Factors Associated with Regulatory Action Involving Investigation of Illnesses Associated with Shiga Toxin–Producing Escherichia coli in Products Regulated by the Food Safety and Inspection Service

Abstract

We described characteristics of the Escherichia coli O157 and Escherichia coli non-O157 illness investigations conducted by the United States Department of Agriculture's Food Safety and Inspection Service (FSIS) during the 5-year period from 2006 through 2010. We created a multivariable logistic regression model to determine characteristics of these investigations that were associated with FSIS regulatory action, which was defined as having occurred if a product recall occurred or if FSIS personnel performed an environmental health assessment (Food Safety Assessment) at the implicated establishment. During this period, FSIS took regulatory action in 38 of 88 (43%) investigations. Illness investigations in which FoodNet states were involved were more likely to result in regulatory action. Illness investigations in which state and local traceback, or FSIS traceback occurred were more likely to result in regulatory action. Reasons for lack of action included evidence of cross-contamination after the product left a regulated establishment, delayed notification, lack of epidemiological information, and insufficient product information.

Introduction

Infection with Shiga toxin–producing Escherichia coli O157:H7 (STEC O157) is estimated to result in 96,000 illnesses each year in the United States (Scallan et al., 2011). Infection with non-O157 STEC results in nearly 113,000 illnesses each year (Batz et al., 2012). Many STEC infections are foodborne (Kassenborg et al., 2004; Laine et al., 2005; Rangel et al., 2005; CDC, 2006; Lynch et al., 2009). Cattle and other ruminants are significant reservoirs for STEC O157 (Ferens and Hovde, 2011). Carriage varies seasonally, peaking in the summertime (Ferens and Hovde, 2011). Contact with contaminated water (Clark et al., 2010), animal contact (NASPHV, 2007; Clark et al., 2010; Mora et al., 2012), and person-to-person transmission (Seto et al., 2007; Snedeker et al., 2009; Januszkiewicz et al., 2012) are additional routes of exposure.

PulseNet, the national molecular network for foodborne disease surveillance coordinated by the Centers for Disease Control and Prevention (CDC), enhances early outbreak detection in addition to the detection of nationally dispersed outbreaks (Gerner-Smidt et al., 2005; Tauxe, 2006). However, clinical laboratories vary widely in testing practices regarding STEC culture and toxin screening: some laboratories test all specimens, others test only those that contain blood, and some test only at the request of a physician (Voetsch et al., 2004). Rates of outbreak reporting and investigation vary substantially by state (Jones et al., 2013).

STEC O157 infections became nationally notifiable in the United States in 1994, the same year that Food Safety and Inspection Service (FSIS) declared STEC O157 as an adulterant in certain classes of beef products. Non-O157 STEC infections became reportable in 2000 (Gould, 2008). The U.S. Department of Agriculture FSIS coordinates with CDC and state and local public health partners to investigate STEC O157 and non-O157 illness clusters that may be linked to FSIS-regulated food products, which include meat, poultry, and egg products (the U.S. Food and Drug Administration [FDA] regulates other food types including shell eggs as well as pet food and animal feed). Notification comes from a variety of sources, including PulseNet, CDC, state public health partners, media outlets, and other federal agencies such as the FDA. Regulatory action in response to a foodborne outbreak takes place when there is a combination of epidemiological, microbiological, and product traceback evidence sufficient to implicate a specific food or establishment.

The purposes of this study were twofold. The first objective was to describe the characteristics of the STEC illness investigations conducted by FSIS during the 5-year period from 2006 through 2010. The second was to determine, using both univariable and multivariable analyses, whether specific factors may be associated with an increased likelihood of FSIS regulatory action.

Materials and Methods

FSIS receives notification regarding pathogen-related illnesses (predominantly STECs, Salmonella, Listeria, and Campylobacter) from a number of sources, including CDC, PulseNet, and state and local public health partners. Key information is tracked at the investigation level; for each of these, the assigned lead epidemiologist enters findings and updates into a centralized database.

FSIS may take action on even a single well-documented case with a plausible link to FSIS-regulated product. An illness investigation was defined as a public health action initiated in response to one or more cases of O157 or non-O157 STEC infection with isolates of the same XbaI pulsed-field gel electrophoresis (PFGE) subtype and plausible FSIS-regulated product exposure. Regulatory action was defined as having occurred if a product recall occurred or if FSIS personnel performed an environmental health assessment (Food Safety Assessment) at the implicated establishment. In most instances, more than one of these regulatory actions resulted from a given illness investigation.

Cases of O157 and non-O157 STEC infection reported to FSIS with earliest onset after December 31, 2005 and prior to December 31, 2010 were included in the study. Illness investigations definitively associated with exposures not under FSIS jurisdiction, such as lettuce, sprouts, mixed nuts, cookie dough, cheese, day-care exposure, animal exposure, custom slaughter, and contaminated water, were excluded from the analysis. We also excluded investigations in which no common exposure could be identified.

Univariable analyses were performed to calculate odd ratios (ORs) and 95% confidence intervals (CIs) describing crude associations between variables of interest and the outcome of FSIS regulatory action. Numeric variables were categorized by quartile to look for natural breakpoints, and dichotomized at the median if appropriate. Variables of interest included frequency of PFGE XbaI subtype, number of states reporting cases, number of cases, number of cases hospitalized, duration of investigation, seasonality, whether a state participating in the Foodborne Diseases Active Surveillance Network (FoodNet) was part of the investigation, and whether state or local traceback or FSIS retail traceback (consisting of determination of establishment origin and production date range) occurred.

The FSIS Outbreaks Section Eastern Laboratory utilizes a categorical naming system for STEC isolates: pattern combinations that occur in PulseNet once or more during a 60-day period are considered “common,” those that occur on average once every 61–120 days are “less common,” every 121–240 days are “rare,” every 241 or more days to two occurrences in PulseNet are considered “very rare,” and those with one appearance in PulseNet are “new.” STEC investigations with XbaI PFGE subtype patterns classified as common and less common were compared with combinations classified as rare, very rare, or new. Investigations associated with more than one XbaI pattern were not included in the frequency analysis.

Duration was defined as the number of days between the earliest illness onset and the latest illness onset. Investigations in which the earliest illness onset was between June and September were classified as occurring during the high season. Investigations in which the first illness onset was between October and May were classified as occurring during the low season.

FoodNet, a collaborative effort by CDC, 10 state health departments, FSIS, and FDA, performs active surveillance for O157 and non-O157 STEC. FoodNet programs within state health departments and collaborating academic institutions have enhanced capacity to respond to outbreaks (Jones et al., 2007). The relationship between having a FoodNet site as part of an illness outbreak investigation and FSIS regulatory action was examined.

In order to estimate statistical associations between variables significant in the univariable analysis, an α<0.20 was selected as the cutoff for eligibility for the multivariable logistic regression model. Eligible variables were examined for collinearity. SAS software version 9.2 (SAS Institute, Cary, NC) was used for all analyses. Using the Wald statistic, variables were removed from the model via stepwise backward elimination until those that remained were significant at p≤0.05.

Results

During 2006–2010, 139 STEC illness clusters (defined as groups of case-patients with the same XbaI PFGE pattern and varying degrees of epidemiologic data) possibly associated with FSIS-regulated products were reported to FSIS. Eighteen of these were excluded from analysis due to documented links to FDA-regulated products, day care exposure, contaminated water, livestock exposure, or custom slaughter. An additional 29 were excluded from analysis because no common exposure could be determined. Finally, four more were excluded because PFGE data were not available.

In the remaining 88 illness investigations (Tables 1 and 2), there were a total of 1195 illnesses (as documented by clinical isolates). FSIS regulatory action was taken in 38 (43%) of these investigations, accounting for 748 (63%) of the case-patients (as documented by clinical isolates), including 29 recalls and 36 food safety assessments: in 27 investigations, both a recall and a food safety assessment occurred. Two recalls occurred without food safety assessments, and nine food safety assessments occurred without recalls. In the 50 investigations for which regulatory action was not taken, documented reasons for lack of regulatory action included evidence of cross-contamination after the product left a regulated establishment, lack of epidemiological information, and insufficient product information. Due to resource shortages or lack of case-patient response, detailed case-patient food histories could not always be obtained by local, state, or federal public health partners. Some case-patients were unable to recall specific food history information during the time period of interest. When insufficient product information was an issue, generally information such as documented purchase history (receipts, shopper card history) or brand and lot code was not available.

Table 1.

Selected Descriptive Characteristics for Shiga Toxin–Producing Escherichia coli Illness Investigations Conducted by Food Safety and Inspection Service, United States, 2006–2010 ^a

Characteristics	Number (%) investigations	Number (%) investigations in category with regulatory action
PFGE subtype
Common	31 (35)	14 (45)
Less common	2 (2)	1 (50)
Rare	32 (36)	13 (41)
Very rare	3 (3)	3 (100)
New	11 (13)	4 (36)
Multiple PFGE patterns	7 (8)	3 (43)
Categorization missing	1 (1)	0 (0)
Total	88 (100)	38 (43)
Number states in cluster
1	35 (40)	7 (20)
2	19 (22)	9 (47)
3–4	14 (16)	7 (50)
5–17	20 (23)	15 (75)
Total	88 (100)	38 (43)
Number cases in cluster
1–4	25 (28)	8 (32)
5–7	21 (24)	7 (33)
8–17	21 (24)	8 (38)
18–117	21 (24)	15 (71)
Total	88 (100)	38 (43)
Number cases hospitalized
0–1	17 (19)	11(65)
2–3	18 (20)	8 (44)
4–5	8 (9)	3 (38)
6–21	12 (14)	11 (92)
Missing data	33 (38)	5 (15)
Total	88 (100)	38 (43)
Number cases HUS
0	24 (27)	15 (63)
1	12 (14)	4 (33)
2	5 (6)	3 (60)
3–6	7 (8)	6 (86)
Missing data	40 (45)	10 (25)
Total	88 (100)	38 (43)
Duration
1–6 days	16 (18)	4 (25)
7–19 days	19 (22)	9 (47)
20–37 days	20 (23)	6 (30)
38–122 days	19 (22)	14 (74)
Missing data	14 (16)	5 (36)
Total	88 (100)	38 (43)
Season
Jan–March	15 (17)	7 (47)
Apr–June	33 (38)	14 (42)
July–Sept	29 (33)	11 (38)
Oct–November	11 (13)	6 (55)
Total	88 (100)	38 (43)

Quartiles used as breakpoints for continuous variables.

PFGE, pulsed-field gel electrophoresis; HUS, hemolytic uremic syndrome.

Table 2.

Univariable Association Between STEC Characteristic and Regulatory Action Regarding USDA-FSIS-Regulated Product, United States, 2006–2010

	Investigations with regulatory action	Investigations without regulatory action	Percent with regulatory action	OR	95% CI
PFGE subtype
Common/less common	15	18	45	Referent
Very rare, rare	20	26	43	0.9	0.4–2.3
New
Total	35	44	44
Number states in investigation^a
1–2	16	38	30	Referent
3–17	22	12	65	4.4	1.7–10.9
Total	38	50	43
Number cases in investigation^a
1–7	15	31	33	Referent
8–117	23	19	55	2.5	1.1–5.9
Total	38	50	43
Number cases hospitalized
0–3	19	16	54	Referent
4–21	14	6	70	2.0	0.6–6.3
Total	33	22	60
HUS cases
0	15	9	63	Referent
1 or more	13	11	54	0.7	0.2–2.2
Total	28	20	58
Duration
0–19 days	18	31	37	Referent
20–122 days	20	19	51	1.8	0.8–4.3
Total	38	50	43
Season
October–March	13	13	50	1.5	0.6–3.7
April–September	25	37	40	Referent
Total	38	50	43
State or local traceback^a
Yes	22	15	59	3.2	1.3–7.8
No	16	35	31	Referent
Total	38	50	43
FSIS retail traceback^a
Yes	31	21	59	6.1	2.3–16.5
No	7	29	19	Referent
Total	38	50	43
FoodNet state involved in Outbreak^a
Yes	32	29	51	3.9	1.4–10.9
No	6	21	22	Referent
Total	38	50	43

Significant chi-square test p<0.05.

OR, odds ratio; CI, confidence interval; PFGE, pulsed-field gel electrophoresis; HUS, hemolytic uremic syndrome; FSIS, Food Safety and Inspection Service.

Of the STEC investigations, 86 were E. coli O157:H7; 1 was E. coli O103, and 1 was E. coli O26. Excluding seven investigations associated with multiple XbaI patterns, 47 PFGE subtypes were observed. CDC PFGE subtype designations EXHX01.0047 (15.9%, n=14), EXHX01.0074 (8.0%, n=7), EXHX01.0200 (6.8%, n=6), and EXHX01.0008 (4.6%, n=4) were the most common XbaI PFGE subtypes associated with the investigations conducted by FSIS. In addition to these, one pattern was represented three times, five patterns were represented twice, and 37 patterns were represented once each. Investigation of illnesses associated with uncommon PFGE subtypes was not more likely to be significantly associated with regulatory action than investigation of common subtype clusters (OR 0.9, 95% CI: 0.4–2.3).

State or local traceback occurred for 37 (42%) of the investigations, and for 22 (58%) of those in which regulatory action occurred (OR 3.2, 95% CI: 1.3–7.8). FSIS retail traceback occurred for 52 (59%) of the investigations, and for 31 (82%) of those in which regulatory action occurred (OR 6.1, 95% CI: 2.3–16.5). FoodNet states were involved in 61 (69%) of the investigations, and 32 (84%) of investigations in which regulatory action occurred (OR 3.9, 95% CI: 1.4–10.9).

The median duration of an investigation was 19 days (range 1–122 days). The association between longer investigations and regulatory action was eligible for model entry at p=0.17 (OR 1.8, 95% CI: 0.8–4.3). While 70% of investigations occurred during the April–September season, as compared with 30% in October–March, the association between season and regulatory action was not significant (OR 1.5 for low season, 95% CI: 0.6–3.7).

The median number of states involved in an investigation was 2 (range 1–17). Investigations involving 3–17 states were significantly more likely to be associated with regulatory action than were investigations involving 1–2 states (OR 4.4, 95% CI: 1.7–10.9). The median number of cases hospitalized was three (range 0–21); the median number of hemolytic uremic syndrome (HUS) cases per investigation was <1; however, the substantial amount of missing hospitalization and HUS data limits the interpretation of these findings. Investigations involving more than three hospitalized case-patients were not significantly more likely to be associated with regulatory action (OR 2.0, 95% CI: 0.6–6.3). The median number of cases was 7 (range 1–117). Investigations involving eight or more cases were significantly more likely to be associated with regulatory action than were investigations involving 1–7 case-patients (OR 4.8, 95% CI: 1.6–14.0).

The variables for duration of investigation, cases, and states were collinear. Of these, the variable for number of cases involved in an investigation was selected for inclusion in stepwise backward elimination regression modeling.

The initial logistic regression model included categorical variables for number of cases in an investigation, FoodNet state in an investigation, local traceback, and FSIS retail traceback. Presence of FoodNet state, local traceback, and FSIS retail traceback remained associated with FSIS regulatory action in the final model (Table 3). The adjusted odds of regulatory action were 3.9 times higher (95% CI: 1.2–12.8) with FoodNet state involvement, 3.9 times higher (95% CI: 1.4–10.8) with local traceback, and 5.5 times higher (95% CI: 1.9–16.1) with FSIS retail traceback.

Table 3.

Final Model—Adjusted Association Between Shiga Toxin–Producing Escherichia coli Characteristic and Regulatory Action Regarding Food Safety and Inspection Service–Regulated Product, United States, 2006–2010

Characteristic	OR	95% CI
FoodNet state involved
Yes	3.9	1.2–12.8
No	Referent
FSIS retail traceback
Yes	5.5	1.9–16.1
No	Referent
State or local traceback
Yes	3.9	1.4–10.8
No	Referent

OR, odds ratio; CI, confidence interval.

Discussion

During the 5-year period from 2006–2010, 88 STEC investigations were conducted and FSIS regulatory action was taken in 38 (43%) of them. Documented reasons for lack of regulatory action included evidence of cross-contamination after the product left a regulated establishment, lack of epidemiological information, and insufficient product information. As regulatory action in response to a foodborne outbreak takes place when there is a combination of epidemiological, microbiological, and product traceback evidence sufficient to implicate a specific food or establishment, collection of detailed food history and product type information including brand, purchase date, and shopper history records can enhance the likelihood of regulatory action. Additionally, state or federal laboratory testing of suspect leftover products in original packaging from case-patients' homes can assist in information-gathering, resulting in regulatory action. Documented handling and chain of custody for leftover products during transport for testing promote regulatory action when products are STEC-positive with isolates indistinguishable from the outbreak strain.

PFGE frequency was not found to be associated with regulatory action. While standardized subtyping has greatly enhanced the detection of geographically dispersed foodborne outbreaks, epidemiological data remain a critical part of outbreak investigation. The results of this study are consistent with others suggesting that PFGE subtype frequency should not be a deciding factor in conducting an investigation (Hedberg and Besser, 2006; Rounds et al., 2012). Recent estimates suggest that an average of 26 cases of STEC O157 and an average of 107 cases of STEC non-O157 occur for each that is reported (Scallan et al., 2011). The Council of State and Territorial Epidemiologists recommends a list of disease-specific data elements to be collected routinely by all states and territories for each identified STEC case that can be interviewed. If resources are limited, priority should be given to STECs known to cause more severe disease, and any case that may be part of a recognized outbreak (CSTE, 2013).

Investigations involving three or more states and eight or more case-patients were univariably associated with an increased likelihood of regulatory action. Resources available for foodborne investigations vary substantially by state and season (Jones et al., 2013); fully 70% of the foodborne illness investigations included in this analysis occurred in the spring and summer. Local and state health resources may be taxed during this time. Timeliness of reporting has also been demonstrated to vary by state (Hedberg et al., 2008). However, investigations involving larger numbers of states and case-patients may provide a greater number of opportunities for obtaining the accurate, detailed case-patient interviews and food purchase histories in the form of shopper card releases and/or purchase receipts needed for epidemiologic investigation of foodborne illness. Additionally, investigations involving multiple states may assist in evaluating similarities and differences in the product supply chain, resulting in more precise traceback to the source of contamination.

Notably, the 10 FoodNet states were involved in 61 of the 88 investigations in this analysis (69%); though it is beyond the scope of this analysis, FoodNet-associated programs in these states are likely to have played a role in initial reporting. In the multivariable model, FoodNet state involvement was associated with an increased likelihood of regulatory action. While a variety of other, more recently established programs such as the Food and Drug Administration's Rapid Response Teams (U.S. Department of Health and Human Services, 2013) and CDC's Foodborne Diseases Centers for Outbreak Response Enhancement (FoodCORE) (CDC, 2013) also play a valuable role in current foodborne disease investigation efforts, the FoodNet program, established in 1996, was chosen as a variable with both historic and current relevance. FoodNet states are well situated with expert knowledge and resources to gather information needed for regulatory action (Jones et al., 2007).

Traceback is needed to determine whether a common establishment and lot or lots of product are associated with illness. State, local, and territorial health workers wishing to coordinate on traceback should ideally notify federal personnel as soon as assistance is sought. Coordination can enhance efforts, avoid duplication, and build relationships that will assist in future traceback efforts. Our findings suggest that traceback at multiple jurisdictional levels is needed to support regulatory action, which can include a recall when definitive lots of product are implicated, as well as other regulatory actions such as increased pathogen testing and a formal, scientific, in-plant review of hazards and critical control points and validation of pathogen interventions. FSIS Directives on traceback and Food Safety Assessment procedures are codified and publicly available (USDA, 2011a, b).

The results of this analysis are based on characteristics of all the states involved in FSIS STEC-related investigations during the 5-year period from 2006 to 2010. Since there is a great degree of variability across states in foodborne disease detection and investigation (Jones et al., 2013), these findings may not be applicable to individual jurisdictions investigating a STEC illness cluster.

Footnotes

Acknowledgments

We thank all of the state and local public health officials who were involved in these investigations.

Disclosure Statement

No competing financial interests exist.

References

Batz

, Hoffmann

, Morris

Jr . Ranking the disease burden of 14 pathogens in food sources in the United States using attribution data from outbreak investigations and expert elicitation. J Food Prot, 2012; 75:1278–1291.

[CDC] Centers for Disease Control and Prevention, Foodborne Diseases Centers for Outbreak Response Enhancement. 2013. Available at: http://www.cdc.gov/foodcore/about.html, accessed November 27, 2013 .

CDC. Ongoing multistate outbreak of Escherichia coli serotype O157:H7 infections associated with consumption of fresh spinach—United States, September 2006. MMWR Morb Mortal Wkly Rep, 2006; 55:1045–1046.

Clark

, Sontrop

, Macnab

, Salvadori

, Moist

, Suri

, Garg

. Long term risk for hypertension, renal impairment, and cardiovascular disease after gastroenteritis from drinking water contaminated with Escherichia coli O157:H7: A prospective cohort study. BMJ, 2010; 341:c6020.

[CSTE] Council of State and Territorial Epidemiologists, Infectious Disease Committee Update to Public Health Reporting for Shiga toxin–producing Escherichia coli (STEC). 2013. Available at: http://c.ymcdn.com/sites/www.cste.org/resource/resmgr/PS/13-ID-01.pdf, accessed January 17, 2014 .

Gould

. Emerging Infections Program FoodNet News, Volume 2, Issue 1, Winter 2008, Available at: http://www.cdc.gov/FoodNet/news/2008/January_FoodNet_News.pdf, accessed February 13, 2013 .

Ferens

, Hovde

. Escherichia coli O157:H7: Animal reservoir and sources of human infection. Foodborne Pathog Dis, 2011; 8:465–487.

Gerner-Smidt

, Kincaid

, Kubota

, Hise

, Hunter

, Fair

, Norton

, Woo-Ming

, Kurzynski

, Sotir

, Head

, Holt

, Swaminathan

. Molecular surveillance of Shiga toxigenic Escherichia coli O157 by PulseNet USA. J Food Prot, 2005; 68:926–931.

Hedberg

, Besser

. Commentary: Cluster evaluation, PulseNet, and public health practice. Foodborne Pathog Dis, 2006; 3:32–35.

10.

Hedberg

, Greenblatt

, Mtyas

, Lemmings

, Sharp

, Skibicki

, Liang

; Enteric Disease Investigation Timeline Study Work Group. Timeliness of enteric disease surveillance in 6 US states. Emerg Infect Dis, 2008; 14:311–313.

11.

Januszkiewicz

, Szych

, Rastawicki

, Wołkowicz

, Chróst

, Leszczyńska

, Kuźma

, Roszkowska-Blaim

, Gierczyński

. Molecular epidemiology of a household outbreak of Shiga-toxin–producing Escherichia coli in Poland due to secondary transmission of STEC O104:H4 from Germany. J Med Microbiol, 2012; 61:552–558.

12.

Jones

, Rosenberg

, Kubota

, Ingram

. Variability among states in investigating foodborne disease outbreaks. Foodborne Pathog Dis, 2013; 10:69–73.

13.

Jones

, Scallan

, Angulo

. FoodNet: Overview of a decade of achievement. Foodborne Pathog Dis, 2007; 4:60–66.

14.

Kassenborg

, Hedberg

, Hoekstra

, Evans

, Chin

, Marcus

, Vugia

, Smith

, Ahuja

, Slutsker

, Griffin

. Farm visits and undercooked hamburgers as major risk factors for sporadic Escherichia coli O157:H7 infections: Data from a case-control study in 5 FoodNet sites. Clin Infect Dis, 2004; 38:S271–S278.

15.

Laine

, Scheftel

, Boxrud

, Vought

, Danila

, Elfering

, Smith

. Outbreak of Escherichia coli O157:H7 infections associated with non-intact blade-tenderized frozen steaks sold by door-to-door vendors. J Food Prot, 2005; 68:1198–1202.

16.

Lynch

, Tauxe

, Hedberg

. The growing burden of foodborne outbreaks due to contaminated fresh produce: Risks and opportunities. Epidemiol Infect, 2009; 137:307–315.

17.

Mora

, Lopez

, Dhabi

, López-Beceiro

, Fidalgo

, Díaz

, Martínez-Carrasco

, Mamani

, Herrera

, Blanco

. Seropathotypes, phylogroups, stx subtypes, and intimin types of wildlife-carried, Shiga toxin–producing Escherichia coli strains with the same characteristics as human-pathogenic isolates. Appl Environ Microbiol, 2012; 78:2578–2585.

18.

[NASPHV] National Association of State Public Health Veterinarians, Inc.. Centers for Disease Control and Prevention (CDC), Council of State and Territorial Epidemiologists, American Veterinary Medical Association. Compendium of measures to prevent disease associated with animals in public settings, 2007: National Association of State Public Health Veterinarians, Inc. (NASPHV). MMWR Morb Mortal Wkly Rep, 2007;Jul 6-56(RR-5):1–14.

19.

Rangel

, Sparling

, Crowe

, Griffin

, Swerdlow

. Epidemiology of Escherichia coli O157:H7 outbreaks, United States, 1982–2002. Emerg Infect Dis, 2005; 11:603–609.

20.

Rounds

, Boxrud

, Jawahir

, Smith

. Dynamics of Escherichia coli O157:H7 outbreak detection and investigation, Minnesota 2000–2008. Epidemiol Infect, 2012; 140:1430–1438.

21.

Scallan

, Hoekstra

, Angulo

, Tauxe

, Widdowson

, Roy

, Jones

, Griffin

. Foodborne illness acquired in the United States—major pathogens. Emerg Infect Dis, 2011; 17:7–15.

22.

Seto

, Soller

, Colford

. Strategies to reduce person-to-person transmission during widespread Escherichia coli O157:H7 outbreak. Emerg Infect Dis, 2007; 13:860–866.

23.

Snedeker

, Shaw

, Locking

, Prescott

. Primary and secondary cases in Escherichia coli O157 outbreaks: A statistical analysis. BMC Infect Dis, 2009; 9:144.

24.

Tauxe

. Molecular subtyping and the transformation of public health. Foodborne Pathog Dis, 2006; 3:4–8.

25.

[USDA] United States Department of Agriculture Food Safety and Inspection Service FSIS Directive 5100.1, Revision 3: Enforcement, Investigations, and Analysis Officer (EIAO) Food Safety Assessment Methodology. 2011a. Available at: http://www.fsis.usda.gov/wps/portal/fsis/topics/regulatory-compliance/food-safety-assessments, accessed January 18, 2014 .

26.

USDA. Food Safety and Inspection Service, FSIS Directive 8010.3, Revision 3: Procedures for Evidence Collection, Safeguarding and Disposal. 2011b. Available at: http://www.fsis.usda.gov/wps/wcm/connect/776dbd09-1e66-488c-b95c-75d15d399f89/8010_58.pdf?MOD=AJPERES, accessed January 18, 2014 .

27.

U.S. Department of Health and Human Services, Food and Drug Administration, Rapid Response Teams. 2013. Available at: http://www.fda.gov/ForFederalStateandLocalOfficials/CooperativeAgreementsCRADAsGrants/ucm297407.htm, accessed November 27, 2013 .

28.

Voetsch

, Angulo

, Rabatsky-Ehr

, Shallow

, Cassidy

, Thomas

, Swanson

, Zansky

, Hawkins

, Jones

, Shillam

, Van Gilder

, Wells

, Griffin

; Emerging Infections Program FoodNet Working Group. Laboratory practices for stool-specimen culture for bacterial pathogens, including Escherichia coli O157:H7, in the FoodNet sites, 1995–2000. Clin Infect Dis, 2004; 38:S190–S197.