.57 0.13 Serotype O157 2355 (45.0) 47 898 7 45 0.98 0.38 - Non-O157 2879 (55.0) 47 898 7 45 1.20 0.19 - Severity Hospitalized 1374 (26.3) 47 898 7 45 0.57 - - HUS 308 (5.9) 47 898 7 45 0.13 - - Age group, y 0–4 1267 (24.2) 2 986 919 8.48 1.65 0.92 5–17 1391 (26.6) 8 103 964 3.43 0.96 0.31 18–49 1615 (30.9) 20 897 307 1.55 0.34 0.02 50–64 455 (8.7) 9 508 512 0.96 0.33 0.01 ≥65 506 (9.7) 6 402 043 1.58 0.70 0.07 Sex Male 2331 (44.5) 23 513 733 1.98 0.52 0.12 Female 2903 (55.5) 24 385 012 2.38 0.62 0.14 Race/ethnicity Hispanic 536 (10.2) 5 253 918 2.04 0.32 0.08 Non-H white 3554 (67.9) 30 919 814 2.30 0.71 0.15 Non-H black 237 (4.5) 7 123 093 0.67 0.18 0.02 Non-H Asian/PI 116 (0.3) 2 277 853 1.02 0.25 0.10 Non-H other/unknown 791 (15.2) 1 382 468 - - - Census tract poverty <5% 1098 (21.0) 8 940 621 2.46 0.62 0.12 5%–<10% 1401 (26.8) 10 888 246 2.57 0.69 0.16 10%–<20% 1701 (32.5) 14 713 758 2.31 0.64 0.17 ≥20% 1031 (19.7) 12 415 515 1.66 0.40 0.08 Year 2010 826 (15.8) 47 898 7 45 1.72 0.52 0.11 2011 982 (18.8) 47 898 7 45 2.05 0.61 0.12 2012 1122 (21.4) 47 898 7 45 2.34 0.60 0.13 2013 1147 (21.9) 47 898 7 45 2.39 0.69 0.20 2014 1157 (22.1) 47 898 7 45 2.42 0.55 0.10 Abbreviations: H, Hispanic; HUS, hemolytic-uremic syndrome; PI, Pacific Islander; STEC, Shiga toxin–producing Escherichia coli. Three cases missing poverty status. Combined population denominator for all FoodNet sites. Based on 2012 US Census Bureau population estimates, except for race/ethnicity and census tract poverty, which are based on the 2010 US Census. Incidence per 100 000 person-years. more recent studies in Connecticut and New York City using appears to be universal across the United States and enduring. the census tract as the area size for analysis and individual Of interest, this relationship does not necessarily hold in other demographic variables rather than percentage of the population developed countries. A  study in Denmark covering national in selected age groups [6, 13]. They each found an association of data from 1993–2004 found no association between individ- lower STEC incidence with higher census tract poverty levels. ual income or education level and STEC [18], and a study from However, as they were local and in the Northeast (mostly urban Alberta, Canada, covering reported data from 2000–2002 found and suburban), it was not clear what would happen with more no association with census subdivision rates of STEC and per- nationally representative data. With our data set, we were able centage of individuals living in low-income households [19]. to confirm this relationship at the national level, that it was con- a Th t the overall findings for STEC O157 and STEC non- sistent across 5 consecutive years, strongest for young children, O157 were similar with respect to poverty is consistent with and present more recently than 1993–2002. Our findings were other aspects of their epidemiology. Although different STEC also consistent across all 10 FoodNet sites by sex and within serotypes may differ in pathogenicity [20, 21], those that cause most race/ethnic groups, and for each of the 3 STEC health severe human disease appear to share a similar ecology and outcomes examined (laboratory-confirmed disease, hospital- many of the same risk factors for acquisition [20–22]. This is ization, and HUS), the relationship between STEC and SES a likely explanation for why they have a similar relationship to 4 • OFID • Hadler et al Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 Table 2. Age-Adjusted Incidence Rate Ratios of Lower (<20% Below Federal Poverty Level) to High (≥20%) Poverty Groups of All STEC, Hospitalized STEC Cases, and Hemolytic-Uremic Syndrome Cases by Demographic Variables, Year, and International Travel Status, FoodNet Sites, 2010–2014 All STEC (n = 5234) Hospitalized Cases (n = 1374) HUS Cases (n = 308) Variable IRR 95% CI IRR 95% CI IRR 95% CI * * * All cases 1.53 1.43–1.64 1.63 1.43–1.87 2.08 1.53–2.81 Age group, y * * * 0–4 1.39 1.22–1.58 1.98 1.43–2.75 2.25 1.42–3.54 * * 5–17 1.69 1.47–1.94 1.95 1.48–2.57 1.71 1.07–2.73 * * 18–49 1.69 1.49–1.91 1.79 1.37–2.35 2.11 0.61–7.22 50–64 1.27 1.01–1.62 0.93 0.65–1.34 1.76 0.21–14.66 ≥65 1.26 1.01–1.57 1.19 0.86–1.65 6.13 0.88–45.65 Sex * * * Male 1.42 1.28–1.57 1.84 1.49–2.26 2.83 1.71–4.68 * * Female 1.56 1.43–1.71 1.48 1.24–1.77 1.60 1.10–2.34 Race/ethnicity Hispanic 1.62 1.32–2.00 1.71 1.07–2.74 3.20 0.91–11.22 * * Non-H white 1.20 1.10–1.30 1.37 1.16–1.61 1.34 0.98–1.85 Non-H black 1.61 1.23–2.11 1.24 0.75–2.04 0.90 0.18–4.48 Non-H Asian/PI 0.71 0.48–1.06 1.43 0.50–4.13 2.87 0.37–22.04 Year * * 2010 1.98 1.64–2.38 1.84 1.32–2.57 1.03 0.56–1.89 2011 1.22 1.06–1.42 1.30 NS 2.33 1.11–4.92 * * * 2012 1.59 1.37–1.84 1.69 1.24–2.29 2.56 1.22–5.37 * * * 2013 1.71 1.47–1.98 1.89 1.41–2.53 2.11 1.22–3.65 * * * 2014 1.52 1.32–1.76 1.57 1.15–2.14 4.04 1.45–11.24 International travel 7 d before symptom onset * * * No (n = 4454) 1.52 1.42–1.64 1.60 1.39–1.83 1.81 1.35–2.42 Yes (n = 435) 1.72 1.35–2.20 2.64 0.79–8.81 - - Abbreviations: CI, confidence interval; H, Hispanic; HUS, hemolytic-uremic syndrome; IRR, incidence rate ratio; PI, Pacific Islander; STEC, Shiga toxin–producing Escherichia coli. P < .01. poverty in the United States, as differences in rates of exposure diarrhea, and to get diagnosed than those in high poverty cen- to STEC are the likely reasons for those in high poverty census sus tracts. This consideration was examined for FoodNet sites tracts to be at lower risk in the United States. during 2000–2003 and found not to be true [23]. In fact, this One important consideration in interpretation of results from analysis of FoodNet population survey data found that those this study is that persons in higher SES census tracts might be in the lowest income category (household income  <$25 000) more likely to seek health care for STEC symptoms, especially were more likely to seek care for acute diarrheal illness. As this All STEC Hospitalized only 2.5 * * * * 1.5 0.5 CA CO CT GA MD MN NM NY OR TN All sites * P ≤ .01 that IRR is >1.0 Figure 2. Age-adjusted incidence rate ratios (IRRs) of lower to high poverty groups of all STEC and hospitalized Shiga toxin–producing Escherichia coli (STEC) cases, by Foodborne Disease Active Surveillance Network site, 2010–2014. Census Tract Poverty and STEC Risk • OFID • 5 IRR Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 