Conducting an Epidemiologic Investigation

Introduction

Epidemiology is “the study of the distribution and determinants of disease frequency in human populations.”  It is the collection and analysis of information to determine whether an association exists between one or more exposures to possible infection and the occurrence of disease.  In practice, epidemiologists often employ statistics to look at who gets sick or injured and why. 

The epidemiologic investigation is an important part of a complete foodborne illness investigation that also includes environmental (see Chapter 8) and laboratory (see Chapter 9) investigations.  Each part complements the other and teamwork and open communication is of utmost importance. 

This chapter addresses the steps in an epidemiologic investigation. The purpose of an epidemiologic investigation is to halt ongoing problems, to prevent further illness, and to learn from the situation how to prevent similar outbreaks in the future. This is done by collecting data, and formulating and testing hypotheses and developing control strategies.  It involves the collection and analysis of data to determine the cause of illness and to implement controls to prevent additional illness.

A questionnaire is often used to assist the investigator in developing better hypotheses about the etiologic agent’s identity, source and transmission. Investigators interview ill and well persons, and calculate and compare rates of illness in both groups.  They make time, place, and person associations and calculate the probability that a particular exposure (food or beverage) was responsible for illness.   

The investigator incorporates results from epidemiological discovered associations and the environmental and laboratory investigations, and uses these data to form and test hypotheses.  Careful development of epidemiologic inferences with persuasive clinical and laboratory evidence and biologic plausibility will almost always provide convincing evidence of the cause agent, source and mode of spread of a disease.  When food and stool testing are negative, the cause of an outbreak is often determined by epidemiological association. 

It is essential to maintain interagency collaboration in the investigation of outbreaks of foodborne illness.  Communication of the results of each agency’s investigation, including those of the Department of Inspections and Appeals (DIA), Iowa Department of Public Health (IDPH),  and University Hygienic Laboratory (UHL) is vital to ensuring a thorough multifaceted investigation.

Epidemiologic Outbreak Investigation

There are essentially 12 steps in an outbreak investigation. The steps are presented here in conceptual order. In practice, however, several may be done at the same time, or they may be done in a different order.

1.  Establish the existence of an outbreak

The first step should almost always be establishing whether or not there is an outbreak going on.  The following questions can help determine the existence of an outbreak:

• Are there two or more people from different households with the same clinical illness resulting from the ingestion of the same food or meal?
• Are the clinical signs and symptoms, along with the incubation period, consistent with a foodborne disease agent?
• Is the number of illnesses more than what would be expected in this group of people and in the population as a whole?
• Are there any other reports of potentially associated cases? 

Answering yes to any of the questions above may warrant further investigation.To determine the number of cases of a disease expected in a given time frame, the current number can be compared with the number of cases from the previous few weeks or months, or from a similar time in the previous few years. Under Iowa Administrative Code 641, Chapter 1, 51, certain infectious diseases must be reported by healthcare providers and testing laboratories to the Iowa Department of Public Health (IDPH).The code requires reporting of many foodborne diseases. This surveillance data could be used to help determine if an outbreak is occurring. Refer to Chapter 5, Conducting Surveillance, for more information on routine disease surveillance.

Further investigation may be warranted, including an epidemiological, environmental, and laboratory investigation. It is often unclear when to conduct a full epidemiologic investigation. It is usually clear when a large number of people get ill at approximately the same time after eating at the same establishment or attending the same event that an investigation is needed.  However, uncertainty arises when sporadic complaints are reported.  Evidence of an association must be determined, necessitating the need to follow-up (i.e., determine the validity of and initiate further action if necessary) on every complaint. It may be that single complaints are related to an outbreak.  Refer to Chapter 6 for more information on handling single complaints, what information to collect and how to collect it.

When an incident exists in which illness has passed and no new cases have been identified, the decision to conduct an epidemiologic investigation should be based on an assessment of what will be gained from it.  As stated, an investigation is always a learning tool.  But, if resources are not available (time, personnel, etc.) or other concerns are more pressing, a full investigation may not be warranted.  Rather, appropriate control measures should be taken to prevent future outbreaks and stop ongoing transmission.

This is especially true of home-based foodborne outbreaks.  In many instances, the illness is confined to a finite number of people in a discrete period. Also, notification sometimes comes after the fact when little material is left for testing, and people have recovered. Food-preparation techniques should be reviewed with responsible parties and the opportunity taken to educate them on proper food handling and preparation.

2.  Verify the diagnosis

This requires a review of the symptoms, incubation periods, length of illness, and laboratory results of patients to determine if they all have the same disease. Knowing the disease can help form a hypothesis about the source of illness.   The first reports about an outbreak may be presumptive based on clinical syndrome and diagnosis. Laboratory results might not be available. Chapter 9, Conducting a Laboratory Investigation, has more information on verifying a diagnosis. 

3.  Develop a case definition

A case definition is a standard set of criteria that should be applied to each individual to decide if they should be classified as a case or not.   A case definition includes:

• Clinical criteria – diagnosis, signs, symptoms, and/or laboratory tests
• Restrictions on time, place, and person

It should not include the parts of the hypothesis, i.e. the exposure or risk factor.

Criteria for the case definition should be simple, practical, and objective. The case definition must be applied to all persons without bias. Initially, it should be broad enough to include most, if not all, actual cases (sensitivity) without including many “false-positives” (specificity).  As more information is gathered, the case definition can be narrowed to increase specificity.

Case Definition Example: 

Persons who developed vomiting or diarrhea on January 1 -3, 2001 and who belong to the church on 45th Street.  

4.  Do a Quick and Dirty Study

Once an outbreak is suspected, it is important to describe the outbreak as soon as possible.  Questions that need to be addressed include:

• The extent of the outbreak – local, statewide, national
• The approximate number of people ill?
• Commonalities – similar or potential exposures
• People affected – school children, IV drug users, people who shop at a particular deli

Answering these questions requires finding as many cases as possible.

Why is case finding important?

The cases reported to the health department may represent only a small portion of the total number of actual outbreak-related cases.  While the reported cases are important, they may not adequately represent all of the people affected.

• To get the full scope of an outbreak, investigators need to know exactly what types of people are getting sick, when they became symptomatic, and where they may have been exposed.
• Finding additional cases will help start refining the case definition.
• Knowing the extent of the outbreak determines the number of resources to allocate to the investigation.
• Case finding helps develop appropriate control measures by defining the population with the exposure of interest.
• When looking for cases at the beginning of an outbreak, it is best to cast a wide net, which can help determine the size and geographic boundaries of the outbreak.

Cases can be identified through active or passive strategies.  Active strategies include soliciting information from health facilities and laboratories and screening an exposed population with a diagnostic test.  Passive strategies include examining county or state surveillance data to identify cases through the mandatory disease-reporting system.

Examples of strategies to find cases:

• Ask health-care providers, clinics, hospitals, and laboratories for information.  Investigators can visit emergency rooms and ask to review records of all patients seen with the illness.  Clinicians can be asked to submit specimens on patients meeting a certain case definition.  Infection-control practitioners can be asked to review the medical records of patients with a certain diagnosis.
• Utilize the media (television, radio, newspapers) to identify people who may have been exposed to a food establishment (imagine a food handler with hepatitis A working at a popular restaurant) or a contaminated food product (recalls of almonds, strawberries, ground beef, etc.).
• Identify potential cases through reviews of wedding invitation lists, guest books, or credit card receipts.
• If the outbreak occurred in a defined setting (church picnic, school event, cruise), it may be effective to interview everyone on symptoms.
• Ask patients if they know of others with a similar illness.

Challenges to case finding may include:

• Laboratory testing for certain pathogens (e.g., Norovirus, Escherichia coli O157:H7) is difficult.
• Diseases that cause a spectrum of symptoms, ranging from mild to severe.  People with mild symptoms may not seek medical attention, removing the possibility of finding them through medical records.  Likewise, people with mild illness may not connect their illness to a publicized outbreak.
• The exposed population may not be well defined.
• Groups involved may not be cooperative.

Information to collect during case finding:

Although the specific information gathered will depend on the outbreak, information can be grouped into four categories:

• