Shigella sonnei Outbreak Investigation During a Municipal Water Crisis—Genesee and Saginaw Counties, Michigan, 2016

Abstract

Objectives. To investigate a shigellosis outbreak in Genesee County, Michigan (including the City of Flint), and Saginaw County, Michigan, in 2016 and address community concerns about the role of the Flint water system.

Methods. We met frequently with community members to understand concerns and develop the investigation. We surveyed households affected by the outbreak, analyzed Shigella isolate data, examined the geospatial distribution of cases, and reviewed available water quality data.

Results. We surveyed 83 households containing 158 cases; median age was 10 years. Index case-patients from 55 of 83 households (66%) reported contact with a person outside their household who wore diapers or who had diarrhea in the week before becoming ill; results were similar regardless of household drinking water source. Genomic diversity was not consistent with a point source. In Flint, no space-time clustering was identified, and average free chlorine residual values remained above recommended levels throughout the outbreak period.

Conclusions. The outbreak was most likely caused by person-to-person contact and not by the Flint water system. Consistent community engagement was essential to the design and implementation of the investigation.

In 2014, administrators of the City of Flint, Michigan, municipal water system switched the source of drinking water from Lake Huron to the Flint River without appropriate corrosion control measures.¹ Following the switch, community members experienced discoloration of their water and widespread exposure to elevated lead levels. Delayed acknowledgment and response by public administrators to this water system failure resulted in the loss of public trust.^2,3 Subsequent investigations into increases in Legionnaire’s disease, rash, and hair loss were affected by an underlying community concern that the water system might be the cause of additional public health problems.^4,5 In this context, an outbreak of shigellosis was detected in Flint, surrounding Genesee County, and neighboring Saginaw County beginning in March 2016.

Shigella is responsible for an estimated 500 000 cases of gastroenteritis annually in the United States⁶ and is a leading cause of outbreaks of diarrheal disease spread by person-to-person contact, especially among children.⁷ Shigella bacteria are spread through the fecal-oral route and are highly infectious, capable of causing prolonged communitywide outbreaks where a single source might be difficult to identify and transmission from person to person can play a major role.⁸ Shigellosis outbreaks in the United States can also be caused by contaminated food⁹ or, rarely, by contaminated recreational or drinking water because of inadequate water treatment.^10–13 Prevention of shigellosis depends on good hygiene practices to prevent person-to-person transmission and on appropriate water disinfection, which is effective at inactivating Shigella in water systems.¹⁴ Shigellosis is a national notifiable disease. In Michigan, all health care providers, laboratories, and local health departments are required to report clinical, laboratory, and epidemiological data for shigellosis cases electronically to the Michigan Disease Surveillance System (MDSS) at the Michigan Department of Health & Human Services (MDHHS). Laboratory diagnostic specimens must also be submitted to the MDHHS when available.

Drinking water is provided to Genesee County residents by multiple public and private water systems, including the Flint municipal water system, which serves residents in an area that approximates City of Flint boundaries. Drinking water in Saginaw County is also provided by many separate water systems. No Saginaw County households are served by the Flint municipal water system.

Initial investigation of the outbreak by the Genesee County Health Department did not identify a source. Preliminary data suggested person-to-person transmission, with many cases among people younger than 18 years (52%) and with 33% of patients reporting another ill person in their household. Despite increased prevention messaging in the community regarding hygiene,^15–17 new cases continued to occur. Community concern emerged that avoidance of tap water in Flint might have compromised hygiene and fueled the outbreak; some expressed concern that shigellosis might have spread through the Flint municipal water system. As these concerns grew, the MDHHS requested Epi-Aid assistance from the Centers for Disease Control and Prevention (CDC) and a field team deployed to assist with the investigation.

METHODS

The investigation team met with state and local public health officials and community members to review data from the outbreak and better understand community concerns. Team members participated in meetings with Genesee and Saginaw County Health Department staff, the Flint Water Recovery Group, the Genesee County Medical Society, and other key community members. Community input informed the design of a multipronged investigation, including (1) a case-household survey to characterize households affected by the outbreak and identify common exposures, (2) a genomic investigation to understand the relatedness of Shigella isolates in the outbreak, and (3) a geospatial investigation to explore shigellosis incidence and examine available water quality data. We provided regular community updates and sought additional input through follow-up meetings during and after the field investigation.

Outbreak Case Definition

We defined an outbreak case as a shigellosis infection diagnosed by a culture-based or culture-independent diagnostic test or a clinically compatible case with epidemiological linkage to a laboratory-confirmed case in a resident of Genesee or Saginaw Counties with illness onset during March to October 2016 reported to the MDSS or identified through the case-household survey.

Case-Household Survey

We identified all Genesee and Saginaw County households containing 1 or more cases of shigellosis reported to the MDSS with illness onset during March to October 2016. We geocoded household addresses and identified households receiving Flint city water using existing water system maps. With input from community stakeholders, we developed a questionnaire to characterize households in the outbreak and identify common exposures among cases. We collected demographic and illness information for all household members and asked about the sources of water used for consumption, hygiene, and household tasks. During administration of the questionnaire, we identified the first person to become ill in each household (referred to as the index case-patient) and asked them (or a parent or guardian when appropriate) about their activities and exposures in the week before becoming ill. Because secondary spread is common within a household during shigellosis outbreaks,¹⁸ we assessed index case-patient exposures to understand how shigellosis might have first entered the home. We did not assess these exposures for secondary cases or other household members. We generated descriptive statistics for household demographics, illness information, and index case-patient exposures. We compared results from households receiving water from the Flint municipal water system with results from households receiving water from all other public or private water systems in Greater Genesee County (which includes all non-Flint households in Genesee County) and Saginaw County using χ², Fisher exact, Wilcoxon–Mann–Whitney, and t tests. We compared individuals and households with the general population using data from intercensal estimates and the Census American Community Survey.

Genomic Investigation

We designed a genomic investigation to assess the relatedness of Shigella sonnei isolates from patients in the outbreak and to consider how these isolates were related to other Michigan isolates. From isolates previously submitted to the Michigan Bureau of Laboratories, we selected a random sample of isolates from patients in the outbreak and a random sample of other Michigan isolates from 2015 to 2017 from patients with no known epidemiological linkage to the outbreak. The Michigan Bureau of Laboratories conducted pulsed-field gel electrophoresis using the CDC PulseNet standard protocol¹⁹ and uploaded the results to the PulseNet USA national Shigella database for comparison with other US isolates from 2015 to 2016. The CDC performed whole-genome sequencing on the Illumina MiSeq using NexteraXT (Illumina Inc, San Diego, CA) DNA library preparations and 2 × 250 base-pair sequencing chemistry and conducted high quality single nucleotide polymorphism (SNP) analysis with Lyve-Set 1.1.4f (https://github.com/lskatz/lyve-SET) using a closely related PacBio sequence as a reference (GenBank: CP022455.1) with phage regions masked. We used the following parameters for SNP calling: 20× coverage and 95% read support with SNPs clustered closer than 5 base pairs filtered out. Phylogenetic trees were prepared by the Enteric Diseases Bioinformatics Team at the CDC, annotated by the PulseNet USA Database Team, and interpreted using epidemiological data reported to the CDC by state and local health departments. All sequences are available at the National Center for Biotechnology Information under BioProject PRJNA218110. Antimicrobial susceptibility testing using broth microdilution was conducted by the National Antimicrobial Resistance Monitoring System at the CDC for outbreak isolates submitted for routine national surveillance of antimicrobial resistance. The Michigan Bureau of Laboratories tested additional randomly selected Shigella isolates via disk diffusion. Resistance was defined by Clinical and Laboratory Standards Institute interpretive standards, when available.

Geospatial Investigation

We developed a geospatial investigation to assess the geographic distribution of cases; detect spatial, temporal, or space-time clustering; and examine water quality data. We calculated shigellosis incidence by census tract using geocoded case-household locations and US Census estimates. For case-households receiving Flint municipal water, we conducted spatial, temporal (by week), and space-time Kulldorf Scan statistics²⁰ to look for clustering that might be associated with a subdivision of the City of Flint municipal water system. We analyzed weekly free chlorine residual testing results from 10 City of Flint monitoring sites and 24 Environmental Protection Agency (EPA) monitoring sites (https://epa.maps.arcgis.com/apps/webappviewer/index.html?id=b0dd83f57f834625b47ccdb7da81db02) and assessed mean values alongside weekly shigellosis case counts. We excluded chlorine measurements of 0.0 mg/ml with no accompanying date (n = 3) from the analysis, as well as weeks (n = 11) in which no data were collected at the EPA monitoring sites.