analysis took place before our study time period, it is possible findings (eg, some poor neighborhoods may have a high prev- that health care–seeking behaviors might have changed. Thus, alence of relatively bacteria-free fast food and lack stores with in addition to laboratory-confirmed, largely outpatient illness, fresh meat and produce, lowering the potential for foodborne we examined the outcomes of hospitalization and HUS, both STEC exposure) and may have a stronger association if the likely to be less influenced by income or health care access. For neighborhood SES is more of a factor than individual-level SES. both outcomes, the association with higher SES was, if any- In addition, not everyone in the neighborhood has the same thing, even stronger. individual SES. Thus, the findings need to be interpreted in this An important risk factor for acquisition of STEC in the context. Second, most of the cases were identified because they United States, especially STEC non-O157, is international travel sought medical care, and stool specimens were taken for diag- [20, 21]. In our study, 8.9% of STEC cases had traveled inter- nosis. Not everyone with STEC infection seeks medical care, nationally in the 7  days before illness onset, and international and not all those who do have specific diagnostic testing. Thus, travel–associated STEC was least likely among persons living the cases used in this population-based study do not comprise in high poverty census tracts. However, aer ex ft cluding interna- the universe of all STEC infections in FoodNet sites [1]. To the tional travel, those in high poverty areas still had the lowest risk extent that there might have been bias by SES in who sought of laboratory-confirmed STEC and its complications. medical care and was diagnosed, the true magnitude of the er Th e are some data to support the hypothesis that adults of associations found could be different in either direction. Third, higher SES status in the United States have had a higher prev- we used relatively broad age groups to age-adjust and to exam- alence of some consumer-level STEC risks other than interna- ine age group–specific associations. However, those ages com- tional travel. Such risks include eating raw or undercooked beef prising each age group were selected because they had a similar [24–26], eating in restaurants [25], eating raw fresh vegetables, age-specific incidence. In addition, there were several statistical fruits, and nuts [27], and poorer hygiene practices to prevent limitations. As previously described, we made a post hoc recat- cross-contamination from raw products despite higher knowl- egorization of 4 census tract poverty levels to 2; thus, only P edge levels [26]. Although more current studies are needed, as values of ≤.01 should be considered statistically significant. In stated in a recent systematic review, “SES should be considered addition, the 95% confidence limits on the incidence rate ratios when targeting consumer level public health interventions for presented in Table  2 were calculated assuming there was no foodborne pathogens,” including STEC [28]. clustering of outcomes within individual census tracts as we did Risk factors for young children (<5 years) in the United States not have individual census tract identifiers; thus, they are pos- have been identified but have not been put in a relative attribu- sibly narrower than they might actually be. Finally, this study tion context or examined by SES, despite these children having was limited to descriptively defining the relationship between the highest age-specific risk of laboratory-confirmed infection census tract poverty used as a single recommended SES surveil- and HUS. Widely recognized means of exposure for young lance variable and STEC incidence in the national foodborne children include person-to-person transmission, particularly disease sentinel surveillance system, FoodNet. Examination of in day care centers and from contact with other children with other area-based SES variables and conducting an analysis to diarrhea, consumption of contaminated foods and beverages, determine the relative importance of census tract poverty com- particularly undercooked beef products and other food items pared with each of the other descriptive variables available were cross-contaminated from them, and direct or indirect contact beyond the scope of this analysis. with farm animals, particularly ruminants. Although all these CONCLUSIONS potential exposure factors could be more common in those In summary, the findings from analysis of population-based of higher SES, documentation of their association with SES data from FoodNet, the national sentinel foodborne disease first, and then of the factors leading to exposure (eg, hygiene surveillance system, confirm those from more localized stud- in day care centers and at home, including handling raw beef ), ies that people living in lower poverty neighborhoods are at is needed to guide prevention efforts targeted at the consumer. higher risk of acquiring STEC infection and suffering its more This study has some notable strengths. These include the high severe complications. These findings point to a need to consider percentage of cases geocoded to the census tract level, partici- higher SES when targeting specific public health interventions pation from all 10 FoodNet sites encompassing more than 48 to prevent STEC infection. However, a current understanding million people enabling generalization to the US population, of SES differences in risk factors, for children in particular, is and analysis by 3 levels of severity of STEC infection. needed to enable potentially effective interventions. er Th e are also some important limitations. Most impor - tantly, this study used census tract SES rather than individual Acknowledgments SES. Although the 2 are usually correlated, they do not measure We thank the many FoodNet surveillance staff at each site and at the exactly the same thing. In particular, census tract SES includes CDC for their work in collecting, geocoding, and collating the data used in this paper. possible neighborhood SES factors contributing to the observed 6 • OFID • Hadler et al Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 Financial support. This work was supported by the Centers for Disease 13. Whitney BM, Mainero C, Humes E, Hurd S, Niccolai L, Hadler JL. Socioeconomic status and foodborne pathogens in Connecticut, USA, 2000–2011. Emerg Infect Control and Prevention (CI05-026, CK12-1202, CK17-1701). Dis 2015; 21:1617–24. Potential conifl cts of interest. All authors: no reported conflicts of 14. Henao OL, Jones TF, Vugia DJ, Griffin PM; Foodborne Diseases Active Surveillance interest. All authors have submitted the ICMJE Form for Disclosure of Network (FoodNet) Workgroup. Foodborne diseases active surveillance network-2 Potential Conflicts of Interest. Conflicts that the editors consider relevant to decades of achievements, 1996–2015. Emerg Infect Dis 2015; 21:1529–36. the content of the manuscript have been disclosed. 15. US Census Bureau. American fact finder advanced. Available at: https://fact- finder.census.gov/faces/nav/jsf/pages/searchresults.xhtml?refresh=t. Accessed 8 March 2018. References 16. Public Health Disparities Geocoding Project. Available at: http://www.hsph.har- 1. Scallan E, Hoekstra RM, Angulo FJ, et  al. Foodborne illness acquired in the vard.edu/thegeocodingproject. Accessed 8 March 2018. United States–major pathogens. Emerg Infect Dis 2011; 17:7–15. 17. Chang M, Groseclose SL, Zaidi AA, Braden CR. An ecological analysis of sociode- 2. Gould LH, Mody RK, Ong KL, et  al. Increased recognition of non-O157 Shiga mographic factors associated with the incidence of salmonellosis, shigellosis, and toxin-producing Escherichia coli infections in the United States during 2000– E. coli O157:H7 infections in US counties. Epidemiol Infect 2009; 137:810–20. 2010: epidemiologic features and comparison with E.  coli O157 infections. 18. Simonsen J, Frisch M, Ethelberg S. Socioeconomic risk factors for bacterial gas- Foodborne Pathog Dis 2013; 10:453–60. trointestinal infections. Epidemiology 2008; 19:282–90. 3. Adams DA, Thomas KR, Jajosky R, et  al. Summary of notifiable infectious dis- 19. Pearl DL, Louie M, Chui L, et  al. A multi-level approach for investigating eases and conditions—United States, 2014. MMWR Morb Mortal Wkly Rep socio-economic and agricultural risk factors associated with rates of reported 2016; 63:1–152. cases of Escherichia coli O157 in humans in Alberta, Canada. Zoonoses Public 4. Krieger N, Chen JT, Waterman PD, et  al. Painting a truer picture of US socio- Health 2009; 56:455–64. economic and racial/ethnic health inequalities: the Public Health Disparities 20. Hedican E, Medus C, Besser JM, et al. Characteristics of O157 versus non-O157 Geocoding Project. Am J Public Health 2005; 95:312–23. Shiga toxin-producing Escherichia coli infections in Minnesota, 2000–2006. Clin 5. Toprani A, Hadler JL. Selecting and applying a standard area-based socioeco- Infect Dis 2009; 49:358–64. nomic status measure for public health data: analysis for New York City. New 21. Hadler JL, Clogher P, Phan Q, et al. Ten year trends and risk factors for non-O157 York City Department of Health and Mental Hygiene: Epi Research Report. Shiga-toxin producing E.  coli found through Shiga-toxin testing, Connecticut, 2013. Available at: http://www1.nyc.gov/assets/doh/downloads/pdf/epi/epire- 2000–2009. Clin Infect Dis 2011; 53:269–76. search-SES-measure.pdf. Accessed 8 March 2018. 22. Bettelheim KA. The non-O157 shiga-toxigenic (verocytotoxigenic) Escherichia 6. Greene SK, Levin-Rector A, Hadler JL, Fine AD. Disparities in reportable com- coli; under-rated pathogens. Crit Rev Microbiol 2007; 33:67–87. municable indicators by census tract-level poverty, New York City, 2006–2013. 23. Scallan E, Jones TB, Cronquist A, et al. Factors associated with seeking medical Am J Public Health 2015; 105:e27–34. care and submitting a stool sample in estimating the burden of foodborne illness. 7. Centers for Disease Control and Prevention. Social determinants of health among Foodborne Pathog Dis 2006; 3:432–8. adults diagnosed with HIV infection in 11 states, the District of Columbia and 24. Taylor EV, Holt KG, Mahon BE, Ayers T, Norton D, Gould LH. Ground beef con- Puerto Rico, 2013. HIV Surveillance Supplemental Report. 2015. http://www.cdc. sumption patterns in the United Sates, FoodNet, 2006 through 2007. J Food Prot gov/hiv/pdf/library/reports/surveillance/cdc-hiv-surveillancereport-vol20-no5. 2012; 75:341–6. pdf. Accessed 8 March 2018. 25. Wagner JA. Foodnet Population Survey 2006–2007: Differences in Prevalence of 8. Boscoe FP, Johnson CJ, Sherman RL, et  al. The relationship between area pov- Health Care Seeking Behaviors and Exposures to Risk Factors for Foodborne Illness erty rate and site-specific cancer incidence in the United States. Cancer 2014; by Socioeconomic Status [master’s thesis]. New Haven, CT: Yale University; 2014. 120:2191–8. Available at: http://elischolar.library.yale.edu/cgi/viewcontent.cgi?article=1304&- 9. Hadler JL, Vugia DJ, Bennett NM, Moore MR. Emerging infections program context=ysphtdl. Accessed 8 March 2018. efforts to address health equity. Emerg Infect Dis 2015; 21:1589–94. 26. Patil SR, Cates S, Morales R. Consumer food safety knowledge, practices, and 10. Hadler JL, Yousey-Hindes K, Perez A, et al. Influenza-related hospitalizations and demographic differences: findings from a meta-analysis. J Food Prot 2005; poverty levels—United States 2010–2012. MMWR MorbMortal Wkly Rep 2016; 68:1884–94. 65:101–5. 27. Dubowitz T, Heron M, Bird CE, et al. Neighborhood socioeconomic status and 11. Soto K, Petit S, Hadler JL. Changing disparities in invasive pneumococcal dis- fruit and vegetable intake among whites, blacks, and Mexican Americans in the ease by socioeconomic status and race in Connecticut, 1998–2008. Public Health United States. Am J Clin Nutr 2008; 87:1883–91. Reports 2011; 126(Suppl 3):81–8. 28. Newman KL, Leon JS, Rebolledo PA, Scallan E. The impact of socioeconomic sta- 12. Bemis K, Marcus R, Hadler JL. Campylobacteriosis incidence is associated with tus on foodborne illness in high-income countries: a systematic review. Epidemiol low poverty: Connecticut, 1999–2009. Emerg Infect Dis 2014; 20:1240–2. Infect 2015; 143:2473–85. Census Tract Poverty and STEC Risk • OFID • 7 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Open Forum Infectious Diseases Oxford University Press

The Relationship Between Census Tract Poverty and Shiga Toxin–Producing E. coli Risk, Analysis of FoodNet Data, 2010–2014

Download PDF
7 pages

Loading next page...
 
/lp/oxford-university-press/the-relationship-between-census-tract-poverty-and-shiga-toxin-9JN9CdJNNj

References (31)

Publisher
Oxford University Press
Copyright
Published by Oxford University Press on behalf of Infectious Diseases Society of America 2018.
eISSN
2328-8957
DOI
10.1093/ofid/ofy148
Publisher site
See Article on Publisher Site

Abstract

Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 Open Forum Infectious Diseases MAJOR ARTICLE e R Th elationship Between Census Tract Poverty and Shiga Toxin–Producing E. coli Risk, Analysis of FoodNet Data, 2010–2014 1 1 2 3 4 5 6 7 8 James L. Hadler, Paula Clogher, Jennifer Huang, Tanya Libby, Alicia Cronquist, Siri Wilson, Patricia Ryan, Amy Saupe, Cyndy Nicholson, 9 10 11 1 Suzanne McGuire, Beletshachew Shiferaw, John Dunn, and Sharon Hurd 1 2 Emerging Infections Program, Yale School of Public Health, New Haven, Connecticut; Division of Foodborne, Waterborne, and Environmental Diseases, Centers for Disease Control and Prevention, 3 4 5 Atlanta, Georgia; California Emerging Infections Program, Oakland, California; Emerging Infections Program, Colorado Department of Public Health and Environment, Denver, Colorado; Emerging 6 7 Infections Program, Georgia Department of Public Health, Atlanta, Georgia; Emerging Infections Program, Maryland Department of Health, Baltimore, Maryland; Emerging Infections Program, Minnesota 8 9 Department of Health, St Paul, Minnesota; Emerging Infections Program, New Mexico Department of Health, Santa Fe, New Mexico; Emerging Infections Program, New York State Department of Health, 10 11 Albany, New York; Emerging Infections Program, Oregon Health Authority, Portland, Oregon; Emerging Infections Program, Tennessee Department of Health, Nashville, Tennessee Background. e r Th elationship between socioeconomic status and Shiga toxin–producing Escherichia coli (STEC) is not well understood. However, recent studies in Connecticut and New York City found that as census tract poverty (CTP) decreased, rates of STEC increased. To explore this nationally, we analyzed surveillance data from laboratory-confirmed cases of STEC from 2010–2014 for all Foodborne Disease Active Surveillance Network (FoodNet) sites, population 47.9 million. Methods. Case residential data were geocoded and linked to CTP level (2010–2014 American Community Survey). Relative rates were calculated comparing incidence in census tracts with <20% of residents below poverty with those with ≥20%. Relative rates of age-adjusted 5-year incidence per 100 000 population were determined for all STEC, hospitalized only and hemolytic-uremic syndrome (HUS) cases overall, by demographic features, FoodNet site, and surveillance year. Results. er Th e were 5234 cases of STEC; 26.3% were hospitalized, and 5.9% had HUS. Five-year incidence was 10.9/100 000 population. Relative STEC rates for the <20% compared with the ≥20% CTP group were >1.0 for each age group, FoodNet site, sur- veillance year, and race/ethnic group except Asian. Relative hospitalization and HUS rates tended to be higher than their respective STEC relative rates. Conclusions. Persons living in lower CTP were at higher risk of STEC than those in the highest poverty census tracts. This is unlikely to be due to health care–seeking or diagnostic bias as it applies to analysis limited to hospitalized and HUS cases. Research is needed to better understand exposure differences between people living in the lower vs highest poverty-level census tracts to help direct prevention efforts. Keywords. census tract; E. coli; incidence; Shiga toxin; poverty. Infections with Shiga toxin–producing Escherichia coli (STEC), and northwestern states have the highest incidence of diag- both O157 and non-O157 serogroups, are an important public nosed infection [2, 3]. health problem, causing an estimated 40 000–570 000 infections Data to determine which demographic groups are most     per year in the United States, including 549–5585 hospital- aeff cted by a disease help to guide public health prevention izations and as many as 113 deaths [1]. Death usually results efforts. Much of our knowledge in the United States of which from intravascular complications such as hemolytic-uremic demographic groups are most aeff cted by STEC has come syndrome (HUS). Although most human exposures to STEC through mandated reporting to state and local health depart- come from contaminated food, people can be exposed from ments, ultimately to the Centers for Disease Control and contaminated water and direct contact with infected ruminants Prevention. As part of disease reporting in general, including and humans. Demographically in the United States, children STEC, there have been no systematic efforts to collect infor - <5 years of age, females, whites, and residents of north-central mation on socioeconomic measures. Although data are not systematically collected with individual socioeconomic infor- mation (eg, education level, income), area-based socioeco- nomic measures, such as census tract poverty level, can be Received 16 April 2018; editorial decision 14 June 2018; accepted 29 June 2018. used if an individual’s residential location is known. Residential Correspondence: J. L. Hadler, MD, MPH, Emerging Infections Program, Yale School of Public Health, One Church Street, 7th floor, New Haven, CT 06511 (hadlerepi@gmail.com). address can be geocoded, matched to the census tract in which Open Forum Infectious Diseases a person lives, and linked to the socioeconomic measures Published by Oxford University Press on behalf of Infectious Diseases Society of America 2018. of that census tract as determined by the US Census and the This work is written by (a) US Government employee(s) and is in the public domain in the US. DOI: 10.1093/ofid/ofy148 American Community Survey [4]. Increasingly in the past Census Tract Poverty and STEC Risk • OFID • 1 Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 decade, area-based socioeconomic measures, particularly cen- whether the patient had traveled internationally in the 7  days sus tract–level poverty, have been used and found to shed a new before onset of illness. perspective on the epidemiology of a number of diseases under For this analysis, each FoodNet site geocode the residential public health surveillance [5–12]. address of all STEC and HUS cases for the years 2010–2014 Analyses of combined data from multiple years of sur- inclusive. Geocoded addresses were assigned to census tracts. veillance in Connecticut and in New York City using census Census tract poverty level, defined as the percentage of house- tract–level poverty found that STEC and HUS incidence were holds in the census tract living below the federal poverty level, consistently higher among those living in census tracts with was determined from the 2010–2014 American Community lower levels of poverty (ie, high socioeconomic status [SES]) Survey 5-Year Estimates [15]. Census tracts were categorized than in those with the highest levels of poverty [6, 13]. To deter- by their percentage of households living below the poverty mine whether these findings were more generalizable to the level (<5%, 5%–9%, 10%–19%, ≥20%), as recommended by the United States, we analyzed geocoded data linked to census tract Public Health Disparities Geocoding Project [4, 16] and as used poverty level from all reported cases of STEC and HUS occur- for other multisite EIP data analysis projects [9, 10]. Census ring in the Foodborne Disease Active Surveillance Network tract–specific denominators were determined from the 2010 (FoodNet) from 2010 to 2014. US Census. Data Analysis METHODS Age-adjusted (2000 US standard population) incidence rates FoodNet is the principal foodborne disease surveillance com- per 100 000 person-years overall and for each of the 4 poverty ponent of the Centers for Disease Control and Prevention’s categories were calculated for all STEC combined, for the O157 (CDC’s) Emerging Infections Program (EIP), a collaboration serotype only, and for all non-O157 serotypes combined. Age of the CDC, the US Department of Agriculture’s Food Safety standardization was done using 5 age categories: 0–4  years, and Inspection Services (USDA-FSIS), the Food and Drug 5–17 years, 18–49 years, 50–64 years, and ≥65 years. These cate- Administration (FDA), and 10 state health departments. gories were based on overall age group–specific incidence rates, FoodNet includes the states of Connecticut, Georgia, Maryland, combining age-specific rates that were similar into the same age Minnesota, New Mexico, Oregon, and Tennessee and selected groups. counties in California, Colorado, and New York. FoodNet staff At this stage of analysis, we made 2 decisions to guide fur- conduct active population-based surveillance for laborato- ther analyses. First, we decided to combine STEC O157 with ry-confirmed cases of STEC, including O157 and non-O157 STEC non-O157 as they had a similar relationship to poverty serogroups, and hemolytic-uremic syndrome (HUS). Enhanced (Figure  1) and to conduct all further analyses using all STEC surveillance methods have been previously described [14]. Data combined only. Second, we decided to combine the 3 lowest collected on each laboratory-confirmed case of STEC and each poverty groups (0%–<5%, 5%–<10% and 10%–20%) into a case of HUS meeting a specific HUS clinical case definition single “lower poverty” group as age-adjusted incidence did not include demographic information (age, sex, race/ethnicity, vary significantly between them (Figure  1). Further analyses street address of residence), whether a person was hospitalized by poverty compared this lower poverty group (<20% below or died, whether the infection was part of an outbreak, and poverty) with the high poverty group (≥20% below poverty). <5% 5%-<10% 10%-<20% 20% and higher 14 ** ** ** 8 ** ** ** ** ** 6 ** O157 Non-O157 All STEC ** P < .001, vs ≥20% group Figure 1. Age-adjusted 5-year incidence of O157, non-O157, and all Shiga toxin–producing Escherichia coli (STEC) by census tract poverty category, Foodborne Disease Active Surveillance Network, 2010–2014. 2 • OFID • Hadler et al 5-year incidence per 100 000 population Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 To account for possible bias introduced by this post hoc deci- Islanders. When examined by FoodNet site, the same was true sion, a P value of <.01 instead of P <.05 was considered statisti- for each site except for California and Colorado where IRRs of cally significant in comparing the newly defined lower poverty >1.0 had 95% confidence limits that overlapped with 1 (Figure 2). group with the high poverty group, and 99% confidence limits For hospitalized and HUS cases, the IRR was higher than for were calculated. all STEC cases overall and for most demographic subgroups We then calculated incidence rate ratios (IRRs) comparing (Table  2). There were 4 states with IRRs for hospitalized cases age-adjusted incidence in the lower vs high poverty categories that were statistically significantly greater than 1 at a P  <  .01 for (1) all STEC cases, (2) hospitalized cases, and (3) HUS cases. level. For all 4, the IRR for hospitalized cases was higher than IRRs were determined by year, by site, by sex, and by race/eth- for all STEC cases (Figure 2). nicity to determine if the relationship was consistent by each of Of the 5234 STEC cases, 437 (8.3%) had traveled internation- these variables. ally in the 7 days before symptom onset, 4455 (85.1%) had not, Finally, STEC cases were separated into international travel– and travel status was unknown for 342 (6.5%). Table  2 shows associated cases and cases acquired in the United States to overall age-adjusted IRRs of lower to high (≥20%) poverty determine how the overall findings of the relationship of census groups for each STEC outcome by international travel status. tract poverty level to STEC and its severe outcomes were influ- For cases in each travel group (international travel and domes- enced by international travel. International travel–associated tic exposure only), infection was associated with living in lower cases were those occurring in a person who had been out of the poverty census tracts. Point estimates of the relative rates were country at any time during the 7 days before illness onset. greater than 1.0 for each age group for both domestic and inter- Statistical analyses were performed using SAS, version 9.3 nationally acquired cases, although for some groups with small (SAS Institute Inc, Cary, NC, USA). IRRs and 95% confidence numbers of cases, this was not statistically significant (data not intervals were calculated using the Statcalc function in Epi shown). Info 7. Because most cases were acquired from domestic exposure, domestic cases were examined by serogroup (O157 vs non- RESULTS O157), age, and surveillance site. The findings were consistent with those from the analyses of all STEC cases: for all groups, There were a total of 5234 cases (96% of total) that were able to sites, and serogroups, the rate of STEC was higher in the less be geocoded to the rooftop level. The characteristics of these poor census tracts than the poorest census tracts and of a simi- cases and average annual incidence of each of the 3 outcomes lar magnitude to that using all STEC cases regardless of interna- are shown in Table 1. Overall, 26.3% were hospitalized and 5.9% tional travel status (data not shown). had HUS. The majority (55%) of STEC cases were non-O157. Crude incidence rates of all 3 outcomes were highest among DISCUSSION the youngest age groups, non-Hispanic whites, residents of lower poverty census tracts, and in the later years of the 5-year This study had several important findings. First, it demon- period. Numbers of cases by site ranged from a low of 223 in strated that persons in the United States of higher SES status, New Mexico to a high of 1274 in Minnesota. Site-specific inci- including very young children, have had a consistently higher dence ranged from a low of 0.99 in Georgia to a high of 4.74 in risk of acquiring clinically consequential STEC infection than Minnesota (data not shown). those living in poverty. Second, the findings were similar for Figure 