You are on page 1of 26

CHAPTER

4
Foodborne Disease Surveillance
and Outbreak Detection

T
he term foodborne disease surveillance is often used to

describe routine monitoring in a population for any enteric

disease. The actual vehicle is usually not known during the

surveillance and early stages of the investigation processes, and

transmission ultimately could be caused by food, water, person-to-

person spread, animal contact, or other exposures.

A primary function of foodborne disease surveillance is detection of

problems in food and water production and delivery systems that might

otherwise have gone unnoticed. Rapid detection and investigation of

outbreaks is a critical first step to abating these active hazards and

preventing their further recurrence (discussed further in Chapter 5).

Broader goals of surveillance include defining the magnitude and

burden of disease in the community, monitoring trends, measuring the

effectiveness of control programs, attributing disease to specific food

vehicles, providing a platform for applied research, and facilitating

understanding of the epidemiology of foodborne diseases. This

chapter focuses on outbreak detection aspects of surveillance.


112 CIFOR | Council to Improve Foodborne Outbreak Response

4.0. Introduction

Unlike food-monitoring programs, which seek recall of hundreds of millions of pounds


to identify problems in food production and of contaminated products and prompting
correct them before illnesses occur, foodborne numerous large and small changes in food-
disease surveillance cannot prevent initial production and food-delivery systems. Many
cases of disease. Nevertheless, surveillance improvements in food safety during the past
is a sensitive tool available for identifying 100 years directly or indirectly resulted from
failures anywhere in food-supply systems. Food outbreak investigations. However, current
monitoring must concentrate on monitoring surveillance practices vary widely, are unevenly
the effectiveness of risk-reduction procedures at resourced, and generally exploit only a fraction
critical control points during the production of of the systems potential.
certain foods. However, the range of possible
food vehicles detectable through foodborne When a possible foodborne disease outbreak
4

disease surveillance includes all food or other is first detected or reported, investigators will
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

substances contaminated at any link in the not know whether the disease is foodborne,
chain from production to ingestion. Foodborne waterborne, or attributable to other causes.
disease surveillance complements regulatory Investigators must keep an open mind in the
and commercial monitoring programs by early stages of the investigation to ensure that
providing primary feedback on the effectiveness potential causes are not prematurely ruled
of prevention programs. out. Although the focus of these Guidelines is
foodborne disease, many of the surveillance and
Over the years, foodborne disease surveillance, detection methods described in this chapter and
coupled with outbreak investigation, has the investigation methods described in Chapter
remained among the most productive 5 apply to a variety of enteric and other
public health activities, resulting in the illnesses, regardless of source of contamination.

4.1. Overview
Disease surveillance is used to identify clusters Pathogen-specific surveillance:
of possible foodborne illness. Investigation Health-care providers and laboratorians
methods (Chapter 5) then are used to identify report individual cases of disease when
common exposures of ill persons in the cluster selected pathogens, such as Salmonella
that distinguish them from healthy persons. enterica or Escherichia coli O157:H7, are
Although, in practice, detecting individual identified in specimens from patients.
foodborne disease outbreaks involves multiple This surveillance method also includes
approaches, two general methods are used specific clinical syndromes with or without
in outbreak detection: pathogen-specific laboratory confirmation, such as hemolytic
surveillance and complaint systems (Table uremic syndrome and botulism, which
4.1). A third method, syndromic surveillance, usually indicate a particular pathogen.
is used in some jurisdictions, but its role Exposure information is gathered by
in detecting foodborne disease outbreaks interviews with cases. Data and pathogens
is limited. Although these methods are collected as part of food, animal, or
presented separately for descriptive purposes, environmental monitoring programs
they are most effective when used together enhance this surveillance method. The
and integrated with food, veterinary, and national notifiable disease reporting system
environmental monitoring programs, as will be and molecular subtyping available through
described later in Chapters 4 and 5. the National Molecular Subtyping Network
Table 4.1. Comparison of foodborne disease surveillance systems
FUNCTIONAL SURVEILLANCE METHOD
CHARACTERISTIC COMPLAINT
OF METHOD PATHOGEN-SPECIFIC SYNDROMIC
GROUP NOTIFICATION INDIVIDUAL COMPLAINT
Inherent speed of outbreak
Relatively slow Fast Fast Potentially fast*
detection
4.1. Overview

Sensitivity to widespread,
low-level contamination High Intermediate Intermediate Low
events (best practices used)
Types of outbreaks Limited to clinically suspected or Any, although effectiveness limited Limited to syndromes
(etiology) that method can laboratory-confirmed diseases under Any to agents with short incubation (or indicators) under
potentially detect surveillance periods surveillance

Multiple independent reports Trend in health


Report of group illnesses
with common exposures in indicator different
Initial outbreak signal (at Cluster of cases in space or time with recognized by health-care
space or time or unique clinical from expected,
public health level) common agent provider, laboratory, or the
presentation recognized by the space/time clusters of
public
agency receiving the reports diagnosed cases

No. cases needed to create


Low to moderate Low Low to moderate High
initial signal
High**
High**
(after interview of cases and collection Low to moderate (after interview
(after interview of cases and
Signal-to-noise ratio of appropriate food history) of cases and collection of Low
collection of appropriate
Even higher when combined with appropriate food history)
food history)
subtyping

* A  n advantage in speed is limited mainly to nonspecific health indicators (preclinical and Exposure histories are not typically obtained.
clinical prediagnostic data). Data must be analyzed, and a follow-up investigation is ** A high signal-to-noise ratio means that even a small number of cases stand out against
required, including comparison with standard surveillance, before public health action a quiet background. A low ratio means a cluster of cases or events is difficult to perceive
can be taken. because it is lost in the many other similar cases or events happening simultaneously

Sensitivity is higher for rare, specific syndromes, such as botulism-like syndrome. similar to a weak radio signal lost in static noise. The signal-to-noise ratio for syndromic
surveillance is lowest for nonspecific health indicators, such as loperamide use or visits
Although outbreaks can be detected without an identified etiology, linking multiple
outbreaks to a common source may require agent information. to the emergency department with diarrheal disease complaints. The ratio increases
with increasing specificity of agent or syndrome information. For highly specific, rare
T he number of cases needed to create a meaningful signal is related to the specificity of
syndromes, such as botulism-like syndrome, the signal-to-noise ratio would approach
the indicator. Indicators that offer an advantage in speed also tend to have low specificity.
2014 | Guidelines for Foodborne Disease Outbreak Response

that of pathogen-specific surveillance.

A ND O U TBR E A K D ETE C TI O N
113

FO O D BO RNE D I S EA S E S U R VEI L L A NC E
4
114 CIFOR | Council to Improve Foodborne Outbreak Response

4.1. Overview

for Foodborne Disease Surveillance System data on nonspecific health indicators that
(PulseNet) are examples of pathogen-specific might reflect increased disease occurrence,
surveillance. such as purchase of loperamide (an
Complaint systems antidiarrheal agent), visits to emergency
Health-care providers or the public identify departments for diarrheal complaints, or
and report suspected disease clusters (group calls to poison control hotlines. Exposure
notifications) or individual complaints. information is not routinely collected.
Exposure information is acquired by This chapter reviews major features, strengths,
interviews with cases. and limitations of each surveillance method
Syndromic surveillance and provides recommendations for increasing
This surveillance method generally involves the effectiveness of each.
4

systematic (usually automated) gathering of


FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

4.2. Pathogen-Specific Surveillance

4.2.1. Purpose Box 4.1. Selected nationally notifiable


diseases that can be foodborne
To systematically collect, analyze, and
disseminate information about laboratory- Anthrax (gastrointestinal)
confirmed illnesses or well-defined syndromes Botulism (foodborne)
as part of prevention and control activities. Cholera
4.2.2. Background Cryptosporidiosis
Cyclosporiasis
Surveillance for typhoid fever began in 1912 Giardiasis
and was extended to all Salmonella spp. in Hemolytic uremic syndrome, postdiarrheal
1942. National serotype-based surveillance of Hepatitis A virus infection, acute
Salmonella began in 1963, making it one of the Listeriosis
oldest pathogen-specific surveillance programs Salmonellosis
and the oldest public health laboratory Shiga toxinproducing Escherichia coli
subtype-based surveillance system. The (STEC) infection
usefulness of pathogen-specific surveillance Shigellosis
is related to the specificity with which agents Trichinellosis (Trichinosis)
are classified (i.e., use of subtyping and Typhoid fever
method), permitting individual cases of Vibrio infection
disease to be grouped with other cases most
likely to share a common food source or other In addition, the following are nationally notifiable:
exposure (Box 4.1). The utility of bacterial Foodborne disease outbreaks
surveillance increased during the 1990s with Waterborne disease outbreaks
the development of PulseNet and molecular
subtyping of selected foodborne pathogens, From CDC. Nationally Notifiable Infectious
including Salmonella, Shiga toxinproducing Diseases. United States 2008. Revised.
Escherichia coli (STEC) O157:H7, Shigella, Available at www.cdc.gov/nndss/
Listeria, and Campylobacter. document/2012_Case%20Definitions.pdf
2014 | Guidelines for Foodborne Disease Outbreak Response 115

4.2. Pathogen-Specific Surveillance

4.2.3. Case Reporting and Laboratory 4.2.4. Epidemiology Process


Submission Process
Information received by the public health
Most diseases included under pathogen- agency through multiple avenues, including
specific surveillance are reportable (i.e., basic clinical and demographic data from
notifiable) diseases. State or local health individual cases of specific laboratory-
agencies establish criteria for voluntary or confirmed illness or well-defined syndromes,
mandatory reporting of infectious diseases, is reconciled and linked with case isolates or
including those that might be foodborne (Table other clinical materials received in the public
4.2). These criteria describe the diseases to health laboratory. Reconciled case reports are
report, to whom, how, and in what time frame. forwarded to higher jurisdictional levels (local
For this type of surveillance, diseases are health agency to state agency, state agency to

4
defined by specific laboratory findings, such federal agency) by a variety of mechanisms.

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
as isolation of Salmonella enterica, or by well- In general, records are redacted (stripped
defined syndromes, such as hemolytic uremic of individual identifiers) when they are sent
syndrome. Diseases are reported primarily outside the reporting states.
by laboratories, medical staff (e.g., physicians,
infection-control practitioners, medical Cases are usually interviewed one or more
records clerks), or both. Disease reports can times about potential exposures and additional
be automatically generated from an electronic clinical and demographic information.
medical record or laboratory information The scope of these interviews varies by
system or reported through a secure website. jurisdiction. Interviews typically cover basic
Legacy systems, such as telephone, mail, or descriptive information and exposures of local
fax reporting, also are used but are slower and importance, such as attendance at a child-
more labor intensive and error prone. Isolates care facility, occupation as a food worker,
or other clinical materials are forwarded and medical follow-up information. Whereas
from laboratories serving primary health- many local agencies collect information
care facilities to public health laboratories for about a limited set of high-risk exposures,
confirmation and further characterization, more detailed exposure interviews might be
as required by state laws or regulations or as collected only when clusters are investigated
requested by the local jurisdiction. or outbreaks are recognized (Chapter 5).
However, routine collection of detailed
States and territories (or sometimes local public exposure information as soon as possible after
health agencies) voluntarily share pathogen- reporting maximizes exposure recall, provides
specific disease surveillance information a basis for rapid cluster investigation, and is
with the Centers for Disease Control and strongly recommended for high-consequence
Prevention (CDC). No personal identifiers enteric pathogens, such as STEC O157:H7
are forwarded, and only minimal information and Listeria monocytogenes. (See Chapter 5 for
is available about cases (e.g., date of onset, further discussion.)
age, sex, race/ethnicity, county of residence).
CDC works with states to compile national Initial cluster identification and cluster
surveillance data. assessment might occur as two processes
conducted, respectively, by the laboratory and
State-specific reporting requirements epidemiology departments or might occur as
can be viewed at www.cste.org/group/ a single process within epidemiology. Agent,
SRCAQueryRes. time, and place are examined individually and
in combination to identify possibly significant
116 CIFOR | Council to Improve Foodborne Outbreak Response

4.2. Pathogen-Specific Surveillance

clusters or trends. This is the critical first step information includes routine food-monitoring
in hypothesis generation. Clusters of unusual test results (see section 4.2.5.2) or concurrent
exposures, abnormal exposure frequencies, group or individual complaints (see section
unusual demographic distributions (e.g., 4.3). The most successful investigators consider
predominance of cases in a particular age information from as wide a variety of sources
group), or connection to food, animal, or as possible.
environmental monitoring studies might be
identified. Clusters of cases are examined as Finally, pathogen-specific data are ideally
a group and, if a common exposure seems compared routinely with complaint data,
likely, investigated further (Chapter 5). In some which offer significant advantages in sensitivity
jurisdictions, cluster detection and triage is a and specificity over either system alone
laboratory function (see section 4.2.5 below). (see section 4.3.6).
4

4.2.5. Laboratory Process


FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

Hypotheses to explain the cluster can be


developed in several ways. If trawling
Clinical diagnostic laboratories forward case
questionnaires (i.e., hypothesis-generating
isolates, specimens that were positive for a
or shotgun questionnaires, or extensive
reportable enteric pathogen by a culture-
interviews of possibly exposed persons,
independent test, or other clinical materials
including food histories) are routinely
to public health laboratories as part of
administered after a case is reported,
mandated or voluntary reporting rules. Such
hypotheses can be generated through
problems as mislabeling, broken-in-transit, or
examination of previously obtained exposure
quantity-not-sufficient are resolved. Receipt of
data based on common exposures above what
samples is recorded, and sample information
would be expected. This approach can be
is entered into the laboratory database. Patient
followed by an iterative follow-up interview
information submitted with the sample may
(see below). In jurisdictions where trawling
be provided to the epidemiology department
questionnaires are not used routinely, such
for comparison with information from cases
interviews might be used only for cases
already reported and to enable reconciliation
suspected to be part of a common-source
of case reports and laboratory samples and
cluster. Unless these interviews identify an
identification of previously unreported cases.
obvious exposure leading to direct public health
intervention, hypotheses are tested during the The agent identification is confirmed, and
ensuing investigation (see Chapter 5). tests used for subtyping (such as serotyping,
virulence assays, molecular subtyping, or
Questionnaire data are not the sole source of
antimicrobial susceptibility tests) are conducted
information available to investigators. The
to further characterize the agent. Reports
basic demographic profile of cases (age, sex,
are issued either singly or in groups to the
occasionally racial or ethnic composition)
epidemiology department. Reports also may
often provides important clues to the identity
be issued to submitters as permitted by local
of commercial food sources. The geographic
policies. Pulsed-field gel electrophoresis (PFGE)
and temporal distribution of cases likewise
or other subtype patterns and accompanying
can suggest (or rule out) certain kinds of
metadata are uploaded to local and national
exposures. Investigators should take advantage
databases. Consolidated daily reports, such
of product distribution data obtained from the
as subtype frequency reports, are often
food distributors or noteworthy outliers (i.e.,
used to facilitate cluster recognition. These
the cases that do not fit an otherwise well-
reports may be automatically generated by
established pattern). Other potentially useful
2014 | Guidelines for Foodborne Disease Outbreak Response 117

4.2. Pathogen-Specific Surveillance

laboratory or epidemiology information severity of disease, matches between human


systems; extracted from the PulseNet cases and food or animal monitoring samples,
database; or facilitated by software, such as the and competing demands for investigators
CIFOR laboratory/epidemiology reporting time. The time window used to delimit clusters
program (http://www.cifor.us/projelr.cfm). varies by agent. For example, a wider window
Case cluster data are enhanced by inclusion is used to evaluate clustering of listeriosis cases
of information about matching isolates or than to evaluate salmonellosis cases because
outbreaks through PulseNet from other of differences in the natural history of the
jurisdictions and by matching isolates from diseases. Although cluster recognition software,
food, animal, or environmental monitoring such as SaTScanTM, cusum outbreak detection
tests that provide information for hypothesis algorithms, and query algorithms in the
generation. Specimen data (including detailed PulseNet Web Portal have been developed, none

4
subtyping results) are additionally uploaded have yet been validated for broad-based enteric

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
to national surveillance systems, such as the disease data. The decision to report or pursue
U.S. Laboratory-based Enteric-Diseases a cluster is an important part of the outbreak
Surveillance (LEDS [in the United States] or detection process but not one that is easily
TESSy [in Europe]). distilled into simple best practices. An increase
in frequency of a strain is only one indication
4.2.5.1 Cluster definition and triage of a potentially significant cluster. Furthermore,
Although, in practice, the term may be used absence of an increase in case numbers from
somewhat casually, a cluster can be defined expected values does not rule out significance.
as two or more cases of disease linked by place,
time, pathogen subtype, or other characteristic. The subject of cluster evaluation will be
Our interest in clusters stems from the fact covered in more detail in Chapter 5. As
that some clusters represent common-source whole-genome sequencing becomes part of
outbreaks. An ill-defined transition in use of routine public health surveillance activities,
the terms cluster to outbreak reflects the new approaches will need to be developed to
certainty that similar cases are in fact related. define and evaluate clusters (also see section
Sometimes transition is immediately and 4.2.9.2). At this writing, real-time whole-
trivially apparent; at other times, doubts linger genome sequencing for outbreak detection and
indefinitely. investigation has been initiated on a pilot basis.
Full transition to genome-based molecular
Clusters may be more or less recognizable surveillance is anticipated in the near future.
and more or less actionable. Although
this chapter focuses on case clusters and 4.2.5.2. Microbiological Screening
outbreaks, it should be clear that for some Microbiological screening of food or other
high-consequence agents or syndromes (e.g., environmental specimens can be useful for an
botulism or paralytic shellfish poisoning), even individual case of botulism and for certain
single cases may merit a prompt and aggressive high-risk exposures reported even by single
public health response. cases of other diseases (e.g., pet reptiles for
Salmonella or raw milk or ground beef for
Clusters are common, and pursuing them all STEC). Targeted screening also might be
with equal vigor is not practical or productive. warranted when specific foods are suspected
The cluster triage process is primarily manual. and reasonable samples are available.
Incoming surveillance data are evaluated Unfocused microbiological screening of
for unusual case counts based on historical multiple foods to investigate clusters is generally
frequencies (accounting for seasonality), the unproductive and always resource-intensive.
118 CIFOR | Council to Improve Foodborne Outbreak Response

4.2. Pathogen-Specific Surveillance

Routine food screening is conducted as 1. Incubation time:


part of larger food safety verification The time from ingestion of a contaminated
programs operated by the Food and Drug food to beginning of symptoms. For
Administration (FDA), U.S. Department of Salmonella, this typically is 13 days,
Agriculture (USDA), and state agriculture sometimes longer.
agencies. Screening information also might be 2. T
 ime to contact with health-care provider
available from the food industry. Incorporating or doctor:
this routine food or animal monitoring or The time from the first symptom to medical
regulatory surveillance test data into the disease care (when a stool sample is collected for
surveillance information stream enhances laboratory testing). This time may be an
hypothesis generation and improves the additional 13 days, sometimes longer.
sensitivity and timeliness of outbreak detection.
4

In the United States, data streams from human 3. T


 ime to diagnosis:
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

disease surveillance, food-testing programs, The time from provision of a sample to lab
and selected live-animal testing are co-mingled identification of the agent in the sample as
in the PulseNet database, although important Salmonella. This may be 13 days from the
product details might not be readily available. time the lab receives the sample.
4. S
 ample shipping time:
4.2.6. Timeline for Case Reporting and The time required to ship the Salmonella
Cluster Recognition isolate from the lab to the state public
Pathogen-specific surveillance requires a series health authorities who will perform
of events from the time a patient is infected serotyping and DNA fingerprinting. This
through the time public health officials usually takes 1-3 days or longer, depending
determine the patient is part of a disease on transportation arrangements within
cluster. This delay is one of the limiting factors a state and distance between the clinical
of this type of surveillance. Minimizing delays lab and the public health department.
by streamlining the individual processes Diagnostic labs are not required by law in
improves the likelihood of overall success. A many jurisdictions to forward Salmonella
sample timeline for Salmonella case reporting is isolates to public health labs, and not all
presented in Figure 4.1. diagnostic labs forward any isolates unless
specifically requested to do so.
Figure 4.1. Sample Salmonella case 5. T
 ime to serotyping and DNA fingerprinting:
reporting timeline The time required for the state public health
authorities to serotype and to perform
Person Eats
Contaminated
Incubation time = 1-3
DNA fingerprinting on the Salmonella isolate
Food and compare it with the outbreak pattern.
Time to contact with healthcare Patient
system = 1-3 days Becomes ill Serotyping typically takes 3 working days
Stool but can take longer. DNA fingerprinting
Time to diagnosis = 1-3 days
Sample can be accomplished in 2 working days (24
Collected
Shipping time = 1-3 days Salmonella hours). However, many public health labs
Identified have limited staff and space and experience
Isolates & Case multiple emergencies simultaneously. In
Reports Received
by Public Serotyping and *DNA practice, serotyping and PFGE subtyping
Health Agency fingerprinting* = 2-10 days Case Confirmed may take several days to several weeks;
as Part of
Cluster faster turnarounds are highly desirable. The
transition to whole genome sequencing for
2014 | Guidelines for Foodborne Disease Outbreak Response 119

4.2. Pathogen-Specific Surveillance

subtyping and serotyping will likely reduce true population of affected persons because
turnaround time for this process. most cases of foodborne disease are not
The total time from onset of illness to diagnosed and reported. The completeness
confirmation of the case as part of an of the reporting and isolate submission
outbreak is typically 23 weeks. processes affects the representativeness
of the reported cases and the potential
4.2.7. Strengths of Pathogen-Specific number and size of outbreaks detected. If
Surveillance for Outbreak Detection the percentage of cases reported or isolates
submitted is low (i.e., sensitivity is low), small
P
 ermits detection of widespread disease outbreaks or outbreaks spread over space and
clusters initially linked only by a common time are likely to be missed. Furthermore, if
agent. Most national and international sensitivity is low, reported cases might differ

4
foodborne disease outbreaks are detected in significantly from cases not reported. This

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
this manner. bias is more likely to influence descriptions of
W
 hen combined with case information clinical illness or the magnitude and severity
from clusters recognized though complaints of illness than associations with any particular
(section 4.3), and when specific exposure vehicle, but it is worth keeping in mind as one
information is obtained, is arguably the develops hypotheses about the source (see
most sensitive single method for detecting Chapter 5).
unforeseen problems in food and water
4.2.9.2. Prevalence of the agent and specificity of
supply systems caused by the agents under
agent classification
surveillance. The specificity of agent or
The more common the agent, the more
syndrome information combined with
difficult it is to identify outbreaks and the
specific exposure information obtained by
more likely sporadic (unrelated) cases
interviews enables the positive association of
will be misclassified as outbreak cases.
small numbers of cases with exposures.
Misclassification reduces the power of the
4.2.8. Limitations of Pathogen-Specific investigation, obscuring trends and diluting
Surveillance outbreak measures of association (type 2
probability error or the possibility of missing
W
 orks only for diseases detected by routine an exposuredisease association when one
testing and reported to a public health truly exists). Consequently, a larger number
agency. of outbreak cases are needed to significantly
I s relatively slow because of the many steps associate illness with exposure.
required, as described in figure 4.1.
Examination of subsets of cases using
4.2.9. Key Determinants of Successful case definitions based on specific agent
Pathogen-Specific Surveillance classifications (e.g., inclusion of subtyping
results) or restricting cases using certain
The following interrelated factors are critical to time, place, or person characteristics
understanding the use of surveillance data to can minimize this impact. For example,
identify potential outbreaks and form the basis Salmonella Typhimurium, a common serotype,
for best practices of cluster investigations (see provides the opportunity for misclassification
Chapter 5). (i.e., grouping together cases resulting from
different exposures). However, Salmonella
4.2.9.1. Sensitivity of case detection Typhimurium cases that are part of a
Surveillance represents a sampling of the common-source outbreak are more likely than
120 CIFOR | Council to Improve Foodborne Outbreak Response

4.2. Pathogen-Specific Surveillance

cases not associated with the outbreak to share attention is suspicion that he or she might have
a PFGE subtype. Therefore, using the PFGE been part of a foodborne disease outbreak.
subtype in the case definition will decrease Routine case interviews should always identify
misclassification (i.e., exclude cases not related group exposures, such as a banquet, after
to the outbreak) and increase the chance of which other persons might have been ill. For
finding a statistically significant association these persons, the event itself largely (but not
between illness and exposure. This is the basic entirely) defines the exposures of interest,
principle behind PulseNet. such as menu items. However, exposures
that need to be considered in pathogen-
Increasing the specificity of strain classification, specific surveillance usually are open-ended;
for example by using serotypes, PFGE results, they include all exposures in a time frame
or whole-genome sequencing, is useful but appropriate to the disease.
4

has drawbacks. Some outbreaks are caused


FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

by more than one pathogen or more than one As noted above, many local agencies collect
subtype of a pathogen. If the strain associated information about a limited set of high-risk
with an outbreak is defined too narrowly exposures when the case is initially reported,
by investigators, truly associated cases with and routine collection of detailed exposure
different subtypes (or no subtyping at all) will be information can provide a basis for real-
eliminated from the investigation. Elimination time evaluation of clusters that might be
of these cases may become problematic justified for enteric pathogens of sufficient
when the number of cases associated with an public health importance. Lack of a list of
outbreak is small. It can result in overlooking specific exposures, such as a menu, makes
an outbreak altogether, but it also can decrease prompting cases during the interview more
study power and the likelihood of implicating difficult. Furthermore, cases identified through
a specific food as the source of the outbreak. pathogen-specific surveillance usually are
In addition, genetic changes can occur as interviewed later after the exposure than are
pathogens multiply over time in food, the those reported as part of specific events. Thus,
human body, or the environment. Pathogens greater attention must be paid to interview
and strains differ in the rate of change. As a timing and content.
result, isolates deriving from the same source
(e.g., a contaminated food) can have slightly 4.2.9.3.1. Timing
different genome sequences. To decrease the time between exposure to
the disease-causing agent and interview of
For these reasons, use of several different the case, reporting of cases by health-care
levels of agent specificity during analysis of providers and laboratories should be as easy as
surveillance data and in the investigation possible. Case interviews should be conducted
of a cluster might be helpful. In addition, as soon as possible because recall will be better
epidemiologic evaluation of whole-genome closer to the time of the exposure and cases
sequences usually involves clustering of will be more motivated to share information
pathogens with closely related genome with investigators closer to the time of their
sequences into larger groupings. Initial illness. Acquiring timely interviews might
discussions are under way to develop entail working outside regular office hours.
international conventions for use of whole-
genome sequence data.1 4.2.9.3.2. Content
In pathogen-specific surveillance, the interview
4.2.9.3. Sensitivity and specificity of interviews of cases form itself must include a broader range of
One reason an ill person seeks medical possible exposures than interview forms for
2014 | Guidelines for Foodborne Disease Outbreak Response 121

4.2. Pathogen-Specific Surveillance

event-driven investigations. Interview forms certain exposures) and data elements (e.g.,
that use a combination of question types will ask about the same high-risk exposures, such
increase the likelihood of detecting the desired as sprouts, raw milk, ground beef, and leafy
exposure information and should be used, as green vegetables) will enhance data sharing
appropriate to the outbreak circumstances. and enable comparisons among jurisdictions
Interview forms can include questions that: in multijurisdictional outbreaksand possibly
speed the resolution of commercial product
C
 ollect information about specific exposures, outbreaks.
such as a broad range of specific food
items and nonfood exposures previously 4.2.9.4. Overall speed of the surveillance and
(or plausibly) associated with the pathogen investigation processes
through closed-ended questions; Delays are inherent in pathogen-specific

4
P
 rompt cases to further describe exposures, surveillance. The usefulness of pathogen-

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
such as brand information and place of specific surveillance in preventing ongoing
purchase or consumption; and transmission of disease from contaminated
food, especially perishable commodities, is
E
 nable cases to identify unanticipated directly related to the speed of the process.
exposures through open-ended questions
(e.g., At which restaurants did you eat?). Once an outbreak investigation is under
way, routine surveillance practices and work
Questionnaire design involves balancing a schedules must be changed to match the
number of competing demands; the end result urgency of the investigation (see Chapter 5).
is always a compromise. Questionnaires with
many open-ended questions require more highly 4.2.10. Routine Pathogen-Specific
trained and skilled personnel than do interviews SurveillanceModel Practices
using more predefined lists of exposures.
Longer questionnaires can cover more possible This section lists model practices for routine
exposures but can task the patience of both case surveillance programs. Practices used in
and interviewer; cases might quit the interview any particular situation depend on a host of
before it is completed. Open-ended questions factors, including circumstances specific to the
generally are more difficult and time-consuming outbreak (e.g., the pathogen and number and
to abstract and for data entry. distribution of cases), staff expertise, structure
of the investigating agency, and agency
No one questionnaire will work for all resources. For example, aggressive identification
investigations or surveillance systems. and investigation of STEC O157:H7 cases
Investigators should consider the specifics of can identify outbreaks and enable the
the outbreak and setting, the importance of implementation of control measures that might
collecting the information, and the likely trade- minimize serious illness and death, whereas
offs before deciding on the content of the investigation of more numerous Campylobacter
interview form. cases is not as likely to lead to public health
interventions. Although a systematic evaluation
Regardless of interview content, use of a of the following practices under different
standardized interview form with which the circumstances has not been performed,
interviewer is familiar will decrease time spent experiences from successful investigations
on staff training and decrease errors in data support their value. Investigators are
collection. In addition, use of standardized encouraged to use a combination of practices
core questions (i.e., questions that use the as appropriate to the specific outbreak.
same wording for collecting information about
122 CIFOR | Council to Improve Foodborne Outbreak Response

4.2. Pathogen-Specific Surveillance

4.2.10.1. Reporting and isolate submission pathogen under study. Undertake subtyping
Increasingly clinicians are diagnosing and as the isolates are submitteddo not
treating patients without collecting and testing wait for a specific number of specimens
clinical specimens. Ongoing communication to accumulate before testing them. Tests
between public health agencies and clinicians such as PFGE and serotyping ideally are
is critical to reinforce the value of collecting performed concurrently to reduce turnaround
and submitting specimens to public health time. Recommended turnaround times are
laboratories for tracking and responding to described in the Association of Public Health
diseases of public health interest. Laboratories/CIFOR yardstick project
(http://www.aphl.org/aphlprograms/food/
Encourage health-care providers to test initiatives/Documents/FS_2012_Yardstick-
patient specimens as part of the routine Self-Assessment-Tool-for-Public-Health-
4

diagnostic process for possible foodborne Food-Safety-Testing.pdf). Post results to


FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

diseases. Increase reporting and isolate national databases as quickly as possible.


submission by clinical laboratories and health- Tests conducted on an as-needed basis during
care providers through: a) education about the a cluster investigation, such as multilocus
value of testing and reporting mechanisms; b) variable number tandem repeat analysis or
regulatory action (such as modifying reporting whole-genome sequencing, should be initiated
rules to mandate isolate submission); c) as soon as the need is recognized.
laboratory audits; and d) provision of easier
methods for compliance, such as automated Use of culture-independent diagnostics
or Web-based reporting, isolate-transport in clinical laboratories is anticipated to be
systems, more consistent reporting across increasing in the coming years. Therefore:
reporting areas, and limitation of the amount J urisdictions should consider amending
of information initially requested. Educate reporting rules to expand the definition of
physicians, laboratorians, and medical required clinical materials for submission to
records clerks by workshops or conferences, include patient specimens (e.g., stool, urine,
newsletters, electronic health alerts, and blood) because isolates currently specified in
regular feedback from public health agencies. most reporting rules might not be available
The medical rationale and specific in the near future.
recommendations for testing can be found in P
 rotocols should be developed for rapidly
Practical Guidelines for the Management of Infectious isolating pathogens from patient specimens.
Diarrhea2 and Diagnosis and management of
foodborne illnesses: a primer for physicians 4.2.10.3. Case interviews
and other health-care professionals.3 The Quality exposure information usually is
latter document provides a series of tables that difficult to obtain and often is the major
give useful information about major foodborne limiting factor of pathogen-specific
pathogens, including signs and symptoms, surveillance. Interview all persons with
incubation periods, and appropriate laboratory laboratory-diagnosed cases of possible
tests, and describes sample patient scenarios to foodborne disease as soon as case reports
help with the diagnostic process. or laboratory isolates are received, when
patient recall and motivation to cooperate
4.2.10.2. Isolate/specimen submission and with investigators is the greatest.
characterization
Confer with the laboratory to determine Obtain an exposure history consistent with
subtyping methods available for the the incubation period of the pathogen
2014 | Guidelines for Foodborne Disease Outbreak Response 123

4.2. Pathogen-Specific Surveillance

identified (see http://www.cdc.gov/foodsafety/ variables for which expected frequency of


outbreaks/investigating-outbreaks/confirming_ exposure is low. For example, because less
diagnosis.html for a table of incubation for the than 20% of the population is expected to eat
most common foodborne agents). raw spinach, asking only whether a case ate
raw spinach should be sufficient to identify
As appropriate to circumstances, construct raw spinach as a possible vehicle. However,
the interview to include a mix of question because more than 75% of the population is
types that will collect the desired exposure expected to eat chicken, additional brand or
information, including: source information is needed. Thus, using a
hybrid approach for collecting basic exposure
S
 pecific closed-ended questions about
information about low-frequency exposures
exposures as a priori hypotheses to be tested
and more specific information about high-

4
(including specific food items that have been
frequency exposures may be the most effective
linked to previous outbreaks or that could

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
approach. The use of open-ended questions
plausibly be associated with the specific
complicates electronic data entry and analysis.
pathogen);
For jurisdictions that rely on electronic data
B
 road open-ended questions to capture entry at the local public health level for rapid
exposures that might not have been communication with the state, answers to
considered; and open-ended questions may need to be captured
Q
 uestions that elicit additional details, as text fields that can be reviewed as needed.
such as brand and place of purchase or
Routine collection of detailed exposure
consumption, for some of the highest
information enables evaluation of clusters
likelihood exposures.
in real time. However, most public health
When possible, use standardized core questions agencies do not have sufficient resources to
and data elements used by other investigators conduct such interviews of every case. Given
to enhance data sharing and comparisons the reality of these resource limitations,
across jurisdictions. Experience can make a two-step interviewing process might be
one a better and more efficient interviewer. the best alternative approach. When first
If investigations are infrequent, achieving reported, all cases should be interviewed
and maintaining proficiency can be difficult; with a standardized questionnaire to collect
centralizing the interview process reduces these exposure information about limited high-
problems and makes questionnaires easier to risk exposures specific to the pathogen.
modify on the fly. Interviewees should be informed that
investigations may require additional
Entering, tabulating, and analyzing information and that they might be
questionnaire data is an essential part of contacted again. When the novelty of the
effective interviewing. Questionnaires should subtype pattern, geographic distribution
be designed with rapid and accurate data of cases, or ongoing accumulation of
entry in mind. The CIFOR Clearinghouse new cases indicate the cluster represents
(www.cifor.us/clearinghouse/keywordsearch. an outbreak possibly associated with a
cfm) provides examples of questionnaires commercially distributed food product, all
used by various health departments to collect cases in the cluster should be interviewed
exposure information for different pathogens. using a detailed exposure questionnaire as
Questions with a yes/no check-box format part of a dynamic cluster investigation (see
are efficient for collecting information about Chapter 5).
124 CIFOR | Council to Improve Foodborne Outbreak Response

4.2. Pathogen-Specific Surveillance

4.2.10.4. Data analysis subtype, geographic or temporal clustering


Use daily laboratory reporting and analysis or lack thereof, or unexpected demographic
systems, where possible, to more easily distribution (also see Chapter 5).
recognize and evaluate clusters. Automated
reports can be developed for laboratory 4.2.10.5. Communication
information management systems or Establish and use routine procedures for
epidemiology systems or by using the CIFOR communicating among epidemiology,
Epi/Lab reporting software. laboratory, and environmental health branches
within an agency and between local and state
Analyses should be able to handle various agencies. Rapidly post subtyping results to
agents (e.g., species, serotype or other subtype, PulseNet, and note the detection of clusters to
more stringent subtype), enabling differing PulseNet and foodborne outbreak electronic
4

types of available information, and should mailing lists to improve communication and
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

include basic demographic information, cooperation within and among local, state,
such as location, sex, and age. Compare and federal public health agencies. Poor
possible clusters to historical frequencies and coordination within and among agencies
national trends. Clusters are triaged on the limits the effectiveness of pathogen-specific
basis of the novelty of a subtype pattern or surveillance.
increased occurrence of a relatively common

CIFOR Keys to Success:


Focus Area 5Pathogen-specific surveillance
Reporting/submission of isolates
S tate has mandatory reporting of diseases and submission of patient isolates that were likely to
have been foodborne.
Staff actively solicit case reports and submission of specimens/isolates to improve completeness
of reporting.
Agency/jurisdiction has system to rapidly transport specimens and isolates from clinical
laboratories to the public health laboratory.
Testing of specimens
P
 ublic health laboratory has the capacity to quickly process and test specimens submitted by
clinical laboratories, including pathogen confirmation and subtyping.
Collection of exposure information
S
 taff collect sufficient demographic and exposure information from patients to recognize possible
patterns and associations between cases in a timely fashion.
Detection of clusters/outbreaks
S
 taff analyze case information (e.g., demographics, exposure information, agent information
including species, serotype, subtype) on a frequent basis to rapidly identify possible clusters or
outbreaks.
Communication
P ublic health laboratory shares test results with epidemiology staff in a timely fashion.
Public health laboratory reports test results to national databases in a timely fashion.
Making changes
A
 gency/jurisdiction has performance indicators related to pathogen-specific surveillance and
routinely evaluates its performance in this Focus Area.
2014 | Guidelines for Foodborne Disease Outbreak Response 125

4.2. Pathogen-Specific Surveillance

4.2.11. Multijurisdictional Considerations 4.2.12. Indicators/Measures for Pathogen-


for Pathogen-Specific Surveillance Specific Surveillance

Because pathogen-specific surveillance does The success of pathogen-specific surveillance


not depend on geographic clustering, it is more at detecting and resolving common-source
sensitive to detection of widespread, low-level outbreaks depends on multiple interrelated
contamination events than surveillance through processes. Indicators for assessing and
complaint systems. Outbreaks detected by improving surveillance programs can be found
pathogen-specific surveillance are more likely in Chapter 8.
to span multiple jurisdictions. See Chapter 7
for Multijurisdictional Investigation Guidelines.

4
4.3. Complaint Systems

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
4.3.1. Purpose cases reported, isolates sent to public health
agencies, and subtyping or further laboratory
Notification or complaint systems are intended testing (see section 4.2.6). Although pathogen-
to receive, triage, and respond to reports from specific surveillance and complaint systems are
the community about possible foodborne treated separately in this chapter, these two
disease events to conduct prevention and systems are synergistic when used together.
control activities. Programs range from ad
hoc response to unsolicited phone reports to 4.3.3. Group Illness and Independent
systematic solicitation and interview of and Complaints
response to community reports.
Complaint reporting involves passive collection
4.3.2. Background of reports of possible foodborne illness from
individuals or groups. Reporting is of two
Receiving and responding to reports of basic types, each with its own dynamics and
disease in the community has been a basic requirements:
function of public health agencies since their
inception. Whereas reports of diseases caused R
 eports from any individual or group who
by specific pathogens generally follow specific observes a pattern of illness affecting a
disease reporting rules, complaints of illnesses group of people, usually after a common
by consumers associated with specific events exposure. Examples include reports of illness
or food establishments generally have been among multiple persons eating at the same
referred to the agency responsible for licensing restaurant or attending the same wedding
the establishment. These consumer complaints and reports from health-care providers of
lead to the identification of most localized unusual patterns of illness, such as multiple
foodborne disease outbreaks and are the only patients with bloody diarrhea in a short time
method for detecting outbreaks caused by span.
agents, such as norovirus, for which there is M
 ultiple independent complaints about
rarely pathogen-specific surveillance. Unlike illness in single persons or households.
pathogen-specific surveillance (described
above) notification and complaint systems Group illness and independent complaints may
do not depend on ill persons seeking medical be used together and linked with data obtained
attention. Therefore, it is not necessary for through pathogen-specific surveillance. In
laboratory tests to be ordered and performed, contrast to pathogen-specific surveillance,
126 CIFOR | Council to Improve Foodborne Outbreak Response

4.3. Complaint Systems

complaint reporting does not require information on each and every independent
identification of a specific agent or syndrome complaint as reported exposures might
or contact with the health-care system. become more significant when also reported by
subsequent complainants.
4.3.4. Epidemiology Process
4.3.5. Public Health Laboratory Process
Notification of group illnesses or independent
complaints can occur at the local, regional, Laboratory activities are not essential for
state, or national level. Some jurisdictions primary detection of outbreaks by this process
mandate reporting of unusual clusters of but are essential for determining etiology,
disease. Reports from health-care providers linking separate events during the investigation,
or other community members of unusual and monitoring the efficacy of control
4

clusters are triaged; occurrence of the same measures (see Chapters 5 and 6). Because of
disease is confirmed; data are analyzed; public health laboratory testing, links may
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

investigations are initiated; and control be seen across jurisdictional boundaries and
measures are implemented as appropriate. beyond; even national outbreaks may then be
For reports of group illness associated with an detected. For instance, an outbreak associated
event or venue, investigation generally involves with a particular restaurant may come to the
obtaining lists of attendees, confirming ill attention of authorities solely on the basis of
persons have the same disease, obtaining a report by a customer who observed illnesses
menus, interviewing cases, performing a cohort among multiple fellow patrons. Laboratory
or casecontrol study, and collecting food and testing and identification of Salmonella
patient specimens (see Chapter 5). Outbreaks Typhimurium can result in refinement of the
detected in this manner can be linked to other case definition used in this investigation, in
outbreaks or to other cases in the community additional testing and restrictions for workers
by a variety of processes, such as PulseNet or found to be carriers, or in connection of
the Foodborne Disease Outbreak Surveillance this outbreak with other outbreaks from a
System, and communication conducted contaminated commodity.
through Epi-X or the U.S. national network of
epidemiologists. 4.3.6. Strengths of Complaint Systems for
Outbreak Detection
Two or more persons with a common
exposure identified through interview of B
 ecause detection does not depend on
independent complaints are used to identify identification of an agent, this system can
clusters of illness in much the same manner as detect outbreaks from any cause, known
common agents are used in pathogen-specific or unknown. Thus, the complaint system
surveillance. Exposure information captured is one of the best methods for detecting
in the initial complaint generally is limited and nonreportable pathogens and new or
biased toward exposures shortly before onset reemerging agents. Recent examples include
of symptoms. Therefore, routine interviews recognition of sapovirus as a significant
are needed for this process to be robust. In agent in norovirus-like outbreaks4 and
the absence of common, suspicious exposures identification of Arcobacter butzleri as the likely
shared by two or more cases, complaints of agent in an outbreak of gastroenteritis at an
individual illness with nonspecific symptoms event.5 In one study, consumer complaint
such as diarrhea or vomitinggenerally are surveillance alone led to detection of 79% of
not worth pursuing. This underscores the confirmed foodborne outbreaks, including
need to collect and record sufficient exposure most norovirus outbreaks.4
2014 | Guidelines for Foodborne Disease Outbreak Response 127

4.3. Complaint Systems

F
 or event-related complaints only: recall For any true outbreak, the absence of an
of food items eaten and other exposures agent makes misclassification of cases more
by cases usually is good for reported events likely. Misclassification of cases makes
because items consumed at the event can identification of an association between an
be identified by menus or other means and outbreak and an exposure more difficult.
specifically included in the interview. W
 ithout a detailed food history (either from
C
 omplaint surveillance systems are the initial report or follow-up interview),
inherently faster than pathogen-specific surveillance of independent complaints is
surveillance because the chain of events sensitive only for short incubation (generally
related to laboratory testing and reporting chemical- or toxin-mediated) illness or
is not required (section 4.1.6). Exposure illness with unique symptoms because most

4
information gained through patient persons associate illness with the last meal
interviews has the potential for being high eaten before onset of symptoms and are

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
quality because patient recall is highest close thus likely to be correct only for exposures
to the exposure event. with short incubation times. This is not a
B
 ecause of the relatively limited number of limitation if full interviews are conducted.
exposures to consider (see 4.3.8.2 below), 4.3.8. Key Determinants of Successful
investigations of event-related notifications Complaint Systems
can be pivotal to solving widespread
outbreaks detected through pathogen- The following factors drive interpretation of
specific surveillance. Recent examples complaint surveillance data, affect the success
include an international outbreak of of investigations, and form the basis for best
Salmonella Bareilly and Salmonella Nchanga practices.
infections associated with a raw scraped
ground tuna product6 and a large outbreak 4.3.8.1. Sensitivity of case or event detection
of Salmonella Typhimurium infections The dynamics of outbreak detection differ
associated with peanut products.7 somewhat for notification involving groups
of illnesses and collection of independent
4.3.7. Limitations of Complaint Systems complaints. Detection of outbreaks by
notification of group illness is limited
N
 otification of illness in groups generally only by the severity of the illness, public
is less sensitive to widespread low-level awareness of where to report the illness,
contamination events than is pathogen- ease and availability of the reporting
specific surveillance because recognition of a process, and investigation resources (to
personplacetime connection among cases determine whether the clusters are in
by a health-care provider or member of the fact outbreaks). In contrast, detection of
community is required. clusters of illnesses from independent
T
 he value of complaints about single complaints relies on analysis by the
possible cases of foodborne disease in public health agency of an entire group
detecting outbreaks is limited by the of complaints collected over time. As
exposure information used to link cases with pathogen-specific surveillance, the
and by the lack of specific agent or disease size and number of outbreaks detectable
information to exclude unrelated cases. using independent complaints as primary
The illness reported by individuals might surveillance data are driven by the number of
or might not be foodborne, and illness individual cases reported, uniqueness of the
presentation might or might not be typical. illness or reported exposure, sensitivity and
128 CIFOR | Council to Improve Foodborne Outbreak Response

4.3. Complaint Systems

specificity of the interview process to detect Because exposures associated with


common exposures, and methods used to group events are relatively few and can
evaluate exposure data. be described specifically, recall tends to
be good and timing is less an issue than
4.3.8.2. Background prevalence of diseasegroup with pathogen-specific surveillance or
complaints independent complaints. In studies of food
When a group illness is reported, some of recall accuracy, the positive predictive value
the cases may be ill for a reason other than of individual food items ranged from 73%
a common group exposure. The likelihood to 97%.9,10 The negative predictive value
of this depends on the background ranged from 79% to 98%. Highly distinctive
prevalence of the disease or complaint. foods tended to be more accurately reported.
For example, unrelated diarrhea cases may Nonetheless, the more specific exposure-
4

inadvertently be grouped with true outbreak- related questions are, the better recall will be.
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

related cases because annually approximately For example, cases asked whether they ate
48 million persons in the United Statesor German potato salad at a particular event
one of sixnormally experience diarrhea.8 are more likely to remember than if they
Inclusion of misclassified cases (i.e., cases were asked whether they ate salad or asked
not associated with the outbreak) hinders the to list the foods they ate. Interviews of food-
detection of associations between exposures preparation staff additionally provide valuable
and disease, thus decreasing the likelihood information because they can list ingredients
of discovery of a common source. When that cases are not likely to recall or even know
reported clusters are small, the possibility must about and that a standardized questionnaire
be considered that the reported cluster results might not include. A good example is the 2011
from coincidence rather than causal association international outbreak of STEC O104:H4
(type I probability errori.e., detection of infections associated with fenugreek sprouts.11
an association between an exposure and a
disease where one does not exist). With unusual The second type of information gathered
syndromes, such as neurologic symptoms in the investigation of group complaints,
associated with botulism or ciguatera fish individual food histories, presents the same
poisoning, the likelihood of misclassification challenges as information collected for
and type 1 probability error is low. The system outbreaks detected through pathogen-specific
specificity can be increased by identifying a surveillance (i.e., includes a broad range
specific agent or disease marker or by increasing of possible exposures among cases and is
the specificity of the symptom information (e.g., associated with difficulties in recall). The
bloody diarrhea or specific mean duration of problems may be even greater because no
illness) or by obtaining exposure information. causative agent has been identified that would
enable investigators to focus on exposures
4.3.8.3. Sensitivity and specificity of case interviews previously associated with that pathogen.
group complaints Hence, cases should be interviewed promptly
Interviews of cases for group complaints for this aspect of the interview to be effective.
capture two types of information:
S
 pecific exposures associated with the 4.3.9. Complaint SystemsModel
reported event and Practices

I ndividual food histories of cases to rule This section lists model practices for
out alternate hypotheses and exclude notification and complaint systems. The
misclassified cases. practices used in any particular situation
2014 | Guidelines for Foodborne Disease Outbreak Response 129

4.3. Complaint Systems

depend on a host of factors, including the The complaint and subsequent interviews can
circumstances specific to the outbreak (e.g., lead to a hypothesis about the pathogen that
the pathogen and number and distribution leads to a different time frame for the exposure
of cases), staff expertise, structure of the history (e.g., vomiting leads to a different
investigating agency, and agency resources. hypothesis and exposure history time frame
For example, reports of bloody diarrhea may than does bloody diarrhea).
warrant aggressive case identification and
investigation to minimize serious illness and Health departments may choose to collect
death. A cluster of possible norovirus infections specimens from independent complaints or
might be investigated less aggressively or not encourage patients to seek health care.
investigated at all. Although these practices
4.3.9.2. Follow-up of food establishments named in
have not been systematically evaluated

4
individual complaints of possible foodborne illness
under different circumstances, experiences
In jurisdictions where visits are not required to

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
from successful investigations support their
every restaurant named in illness complaints,
value. Investigators are encouraged to use a
health department staff must decide
combination of these practices as appropriate
whether investigation of a commercial food
to the specific outbreak.
establishment is likely to be beneficial. To make
4.3.9.1. Interviews related to individual complaints this decision, investigators should consider
Detection of outbreaks based on multiple details of the complainants illness and the foods
individual complaints requires a system for eaten at the establishment. In the following
recording complaints and comparing food situations, investigation of a named commercial
histories and other exposures reported by food establishment might be warranted:
individuals.
T
 he confirmed diagnosis and/or clinical
A detailed 5-day exposure history is symptoms are consistent with the foods eaten
recommended for individual complaints and the timing of illness onset (e.g., a person
because common exposures are the sole in whom salmonellosis is diagnosed reports
mechanism to link cases. Although outbreaks eating poorly cooked eggs 2 days before
caused by agents with short incubation becoming ill).
periods may be able to be identified on the T
 he complainant observed specific food-
basis of information provided during initial preparation or serving procedures likely
complaints only, the signal-to-noise ratio would to lead to a food-safety problem at the
be low, and investigations would tend to be establishment.
nonproductive. Therefore, a detailed interview,
T
 wo or more persons with a similar illness
using a standardized form that includes both
or diagnosis implicate a food, meal, or
food and nonfood exposures, is preferred.
establishment and have no other shared food
Collection of a 5-day exposure history is also history or evident source of exposure.
recommended when an investigation begins
As noted below (section 4.3.9.6), regular review
that is based on multiple individual complaints.
of individual complaints is critical in recognizing
Given the ubiquity of norovirus infections, the
that multiple persons have a similar illness or
investigator should pay particular attention
diagnosis and share a common exposure.
to exposures in the 2448 hours before onset
whenever norovirus is suspected. As more Clues that a follow-up investigation of a food
information about the likely etiologic agent establishment is unlikely to be productive
is collected, this approach can be modified. include:
130 CIFOR | Council to Improve Foodborne Outbreak Response

4.3. Complaint Systems

C
 onfirmed diagnoses and/or clinical the laboratory that will do the analysis. Store
symptoms that are not consistent with the the food appropriately, but generally test the
foods eaten at the establishment and/or food only after epidemiologic implication or
the onset of illness (e.g., bloody diarrhea identification of specific food-safety problems
associated with a well-cooked hamburger through an environmental health assessment.
eaten the night before illness onset). Food samples that are frozen when collected
S
 igns and symptoms (or confirmed should remain frozen until examined. Samples
diagnoses) among affected persons that should be analyzed within 48 hours after
suggest they might not have the same illness. receipt. If sample analysis is not possible
within 48 hours, then perishable foods should
I ll persons who are not able to provide be frozen (40oC to 80oC). Storage under
adequate information for investigation, refrigeration can be longer than 48 hours, if
4

including date and time of illness onset, necessary, but the length of the storage period is
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

symptoms, or complete food histories. food dependent. Because certain bacteria (e.g.,
R
 epeated complaints by the same person(s) Campylobacter jejuni) die when frozen, affecting
for which prior investigations revealed no laboratory results, immediate examination of
significant findings. samples without freezing is encouraged. Food
samples can be collected as part of the process
4.3.9.3. Interviews related to reported illnesses in groups of removing suspected food from service.
Complaints of illness among groups
often are tantamount to outbreak reports. A Note: Food testing has inherent limitations
report of illness among 812 people who ate because most testing is agent-specific, and
together merits a different response than an demonstration of an agent in food, especially
isolated report of diarrhea. viruses, is not always possible or necessary
before implementation of public health action.
Focus interviews on the event shared by Detection of microbes or toxins in food is
members of the group. However, be aware most important for outbreaks involving
they might have more than one event in preformed toxins such as enterotoxins of
common, and explore that possibility. For Staphylococcus aureus or Bacillus cereus,
example, an outbreak associated with a wedding where detection of toxin or toxin-producing
reception might actually result from the organisms in human specimens frequently
rehearsal dinner, which involves many of the is problematic. In addition, organisms such
same people. Interviews should ask about other as S. aureus and Clostridium perfringens, which are
possible exposures either for the interviewee or commonly found in the human intestinal tract,
for others he or she might have contacted, such can confound interpretation of culture results.
as child-care attendance, employment as a food
worker, or ill family members. Furthermore, results of testing are often
difficult to interpret. Because contaminants
4.3.9.4. Clinical specimens and food samples related to in food change with time, samples collected
group illness during an investigation might not represent
Obtain clinical specimens from members food ingested when the outbreak occurred.
of the ill group. If the presumed exposure Subsequent handling or processing of food
involves food, collect and storebut do not might result in the death of microorganisms,
testfood from the implicated event. All multiplication of microorganisms originally
sampling must be conducted using legally present in low levels, or introduction of new
defensible procedures (e.g., chain-of- contaminants. If the food is not uniformly
custody) and using protocols as guided by contaminated, the sample collected might miss
2014 | Guidelines for Foodborne Disease Outbreak Response 131

4.3. Complaint Systems

the contaminated portion. Finally, because control centers, agricultural agencies, facility-
food usually is not sterile, microorganisms licensing agencies, and grocery stores. Identify
can be isolated from samples but not be the agencies/organizations in the community
responsible for the illness under investigation. that are likely to receive complaints.
As a result, food testing should not be routinely
undertaken but should instead be based on Improve communication and cooperation
meaningful associations identified through data among agencies that receive illness
analysis of interviews with suspected cases or complaints. Regular communication should
during environmental health assessments at the be established between agencies that receive
implicated food-service establishment. illness complaints, epidemiology staff, and
laboratory staff. Contact information should be
If food testing is determined to be kept current at all times. Because complaints

4
necessaryfor example, if a food has been might be made to multiple agencies, having

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
epidemiologically implicatedofficial reference a robust method of sharing information is
testing methods must be used at a minimum important. If possible, set up a database that
for regulated products (e.g., pasteurized eggs or public health agencies can access and review.
commercially distributed beef).
4.3.9.8. Other potentially useful tools
4.3.9.5. Establishment of etiology through laboratory Check complaint information against
testing national databases, such as the USDA/
Even though the etiology is not essential for Food Safety and Inspection Service (FSIS)
primary linkage of cases, as it is for pathogen- Consumer Complaint Monitoring System
specific surveillance, information about agents (CCMS). Recognizing that consumers are
is important for understanding the outbreak one of the many important resources for
and for implementing rational intervention complaint information possibly linked to its
and facilitates establishing links to other products, FSIS released a new online tool,
outbreaks or sporadic cases by PulseNet the Electronic Consumer Complaint Form
and the Foodborne Disease Outbreak (eCCF) to enhance its current surveillance
Surveillance System. Further information of the food supply. Before eCCF, consumer
about investigation methods and establishing complaints were reported to FSIS through
etiology is available in Chapter 5. its field offices or through calls to the USDAs
Meat and Poultry Hotline. The eCCF now
4.3.9.6. Regular review of interview data offers all consumers, including state and local
Review interview data regularly to look for health departments and schools receiving
trends or commonalities. Compile interview USDA-inspected products through the
data in a single database, and examine daily National School Lunch Program, an additional
for exposure clustering. Comparison with channel to report complaints to FSIS that is
exposure data obtained through pathogen- available 24 hours a day. Increased consumer
specific surveillance interviews might reveal a reporting through the eCCF will enhance FSIS
possible connection among cases and increase surveillance activities to characterize, prevent,
the sensitivity of both surveillance systems for and respond rapidly to potential threats from
detecting outbreaks. FSIS-regulated products.
4.3.9.7. Improvement of interagency cooperation and 4.3.9.9. Simplification of reporting process
communication To increase surveillance sensitivity, remove
Consumers may submit complaints to multiple barriers to reporting by making the
organizations and agencies, such as poison reporting process as simple as possible for
132 CIFOR | Council to Improve Foodborne Outbreak Response

4.3. Complaint Systems

the public. For example, provide one 24/7 toll- unfounded allegations of foodborne illness.
free telephone number or one website. Such
systems enable callers to leave information that 4.3.9.11. Centralized reporting or report review process
public health staff can follow up. Set up the reporting process so all reports
go through one person or one person
4.3.9.10. Increased public awareness of reporting routinely reviews reports. Centralization
process of the reporting or review process increases
Promote reporting by routine press the likelihood that patterns among individual
releases that educate the public about food complaints and seemingly unrelated outbreaks
safety, and advertise the contact phone will be detected.
number or website for reports of illness.
Use a telephone number that easily can 4.3.10. Multijurisdictional Considerations
4

be remembered or found in the telephone for Complaint Systems


FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

directory. Train food managers and workers


Outbreaks discovered through complaints
about the importance of reporting unusual
might span multiple jurisdictions, as
patterns of illness among workers or customers
evidenced by the 1998 parsley-associated
and food code requirements for disease
shigellosis outbreak and the 2006 multistate
reporting. Communicate the value of such
lettuce-associated E. coli O157:H7 outbreak
reporting, not just to protect public health,
in taco restaurants12. See Chapter 7 for
but also to protect food establishments from
Multijurisdictional Investigation Guidelines.

CIFOR Keys to Success:


Focus Area 4Complaint systems
Soliciting and receiving reports
A gency/jurisdiction has an established process for receiving reports from the public about
possible foodborne illness(es).
Public knows how to report possible foodborne illnesses to the agency/jurisdiction.
Agency/jurisdiction solicits reports of possible foodborne illness from other agencies and
organizations likely to receive these reports (e.g., poison control center, industry) inside and
outside the jurisdiction.
Agency/jurisdiction works with the local media to solicit reports of possible foodborne illness
from the public.
Detection of clusters/outbreaks
S taff collect specified pieces of information about each foodborne illness report and record the
information in an electronic data system.
Staff regularly review reports of foodborne illness to identify cases with common characteristics
or suspicious exposures that might represent a common-source outbreak.
Responding to complaints
S
 taff triage and respond to complaints in a manner consistent with the likely resulting public
health intervention (e.g., investigate reports of group illnesses more aggressively than isolated
independent illnesses).
Making changes
A
 gency/jurisdiction has performance indicators related to complaint systems and routinely
evaluates its performance in this Focus Area.
2014 | Guidelines for Foodborne Disease Outbreak Response 133

4.3. Complaint Systems

4.3.11. Indicators/Measures source outbreaks depends on multiple


interrelated processes. Indicators for assessing
The success of complaint-based surveillance and improving surveillance programs can be
systems at detecting and resolving common- found in Chapter 8.

4.4. Syndromic Surveillance

4.4.1. Overview and, therefore, takes less time)emergency


department chief complaint, ambulance
The utility of syndromic surveillance for dispatch, lab test orders. . Surveillance for

4
non-specific health indicators has not been specific syndromes, such as symptoms and

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
established for enteric disease surveillance non-pathogen related laboratory findings
and outbreak investigation. In theory, the associated with botulism or hemolytic
electronic collection of such indicators could uremic syndrome (HUS) generally fall in this
permit rapid detection of significant trends, category.
including outbreaks. In practice, the right mix
P
 ostdiagnostic datahospital discharge
of sensitivity and specificity has proven difficult
codes (ICD-9, ICD-10).
to find, and the utility of such systems may
be marginal. Surveillance for highly specific 4.4.4. Epidemiology Process
syndromes such as HUS or botulism is a
critical public health function. Epidemiology or emergency preparedness
groups evaluate alerts triggered by the
4.4.2. Background syndromic surveillance system. The
effectiveness of syndromic surveillance using
Syndromic surveillance is a relatively
non-specific health indicators in detecting
new concept, developed in the 1990s and
outbreaks has not been demonstrated.
expanded after the 2001 postal system anthrax
Presumably, cases would be interviewed
attacks in an attempt to improve readiness
and exposures determined if an alert were
for bioterrorism. One of the first systems
determined likely to represent a true outbreak.
implemented was in New York City in 2001.
4.4.5. Laboratory Process
4.4.3. Reporting
Laboratories do not play a direct role in
Syndromic surveillance typically relies on
preclinical syndromic surveillance. Various
automated extraction of health information:
types of laboratory data may be utilized for
P
 reclinical (i.e., not dependent on access clinical pre-diagnostic and post-diagnostic
to health care, consequently less specific data-based syndromic surveillance. Public
and potentially less useful)school and health laboratories would be involved during
work absenteeism, nurse help-lines, sales of epidemiologic investigations triggered by a
over-the-counter drugs, complaints to water syndromic surveillance signal.
companies, calls to poison control centers.
4.4.6. Strengths of Syndromic Surveillance
C
 linical prediagnostic (i.e., requires contact
with the health-care system but does not rely I n theory, syndromic surveillance using non-
on a full work-up or laboratory confirmation specific health indicators has the potential to
identify clusters of disease before definitive
134 CIFOR | Council to Improve Foodborne Outbreak Response

4.4. Syndromic Surveillance

diagnosis and reporting, thus generating M


 ore specific signals, such as discharge
a faster signal than can be expected with diagnoses, are less timely and do not appear
pathogen-specific surveillance. to offer advantage over standard surveillance
A
 s with complaint systems, outbreaks from methods.
any cause, known or unknown, potentially T
 he usefulness of syndromic surveillance
can be detected. Included are clusters of using non-specific health indicators has not
cases identified with discharge diagnoses that been demonstrated for foodborne disease.
include specific agents not part of standard After examination of 2.5 million patient
surveillance. records in its first year of operation, the
S
 yndromic surveillance may be able to New York City surveillance system identified
detect large, undiagnosed events, such as 18 diarrhea or vomiting alerts during
4

an increase in gastrointestinal illness among three outbreak periods. Five institutional


outbreaks were identified during one of
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

persons of all ages consistent with norovirus,


an increase in diarrheal illness among young these periods, but whether the data were
children consistent with rotavirus, and the sufficiently specific to allow for public health
arrival of epidemic influenza. intervention is not clear.13,14,15

M
 ost syndromic surveillance systems have T
 he cost of developing syndromic
been built with automated electronic data surveillance systems is substantial, and
transfer. This infrastructure should be useful if development occurs at the expense
for other types of surveillance and public of maintaining or upgrading routine
health activities. surveillance, results of surveillance are
degraded, rather than enhanced.
V
 ery specific syndromes, such as botulism
or HUS, are important indicators of serious 4.4.8. Key Determinants of Successful
public health problems. Surveillance Syndromic Surveillance Systems
for specific syndromes with or without
identification of an agent is a critical The following factors drive the interpretation
function of health agencies, and is not of syndromic surveillance data, affect the
subject to artifacts introduced by changes in success of investigations, and form the basis for
microbiology testing methodologies. best practices.

4.4.7. Limitations of Syndromic 4.4.8.1. Specificity and speed


Surveillance Although the potential speed of syndromic
surveillance is its chief strength, speed is
L
 ack of specificity for most syndromic inversely proportional to the specificity
surveillance indicators in the area of of the indicator disease information.
foodborne disease makes for an unfavorable Preclinical information, such as sales of over-
signal-to-noise ratio, meaning that only the the-counter drugs is generally available sooner
largest events would be detected, and many and is less specific than clinical, prediagnostic
false-positive signals would be expected. signals (such as laboratory test orders).
Responding to false-positive signals drains an Prediagnostic signals, in turn, are available
agencys resources substantially. sooner and are less specific than postdiagnostic
E
 valuating a signal usually means cross- signals (such as hospital discharge data).
checking it with routine surveillance Lack of specificity at any level results in type
reports, meaning it cannot replace routine 1 probability error (the suggestion of an
surveillance.
2014 | Guidelines for Foodborne Disease Outbreak Response 135

4.4. Syndromic Surveillance

association between a signal and a significant Insurance Portability and Accountability


health event when, in fact, none exists) and Act (HIPAA). Respondents noted that many
type 2 probability error (the lack of signal health-care providers and medical staff did
suggests a disease event is not occurring, not understand HIPAA and so tended to
when, in fact, it is). Less specificity means give minimal patient information. Questions
that more cases are needed to overcome also were raised about whether syndromic
background noise and that false-positive surveillance falls under the same regulations
alerts are likely. as reports of diagnosis-related disease. For
example, whether health departments have
The most specific signalshospital discharge the legal authority to collect these data is
datainclude both nonspecific diagnoses (e.g., not always clear. Most respondents were
diarrhea of infectious origin, ICD-9 #009.3) using current disease reporting regulations

4
and diagnoses based on specific agents (e.g., to cover syndromic surveillance. Many

A ND O U TBR E A K D ETE C TI O N
FO O D BO RNE D I S EA S E S U R VEI L L A NC E
Salmonella gastroenteritis, ICD-9 #003.0). respondents believed more specific syndromic
Discharge signals for reportable disease, such indicators are needed to incorporate them
as salmonellosis, should not offer any time into regulations. Most agencies that had
advantage over standard surveillance methods implemented a syndromic surveillance
because: system used deidentified data, which slows
investigations of positive signals from the
T
 he diagnoses requires agent identification
surveillance system.17
and would have the same limitations as
pathogen-specific surveillance, 4.4.9. Practices for Improving Syndromic
S
 tandard investigation probably would be Surveillance
required for public health action, and
Because the usefulness of syndromic
I dentification of illness may precede surveillance for detecting foodborne
discharge. disease events has not been demonstrated,
the need for additional investment is not
Signals from rare, specific syndromes without
clear, especially if these systems compete
laboratory confirmation, such as botulism-like
for resources with underresourced
syndrome, should be as effective as pathogen-
standard surveillance systems. If an agency
specific surveillance. This is the basis for the
implements or seeks to improve a syndromic
national botulism surveillance program at
surveillance system, it needs to consider the
CDC, which provides emergency clinical,
following practices:
epidemiologic, and microbiologic consultation
and antitoxin treatment for persons with B
 etter electronic and process integration
suspected botulism because of the extremely with standard surveillance systems might
serious nature of that illness and the possibility improve usefulness.
that one case might herald other cases from
the same exposure.8,16 (http://www.cdc.gov/ S
 yndromic surveillance data are most useful
ncidod/dbmd/diseaseinfo/files/botulism.PDF). when corroborated with data from multiple
sources (e.g., increased sales of over-the-
4.4.8.2. Personal information privacy issues counter diarrheal medicines associated
In a survey on implementation of syndromic with a rise in emergency department chief
surveillance systems, more than half (54.2%) complaints of diarrhea). As historical data
of respondents reported some or substantial accumulate, fine-tuning detection algorithms
problems caused by real or perceived patient to reduce false-positive signals might be
confidentiality concerns and the Health possible.
136 CIFOR | Council to Improve Foodborne Outbreak Response

4.5. References

1 Kupferschmidt K. Outbreak detectives embrace the 10 D


 ecker MD, Booth AL, Dewey MJ, Fricker RS,
genome era. Science 2011;333:18189. Hutcheson RH, Schaffner W. Validity of food
2 Guerrant RL, Van Gilder T, Steiner, TS, et al. Practical consumption histories in a foodborne outbreak
guidelines for the management of infectious diarrhea. investigation. Am J Epidemiol 1986;124:85963.
Clin infect Dis 2001;32:33150. 11 K
 ing L, Nogareda F, Weill F, et al. Outbreak of Shiga
3 American Medical Association, American Nursing toxinproducing Escherichia coli O104:H4 associated
Association, CDC, FDA, USDA/FSIS. Diagnosis with organic fenugreek sprouts, France, June 2011.
and management of foodborne illnesses: a primer for Clin Infect Dis 2012;54:158894.
physicians and other health care providers. Available at 12 C
 enters for Disease Control and Prevention.
www.cdc.gov/mmwr/pdf/rr/rr5304.pdf (accessed Outbreaks of Shigella sonnei Infection Associated
October 15, 2013). with Eating Fresh Parsley United States and
4 Lee LE, Cebelinski EA, Fuller C, et al. Sapovirus Canada, July-August 1998. http://www.cdc.gov/
outbreaks in long-term care facilities, Oregon and mmwr/preview/mmwrhtml/00056895.htm
4

Minnesota, USA, 20022009. Emerg Infect Dis (accessed January 6, 2014).


FO O D BO RNE D I S EA S E S U R VEI L L A NC E
A ND O U TBR E A K D ETE C TI O N

2012;18:8736. 13 D
 as D, Metzger K, Heffernan R, et al. Monitoring
5 Lappi V, Archer JR, Cebelinski E, et al. An outbreak over-the-counter medication sales for early detection
of foodborne illness among attendees of a wedding of disease outbreaksNew York City. MMWR Morb
peception in Wisconsin likely caused by Arcobacter Mortal Wkly Rep 2005;54(Suppl):416.
butzleri. Foodborne Pathog Dis 2013;10:2505. 14 H
 effernan R, Mostashari F, Das D, et al. New York
6 Centers for Disease Control and Prevention. Multistate City syndromic surveillance systems. MMWR Morb
outbreak of Salmonella Bareilly and Salmonella Nchanga Mortal Wkly Rep 2004;53(Suppl):237.
infections associated with a raw scraped ground tuna 15 B
 esser JM; Systems to detect microbial contamination
product (final update). www.cdc.gov/salmonella/ of the food supply. Institute of Medicine Forum on
bareilly-04-12/index.html (accessed October 2, Microbial Threats. Addressing Foodborne Threats to
2013). Health; Policies, Practices, and Global Coordination.
7 Cavallaro E, Date K, Medus C, et al. Salmonella 2006. Washington, DC: National Academies Press;
Typhimurium infections associated with peanut 2006:17889.
products. N Engl J Med 2011;365:60110. 16 C
 DC. Botulism in the United States, 18991996.
8 Centers for Disease Control and Prevention. In: Handbook for epidemiologists, clinicians, and
Estimating foodborne illness: an overview. www.cdc. laboratory workers. 1998. Available at http://
gov/foodborneburden/estimates-overview.html www.cdc.gov/ncidod/dbmd/diseaseinfo/files/
(accessed October 3, 2013). botulism.PDF(accessed January 6, 2014).

9 Mann JM. A prospective study of response error 17 D


 rociuk D, Gibson J, Hodge J. Health information
in food history questionnaires: Implications for privacy and syndromic surveillance systems. MMWR
foodborne outbreak investigation. Am J Public Health Morb Mortal Wkly Rep 2004;53(Suppl);2215.
1981;71:13626.