RESULTS

Between March and October 2016, 177 cases of shigellosis were reported to the MDSS from 126 households, including 127 cases from 83 Genesee County households and 50 cases from 43 Saginaw County Households (Figure 1). We successfully interviewed 83 of 126 total households (66%). Within these 83 households, we identified 353 total household members, including 158 people meeting the outbreak case definition; 115 cases (73%) had been reported to the MDSS and 43 had not previously been reported. Of the 83 households interviewed, 24 (29%) received Flint municipal water and 59 (71%) received water from other drinking water systems. Median household size was similar for households in Saginaw County, Flint, and Greater Genesee County (Table 1). Households in the outbreak were significantly larger on average than households in the general population in these counties (4.2 people vs 2.5 people, P < .001; data from US Census Bureau, American Community Survey). Secondary attack rates were similar across households in Genesee and Saginaw Counties.

FIGURE 1—

Shigellosis Cases Reported to the Michigan Disease Surveillance System (MDSS) by Week: Genesee and Saginaw Counties, Michigan, March 1–October 30, 2016

Note. n = 177.

TABLE 1—

Characteristics of Households and Index Case-Patients From a Survey of Households Affected by an Outbreak of Shigellosis: Genesee and Saginaw Counties, Michigan, 2016

			Genesee County
	All Households	Saginaw County	Greater Genesee County	Flint	P (Flint vs All Non-Flint)
Household characteristics
Total households	83	29	30	24a	. . .
Total household members	353	119	126	108	. . .
Total cases	158	59	49	50	. . .
Median household size (IQR)	4 (3–5)	4 (3–6)	4 (3–5)	4 (3–5)	.65b
Median cases per household (IQR)	1 (1–2)	1 (1–3)	1 (1–2)	2 (1–2)	.32b
Secondary attack rate, %	28	33	20	31	.43c
Patient demographics
Median age, y (IQR)	10 (28)	12 (30)	10 (26)	10 (28)	.89b
Age range, y	< 1–86	< 1–86	1–70	1–73	. . .
Black race, self-report, no./total no. (%)	94/158 (59)	45/59 (76)	19/49 (39)	30/50 (60)	.93c
Female gender, self-report, no./total no. (%)	91/158 (58)	29/59 (49)	34/49 (69)	28/50 (56)	.78c
Index case-patient exposures, no./total no. (%)
Contact with a person wearing diapers who did not live with them	50/83 (60)	21/29 (72)	14/30 (47)	15/24 (63)	.79c
Contact with a person with diarrhea who did not live with them	15/83 (18)	7/29 (24)	2/30 (7)	6/24 (25)	.35d
Contact with a person wearing diapers or having diarrhea who did not live with them	55/83 (66)	23/29 (79)	15/30 (50)	17/24 (71)	.57c
Attended day care	10/83 (12)	5/29 (17)	1/30 (3)	4/24 (17)	.46d
Attended child care in another home	7/83 (8)	4/29 (14)	1/30 (3)	2/24 (8)	.99d
Household water uses, no./total no. (%)
Tap water used for drinking	51/83 (61)	29/29 (100)	18/30 (60)	4/24 (17)	< .001c
Bottled water used for drinking	54/83 (65)	13/29 (45)	19/30 (63)	22/24 (92)	.001c
Tap water used for handwashing	67/83 (81)	28/29 (97)	19/30 (63)	20/24 (83)	.99d
Hand sanitizer used for handwashing	21/83 (25)	9/29 (31)	7/30 (23)	5/24 (21)	.55c
Tap water used for bathing/showering	72/83 (87)	28/29 (97)	23/30 (77)	21/24 (88)	.99d
Changed handwashing habits since Flint water crisis	17/77 (22)	2/29 (7)	7/27 (26)	8/21 (38)	.06d
Changed bathing/showering habits since Flint water crisis	15/78 (19)	0/29 (0)	4/28 (14)	11/21 (52)	< .001d

Note. IQR = interquartile range. Greater Genesee County includes all non-Flint households in Genesee County.

^aAll Flint households (n = 24) are supplied by the City of Flint municipal water system; Greater Genesee County (n = 30) and Saginaw County (n = 29) households are supplied by other water systems.

^bWilcoxon rank-sum test.

^cχ² square test.

^dFisher exact test, 2-sided.

Median age of patients was similar between counties and younger than the median age of the general population (Saginaw: 12 years vs 40 years; Genesee: 10 years vs 39 years; P < .001 for both counties; data from US Census Bureau, American Community Survey). Black race was more frequent among patients in the outbreak than in the general population (Saginaw: 76% vs 19%, P < .001; Genesee: 49% vs 20%, P < .001). Female gender was more frequent among Genesee County patients than Saginaw County patients, and Genesee County patients were more likely to be female than the general population of Genesee County (63% vs 52%). Twenty-eight patients (18%) reported hospitalization for their illness; no deaths were reported.

When asked about exposures in the week before becoming ill, 50 of 83 index case-patients (60%) reported contact with a person wearing diapers who did not live with them, 15 of 83 (18%) reported contact with a person with diarrhea who did not live with them, and 55 of 83 (66%) reported 1 or both of these exposures. Few index case-patients reported attending day care or attending child care in someone else’s home in the week before becoming ill. These exposures were similar among households in Saginaw County, Flint, and Greater Genesee County. No common event, pattern of travel, restaurant, or recreational water venue was identified.

When asked about all household sources of water for drinking, making drinks, or making ice, few households supplied by the Flint municipal water system (4 of 24, 17%) reported using tap water compared with households receiving water from other drinking water systems (47 of 59, 80%; P < .001). Use of bottled water for drinking, making drinks, or making ice was more frequent among households supplied by the Flint municipal water system (22 of 24, 92%) than among other households in the outbreak (32 of 59, 54%; P < .001). Flint households did not differ from other households in the use of tap water or hand sanitizer for handwashing. In both counties, households frequently reported using tap water for bathing or showering; 1 Flint household reported using bottled water. No Flint households reported using hand sanitizer or cleaning wipes for bathing or showering. More households in Flint reported changing bathing or showering habits and handwashing habits since the Flint water crisis compared with other households in the outbreak.

Genomic Investigation

We analyzed 27 isolates from patients in the outbreak and 57 isolates from Michigan patients without known epidemiological linkage to the outbreak (isolate date 2015–2017) using pulsed-field gel electrophoresis and whole-genome sequencing (84 isolates total). Pulsed-field gel electrophoresis analysis revealed 4 similar but distinct patterns among outbreak isolates and 19 distinct patterns among non-outbreak-associated isolates. We constructed a phylogenetic tree from high quality SNP analysis of all 84 Michigan isolates and 2 reference isolates from another state (Figure 2). Most isolates sorted into 2 clades: clade 1 (62 isolates, range = 0–58 SNPs) contained all 27 outbreak isolates; clade 2 contained 12 isolates (range = 1–42 SNPs). Twelve isolates did not fall into either clade, including the 2 reference isolates. Within clade 1, outbreak isolates from Genesee County (n = 15) differed by 0 to 25 SNPs, whereas isolates from Saginaw County (n = 12) differed by 1 to 9 SNPs. Genesee County outbreak isolates differed from Saginaw County outbreak isolates by 15 to 33 SNPs. Isolates from 9 patients whose household tap water was supplied by the Flint municipal water system differed by 0 to 25 SNPs. We detected no clinically significant antimicrobial resistance for 3 isolates tested by the National Antimicrobial Resistance Monitoring System or 14 isolates tested by the Michigan Bureau of Laboratories.

FIGURE 2—

Phylogenetic Tree From High Quality Single Nucleotide Polymorphism Analysis of Shigellosis Outbreak Isolates From 2016, Other Michigan Isolates From 2015 to 2017, and Reference Isolates

Note. n = 27 for shigellosis outbreak isolates from 2016, n = 57 for other Michigan isolates from 2015–2017, and n = 2 for reference isolates.

Geospatial Investigation

Shigellosis incidence during the outbreak varied by census tract in Genesee and Saginaw Counties; highest rates occurred in the cities of Flint and Saginaw (Figure A, available as a supplement to the online version of this article at http://www.ajph.org). We did not identify any statistically significant spatial or spatial-temporal clusters of cases among case-households receiving Flint municipal water during the outbreak period. Average weekly free chlorine residual values from City of Flint and EPA monitoring sites ranged from 0.63 to 1.25 milligrams per liter during the outbreak period. Among 8 City of Flint monitoring sites with additional historical data available, average free chlorine residual values varied in the years leading up to the shigellosis outbreak but remained above 0.2 milligrams per liter throughout the outbreak period (Figure 3).

FIGURE 3—

Average Free Chlorine Residual Results for City of Flint Municipal Water System Monitoring Sites and Cases of Shigellosis in Flint Reported to the Michigan Disease Surveillance System (MDSS) by Week: 2012–2016

Note. n = 8 for monitoring sites. Dashed line indicates an average free chlorine residual level of 0.2 mg/L. This is not a regulatory requirement, but rather a recommended minimum level for free chlorine residual testing in a water distribution system.²¹

DISCUSSION

Our investigation found that the outbreak was not caused by a common source, such as the City of Flint municipal water system, but was instead caused by person-to-person transmission across communities in Genesee and Saginaw Counties. As in other communitywide person-to-person shigellosis outbreaks, this outbreak spanned several months, disproportionately affected children, and frequently affected more than 1 household member. Index case-patients frequently reported contact with people wearing diapers or people with diarrhea in the week prior to becoming ill, suggesting increased risk for acquiring shigellosis, which passes easily from person to person. These findings were similar across households in Genesee and Saginaw Counties, suggesting a common mode of transmission. As is often the case in large communitywide outbreaks, no single source or setting was implicated; this is likely because of the ease with which shigellosis spreads through schools, workplaces, households, and other community settings. We identified an additional 43 cases in the case-household survey not previously reported to public health. This is not surprising, because only a subset of people with a diarrheal illness seeks medical care, undergoes laboratory testing, and is reported to public health.²² However, this indicates that public health surveillance data likely underestimated the population affected by this communitywide outbreak.

Compared with the other jurisdictions, households in Flint more frequently reported changing their bathing, showering, and handwashing habits since the Flint water crisis. Although we did not ask respondents for additional details of their behavior change, it is possible that these changes could have compromised hygiene and contributed to the outbreak in Flint. However, given similarities in household uses of water for hygiene, secondary attack rates, and other household characteristics, it is not likely that these changes in Flint caused the outbreak. An assessment of knowledge, attitudes, and practices related to water, sanitation, and hygiene might provide additional insight, but it was outside the scope of this household survey. Critically, these results underscore the importance of effective messaging about hand hygiene and surface cleaning to prevent shigellosis outbreaks caused by person-to-person contact.

The major community concerns identified by the investigation team involved the possible role of the Flint municipal water system in the outbreak. Our results indicate it is unlikely that contamination of the water system caused the outbreak. Shigellosis incidence gradually increased and decreased in this outbreak, unlike the rapid rise and fall that might be seen with point-source transmission via a water contamination event. Reported exposure to water from the Flint municipal water system was low among households in the outbreak: fewer than half of households affected by the outbreak received tap water from the Flint municipal water system, and no pattern of travel was identified for non-Flint residents to indicate more widespread exposure to Flint municipal water. Few households receiving water from the Flint municipal water system reported consuming tap water during the outbreak period.

Water quality data from the Flint municipal water system also suggest that the outbreak was not caused by a widespread water contamination event. Average free chlorine residual levels from monitoring sites in Flint remained above 0.5 milligrams per liter throughout the outbreak, indicating that water treatment was adequate to inactivate Shigella and similar waterborne pathogens.^14,23 Among other chlorine-sensitive enteric pathogens that could also spread via widespread water contamination (e.g., Escherichia coli, Salmonella spp., Campylobacter spp.), no similar increase in case counts occurred during the outbreak period. No spatial or space-time clusters were identified among households served by the Flint municipal water system, suggesting that cases did not cluster near a subsection of the water system. These data do not, however, exclude the possibility that exposure to water contaminated at a point between the distribution system and the household tap contributed to the outbreak, as in the case of local sewage exposure or premise plumbing deficiencies. To further explore this, we considered conducting prospective household water sampling for case-households affected by the outbreak, but we did not pursue this because of the low number of new cases reported during the investigation, input from Flint community members, and findings from other investigation activities. Shigellosis outbreaks are rarely associated with drinking water systems in the United States; only 1 has been reported to the CDC’s National Outbreak Reporting System²⁴ since 2009. Standard disinfection practices like chlorination are effective for preventing shigellosis outbreaks in drinking water systems and play a vital role in maintaining a safe drinking water supply.²³

Community engagement played a critical role in understanding community concerns and preferences, designing the investigation, and communicating progress and results. We met regularly with established stakeholder groups and sought recommendations for other key stakeholders to engage. Our participation in Flint Water Recovery Group meetings, which brought together diverse stakeholders and were open to the general public, began before the field investigation and continued afterward for several weeks; these meetings provided valuable opportunities for community input and for sharing progress updates every 1 to 2 weeks during the investigation. Overall, our community engagement efforts resulted in at least 15 stakeholder meetings involving a total of more than 200 individuals from Genesee and Saginaw Counties, including public health officials, health care providers, researchers, political leaders, state and federal agency representatives, activists, nonprofit organizations, neighborhood leaders, and other members of the public. Community input directly informed our decision to conduct a case-household survey instead of a communitywide case–control study. This allowed us to assess a wider range of hypotheses about the outbreak and honor community preference not to conduct a larger and potentially more disruptive study unless a specific hypothesis suggested it was necessary. We deployed a health communication specialist to assist with gathering community feedback and to pilot the household survey with community members. In all, we developed 12 new health communication products, including letters to community members and health care providers, update presentations and reports, and prevention materials valuable for this and future response efforts.

Whole genome sequencing data provided valuable insight into this outbreak. All outbreak isolates sorted into a single large clade, which also contained isolates from other counties and from months before and after the recognized outbreak in Genesee and Saginaw Counties. This clade is notable for its overall diversity (0–58 SNPs) and for many small groups of closely related isolates (0–2 SNPs). This combination of diversity and interrelatedness is not consistent with a point source (such as a water system contamination event) but instead suggests spread across time and space through many smaller clusters of infections, as seen in other communitywide outbreaks of shigellosis.^25,26 Our analysis found genetic links between this outbreak and other shigellosis cases in the state not previously known to be connected; this provides additional insight into the extent to which shigellosis can spread between communities. These data cannot pinpoint the origin of the outbreak, since only a subset of cases have specimens available for analysis and SNP differences alone do not reveal direction of spread between patients. Future advances in Shigella genetic analysis are needed to further characterize thresholds for relatedness and better understand spread through a population.

Limitations

Our investigation was subject to at least 2 limitations. First, our investigation took place near the end of the long outbreak period, which likely affected recall of illness and exposure information, especially among patients with illness earlier in the outbreak. Second, it was challenging to gather, synthesize, and incorporate feedback from across a community where complex dynamics related to the Flint water crisis had developed prior to our involvement. Although we were not able to address every proposed hypothesis or implement every suggestion, we believe our approach to community engagement was essential for partnering with community members who had previously lost trust in public officials.

Public Health Implications

This investigation of a communitywide outbreak of shigellosis used a multimethod approach to identify transmission dynamics while addressing a wide variety of community concerns, including the possibility that the outbreak spread through a drinking water system. This required consistent community engagement to design an appropriate investigation, develop prevention messaging, and communicate progress and results. Although the unique circumstances of this outbreak prompted the development of this multimethod approach, a similar approach should be considered for future outbreak investigations involving pathogens with challenging transmission dynamics, compromised infrastructure, or loss of public trust. This combination of community engagement and optimal scientific methodology is critical for building lasting partnerships and protecting the health of the public.