1 shows age-adjusted rates of STEC O157, STEC non- O157 and non-O157 STEC. These findings have implications O157, and all STEC by census tract poverty level. Overall and for both current intervention efforts and future research. for each group of serotypes, the highest poverty group (≥20%) Before this analysis, there have been 3 published analyses had the lowest rate, and it was statistically significantly lower of the relationship of SES to STEC in the United States, all than for each of the other poverty groups (P < .001). There was using area-based SES measures, all with similar findings. The no consistent hierarchical relationship of incidence rate within first was a national study from the National Notifiable Disease the other poverty groups. Based on these findings, we combined System covering nationally reported data from 1993–2002. O157 and non-O157 serotypes and collapsed the 3 lowest pov- Using counties as the area-based unit of analysis, it found lower erty groups into 1 group <20% below the federal poverty level county-level incidence with higher county poverty levels [17]. for subsequent analyses. However, its findings were limited. It analyzed nationally noti- Age-adjusted IRRs comparing incidence of all STEC, hos- fiable data for E. coli O157:H7 (not all STEC) from 1993–2002, pitalized cases, and HUS cases in the combined lower poverty used a very broad area-based level, and only used county-level, group with the highest poverty group are shown in Table  2. not individual, variables. It was further limited in shedding For all STEC, IRRs were consistently significantly greater than no light on the contribution of neighborhood poverty to risk 1.0 for each age group, by sex, by surveillance year, and for all in specific age groups or on whether the risk associated with major race/ethnic groups except for non-Hispanic Asian/Pacific neighborhood poverty changed over time. The other 2 were the Census Tract Poverty and STEC Risk • OFID • 3 Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 Table 1. STEC Case Characteristics and Incidence by Severity, Serotype, and Selected Demographic Characteristics, FoodNet 2010–2014 a b c c c Characteristic No (%) of STEC Cases Population STEC Incidence Incidence of Hospitalized Cases Incidence of HUS All cases 5234 47 898 7 45 2.19 0.57 0.13 Serotype O157 2355 (45.0) 47 898 7 45 0.98 0.38 - Non-O157 2879 (55.0) 47 898 7 45 1.20 0.19 - Severity Hospitalized 1374 (26.3) 47 898 7 45 0.57 - - HUS 308 (5.9) 47 898 7 45 0.13 - - Age group, y 0–4 1267 (24.2) 2 986 919 8.48 1.65 0.92 5–17 1391 (26.6) 8 103 964 3.43 0.96 0.31 18–49 1615 (30.9) 20 897 307 1.55 0.34 0.02 50–64 455 (8.7) 9 508 512 0.96 0.33 0.01 ≥65 506 (9.7) 6 402 043 1.58 0.70 0.07 Sex Male 2331 (44.5) 23 513 733 1.98 0.52 0.12 Female 2903 (55.5) 24 385 012 2.38 0.62 0.14 Race/ethnicity Hispanic 536 (10.2) 5 253 918 2.04 0.32 0.08 Non-H white 3554 (67.9) 30 919 814 2.30 0.71 0.15 Non-H black 237 (4.5) 7 123 093 0.67 0.18 0.02 Non-H Asian/PI 116 (0.3) 2 277 853 1.02 0.25 0.10 Non-H other/unknown 791 (15.2) 1 382 468 - - - Census tract poverty <5% 1098 (21.0) 8 940 621 2.46 0.62 0.12 5%–<10% 1401 (26.8) 10 888 246 2.57 0.69 0.16 10%–<20% 1701 (32.5) 14 713 758 2.31 0.64 0.17 ≥20% 1031 (19.7) 12 415 515 1.66 0.40 0.08 Year 2010 826 (15.8) 47 898 7 45 1.72 0.52 0.11 2011 982 (18.8) 47 898 7 45 2.05 0.61 0.12 2012 1122 (21.4) 47 898 7 45 2.34 0.60 0.13 2013 1147 (21.9) 47 898 7 45 2.39 0.69 0.20 2014 1157 (22.1) 47 898 7 45 2.42 0.55 0.10 Abbreviations: H, Hispanic; HUS, hemolytic-uremic syndrome; PI, Pacific Islander; STEC, Shiga toxin–producing Escherichia coli. Three cases missing poverty status. Combined population denominator for all FoodNet sites. Based on 2012 US Census Bureau population estimates, except for race/ethnicity and census tract poverty, which are based on the 2010 US Census. Incidence per 100 000 person-years. more recent studies in Connecticut and New York City using appears to be universal across the United States and enduring. the census tract as the area size for analysis and individual Of interest, this relationship does not necessarily hold in other demographic variables rather than percentage of the population developed countries. A  study in Denmark covering national in selected age groups [6, 13]. They each found an association of data from 1993–2004 found no association between individ- lower STEC incidence with higher census tract poverty levels. ual income or education level and STEC [18], and a study from However, as they were local and in the Northeast (mostly urban Alberta, Canada, covering reported data from 2000–2002 found and suburban), it was not clear what would happen with more no association with census subdivision rates of STEC and per- nationally representative data. With our data set, we were able centage of individuals living in low-income households [19]. to confirm this relationship at the national level, that it was con- a Th t the overall findings for STEC O157 and STEC non- sistent across 5 consecutive years, strongest for young children, O157 were similar with respect to poverty is consistent with and present more recently than 1993–2002. Our findings were other aspects of their epidemiology. Although different STEC also consistent across all 10 FoodNet sites by sex and within serotypes may differ in pathogenicity [20, 21], those that cause most race/ethnic groups, and for each of the 3 STEC health severe human disease appear to share a similar ecology and outcomes examined (laboratory-confirmed disease, hospital- many of the same risk factors for acquisition [20–22]. This is ization, and HUS), the relationship between STEC and SES a likely explanation for why they have a similar relationship to 4 • OFID • Hadler et al Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 Table 2. Age-Adjusted Incidence Rate Ratios of Lower (<20% Below Federal Poverty Level) to High (≥20%) Poverty Groups of All STEC, Hospitalized STEC Cases, and Hemolytic-Uremic Syndrome Cases by Demographic Variables, Year, and International Travel Status, FoodNet Sites, 2010–2014 All STEC (n = 5234) Hospitalized Cases (n = 1374) HUS Cases (n = 308) Variable IRR 95% CI IRR 95% CI IRR 95% CI * * * All cases 1.53 1.43–1.64 1.63 1.43–1.87 2.08 1.53–2.81 Age group, y * * * 0–4 1.39 1.22–1.58 1.98 1.43–2.75 2.25 1.42–3.54 * * 5–17 1.69 1.47–1.94 1.95 1.48–2.57 1.71 1.07–2.73 * * 18–49 1.69 1.49–1.91 1.79 1.37–2.35 2.11 0.61–7.22 50–64 1.27 1.01–1.62 0.93 0.65–1.34 1.76 0.21–14.66 ≥65 1.26 1.01–1.57 1.19 0.86–1.65 6.13 0.88–45.65 Sex * * * Male 1.42 1.28–1.57 1.84 1.49–2.26 2.83 1.71–4.68 * * Female 1.56 1.43–1.71 1.48 1.24–1.77 1.60 1.10–2.34 Race/ethnicity Hispanic 1.62 1.32–2.00 1.71 1.07–2.74 3.20 0.91–11.22 * * Non-H white 1.20 1.10–1.30 1.37 1.16–1.61 1.34 0.98–1.85 Non-H black 1.61 1.23–2.11 1.24 0.75–2.04 0.90 0.18–4.48 Non-H Asian/PI 0.71 0.48–1.06 1.43 0.50–4.13 2.87 0.37–22.04 Year * * 2010 1.98 1.64–2.38 1.84 1.32–2.57 1.03 0.56–1.89 2011 1.22 1.06–1.42 1.30 NS 2.33 1.11–4.92 * * * 2012 1.59 1.37–1.84 1.69 1.24–2.29 2.56 1.22–5.37 * * * 2013 1.71 1.47–1.98 1.89 1.41–2.53 2.11 1.22–3.65 * * * 2014 1.52 1.32–1.76 1.57 1.15–2.14 4.04 1.45–11.24 International travel 7 d before symptom onset * * * No (n = 4454) 1.52 1.42–1.64 1.60 1.39–1.83 1.81 1.35–2.42 Yes (n = 435) 1.72 1.35–2.20 2.64 0.79–8.81 - - Abbreviations: CI, confidence interval; H, Hispanic; HUS, hemolytic-uremic syndrome; IRR, incidence rate ratio; PI, Pacific Islander; STEC, Shiga toxin–producing Escherichia coli. P < .01. poverty in the United States, as differences in rates of exposure diarrhea, and to get diagnosed than those in high poverty cen- to STEC are the likely reasons for those in high poverty census sus tracts. This consideration was examined for FoodNet sites tracts to be at lower risk in the United States. during 2000–2003 and found not to be true [23]. In fact, this One important consideration in interpretation of results from analysis of FoodNet population survey data found that those this study is that persons in higher SES census tracts might be in the lowest income category (household income  <$25 000) more likely to seek health care for STEC symptoms, especially were more likely to seek care for acute diarrheal illness. As this All STEC Hospitalized only 2.5 * * * * 1.5 0.5 CA CO CT GA MD MN NM NY OR TN All sites * P ≤ .01 that IRR is >1.0 Figure 2. Age-adjusted incidence rate ratios (IRRs) of lower to high poverty groups of all STEC and hospitalized Shiga toxin–producing Escherichia coli (STEC) cases, by Foodborne Disease Active Surveillance Network site, 2010–2014. Census Tract Poverty and STEC Risk • OFID • 5 IRR Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 analysis took place before our study time period, it is possible findings (eg, some poor neighborhoods may have a high prev- that health care–seeking behaviors might have changed. Thus, alence of relatively bacteria-free fast food and lack stores with in addition to laboratory-confirmed, largely outpatient illness, fresh meat and produce, lowering the potential for foodborne we examined the outcomes of hospitalization and HUS, both STEC exposure) and may have a stronger association if the likely to be less influenced by income or health care access. For neighborhood SES is more of a factor than individual-level SES. both outcomes, the association with higher SES was, if any- In addition, not everyone in the neighborhood has the same thing, even stronger. individual SES. Thus, the findings need to be interpreted in this An important risk factor for acquisition of STEC in the context. Second, most of the cases were identified because they United States, especially STEC non-O157, is international travel sought medical care, and stool specimens were taken for diag- [20, 21]. In our study, 8.9% of STEC cases had traveled inter- nosis. Not everyone with STEC infection seeks medical care, nationally in the 7  days before illness onset, and international and not all those who do have specific diagnostic testing. Thus, travel–associated STEC was least likely among persons living the cases used in this population-based study do not comprise in high poverty census tracts. However, aer ex ft cluding interna- the universe of all STEC infections in FoodNet sites [1]. To the tional travel, those in high poverty areas still had the lowest risk extent that there might have been bias by SES in who sought of laboratory-confirmed STEC and its complications. medical care and was diagnosed, the true magnitude of the er Th e are some data to support the hypothesis that adults of associations found could be different in either direction. Third, higher SES status in the United States have had a higher prev- we used relatively broad age groups to age-adjust and to exam- alence of some consumer-level STEC risks other than interna- ine age group–specific associations. However, those ages com- tional travel. Such risks include eating raw or undercooked beef prising each age group were selected because they had a similar [24–26], eating in restaurants [25], eating raw fresh vegetables, age-specific incidence. In addition, there were several statistical fruits, and nuts [27], and poorer hygiene practices to prevent limitations. As previously described, we made a post hoc recat- cross-contamination from raw products despite higher knowl- egorization of 4 census tract poverty levels to 2; thus, only P edge levels [26]. Although more current studies are needed, as values of ≤.01 should be considered statistically significant. In stated in a recent systematic review, “SES should be considered addition, the 95% confidence limits on the incidence rate ratios when targeting consumer level public health interventions for presented in Table  2 were calculated assuming there was no foodborne pathogens,” including STEC [28]. clustering of outcomes within individual census tracts as we did Risk factors for young children (<5 years) in the United States not have individual census tract identifiers; thus, they are pos- have been identified but have not been put in a relative attribu- sibly narrower than they might actually be. Finally, this study tion context or examined by SES, despite these children having was limited to descriptively defining the relationship between the highest age-specific risk of laboratory-confirmed infection census tract poverty used as a single recommended SES surveil- and HUS. Widely recognized means of exposure for young lance variable and STEC incidence in the national foodborne children include person-to-person transmission, particularly disease sentinel surveillance system, FoodNet. Examination of in day care centers and from contact with other children with other area-based SES variables and conducting an analysis to diarrhea, consumption of contaminated foods and beverages, determine the relative importance of census tract poverty com- particularly undercooked beef products and other food items pared with each of the other descriptive variables available were cross-contaminated from them, and direct or indirect contact beyond the scope of this analysis. with farm animals, particularly ruminants. Although all these CONCLUSIONS potential exposure factors could be more common in those In summary, the findings from analysis of population-based of higher SES, documentation of their association with SES data from FoodNet, the national sentinel foodborne disease first, and then of the factors leading to exposure (eg, hygiene surveillance system, confirm those from more localized stud- in day care centers and at home, including handling raw beef ), ies that people living in lower poverty neighborhoods are at is needed to guide prevention efforts targeted at the consumer. higher risk of acquiring STEC infection and suffering its more This study has some notable strengths. These include the high severe complications. These findings point to a need to consider percentage of cases geocoded to the census tract level, partici- higher SES when targeting specific public health interventions pation from all 10 FoodNet sites encompassing more than 48 to prevent STEC infection. However, a current understanding million people enabling generalization to the US population, of SES differences in risk factors, for children in particular, is and analysis by 3 levels of severity of STEC infection. needed to enable potentially effective interventions. er Th e are also some important limitations. Most impor - tantly, this study used census tract SES rather than individual Acknowledgments SES. Although the 2 are usually correlated, they do not measure We thank the many FoodNet surveillance staff at each site and at the exactly the same thing. In particular, census tract SES includes CDC for their work in collecting, geocoding, and collating the data used in this paper. possible neighborhood SES factors contributing to the observed 6 • OFID • Hadler et al Downloaded from https://academic.oup.com/ofid/article-abstract/5/7/ofy148/5047963 by guest on 16 October 2019 Financial support. This work was supported by the Centers for Disease 13. Whitney BM, Mainero C, Humes E, Hurd S, Niccolai L, Hadler JL. Socioeconomic status and foodborne pathogens in Connecticut, USA, 2000–2011. Emerg Infect Control and Prevention (CI05-026, CK12-1202, CK17-1701). Dis 2015; 21:1617–24. Potential conifl cts of interest. All authors: no reported conflicts of 14. Henao OL, Jones TF, Vugia DJ, Griffin PM; Foodborne Diseases Active Surveillance interest. All authors have submitted the ICMJE Form for Disclosure of Network (FoodNet) Workgroup. Foodborne diseases active surveillance network-2 Potential Conflicts of Interest. Conflicts that the editors consider relevant to decades of achievements, 1996–2015. Emerg Infect Dis 2015; 21:1529–36. the content of the manuscript have been disclosed. 15. US Census Bureau. American fact finder advanced. Available at: https://fact- finder.census.gov/faces/nav/jsf/pages/searchresults.xhtml?refresh=t. Accessed 8 March 2018. References 16. Public Health Disparities Geocoding Project. Available at: http://www.hsph.har- 1. Scallan E, Hoekstra RM, Angulo FJ, et  al. Foodborne illness acquired in the vard.edu/thegeocodingproject. Accessed 8 March 2018. United States–major pathogens. Emerg Infect Dis 2011; 17:7–15. 17. Chang M, Groseclose SL, Zaidi AA, Braden CR. An ecological analysis of sociode- 2. Gould LH, Mody RK, Ong KL, et  al. Increased recognition of non-O157 Shiga mographic factors associated with the incidence of salmonellosis, shigellosis, and toxin-producing Escherichia coli infections in the United States during 2000– E. coli O157:H7 infections in US counties. Epidemiol Infect 2009; 137:810–20. 2010: epidemiologic features and comparison with E.  coli O157 infections. 18. Simonsen J, Frisch M, Ethelberg S. Socioeconomic risk factors for bacterial gas- Foodborne Pathog Dis 2013; 10:453–60. trointestinal infections. Epidemiology 2008; 19:282–90. 3. Adams DA, Thomas KR, Jajosky R, et  al. Summary of notifiable infectious dis- 19. Pearl DL, Louie M, Chui L, et  al. A multi-level approach for investigating eases and conditions—United States, 2014. MMWR Morb Mortal Wkly Rep socio-economic and agricultural risk factors associated with rates of reported 2016; 63:1–152. cases of Escherichia coli O157 in humans in Alberta, Canada. Zoonoses Public 4. Krieger N, Chen JT, Waterman PD, et  al. Painting a truer picture of US socio- Health 2009; 56:455–64. economic and racial/ethnic health inequalities: the Public Health Disparities 20. Hedican E, Medus C, Besser JM, et al. Characteristics of O157 versus non-O157 Geocoding Project. Am J Public Health 2005; 95:312–23. Shiga toxin-producing Escherichia coli infections in Minnesota, 2000–2006. Clin 5. Toprani A, Hadler JL. Selecting and applying a standard area-based socioeco- Infect Dis 2009; 49:358–64. nomic status measure for public health data: analysis for New York City. New 21. Hadler JL, Clogher P, Phan Q, et al. Ten year trends and risk factors for non-O157 York City Department of Health and Mental Hygiene: Epi Research Report. Shiga-toxin producing E.  coli found through Shiga-toxin testing, Connecticut, 2013. Available at: http://www1.nyc.gov/assets/doh/downloads/pdf/epi/epire- 2000–2009. Clin Infect Dis 2011; 53:269–76. search-SES-measure.pdf. Accessed 8 March 2018. 22. Bettelheim KA. The non-O157 shiga-toxigenic (verocytotoxigenic) Escherichia 6. Greene SK, Levin-Rector A, Hadler JL, Fine AD. Disparities in reportable com- coli; under-rated pathogens. Crit Rev Microbiol 2007; 33:67–87. municable indicators by census tract-level poverty, New York City, 2006–2013. 23. Scallan E, Jones TB, Cronquist A, et al. Factors associated with seeking medical Am J Public Health 2015; 105:e27–34. care and submitting a stool sample in estimating the burden of foodborne illness. 7. Centers for Disease Control and Prevention. Social determinants of health among Foodborne Pathog Dis 2006; 3:432–8. adults diagnosed with HIV infection in 11 states, the District of Columbia and 24. Taylor EV, Holt KG, Mahon BE, Ayers T, Norton D, Gould LH. Ground beef con- Puerto Rico, 2013. HIV Surveillance Supplemental Report. 2015. http://www.cdc. sumption patterns in the United Sates, FoodNet, 2006 through 2007. J Food Prot gov/hiv/pdf/library/reports/surveillance/cdc-hiv-surveillancereport-vol20-no5. 2012; 75:341–6. pdf. Accessed 8 March 2018. 25. Wagner JA. Foodnet Population Survey 2006–2007: Differences in Prevalence of 8. Boscoe FP, Johnson CJ, Sherman RL, et  al. The relationship between area pov- Health Care Seeking Behaviors and Exposures to Risk Factors for Foodborne Illness erty rate and site-specific cancer incidence in the United States. Cancer 2014; by Socioeconomic Status [master’s thesis]. New Haven, CT: Yale University; 2014. 120:2191–8. Available at: http://elischolar.library.yale.edu/cgi/viewcontent.cgi?article=1304&- 9. Hadler JL, Vugia DJ, Bennett NM, Moore MR. Emerging infections program context=ysphtdl. Accessed 8 March 2018. efforts to address health equity. Emerg Infect Dis 2015; 21:1589–94. 26. Patil SR, Cates S, Morales R. Consumer food safety knowledge, practices, and 10. Hadler JL, Yousey-Hindes K, Perez A, et al. Influenza-related hospitalizations and demographic differences: findings from a meta-analysis. J Food Prot 2005; poverty levels—United States 2010–2012. MMWR MorbMortal Wkly Rep 2016; 68:1884–94. 65:101–5. 27. Dubowitz T, Heron M, Bird CE, et al. Neighborhood socioeconomic status and 11. Soto K, Petit S, Hadler JL. Changing disparities in invasive pneumococcal dis- fruit and vegetable intake among whites, blacks, and Mexican Americans in the ease by socioeconomic status and race in Connecticut, 1998–2008. Public Health United States. Am J Clin Nutr 2008; 87:1883–91. Reports 2011; 126(Suppl 3):81–8. 28. Newman KL, Leon JS, Rebolledo PA, Scallan E. The impact of socioeconomic sta- 12. Bemis K, Marcus R, Hadler JL. Campylobacteriosis incidence is associated with tus on foodborne illness in high-income countries: a systematic review. Epidemiol low poverty: Connecticut, 1999–2009. Emerg Infect Dis 2014; 20:1240–2. Infect 2015; 143:2473–85. Census Tract Poverty and STEC Risk • OFID • 7

Journal

Open Forum Infectious Diseases – Oxford University Press

Published: Jul 1, 2018

There are no references for this article.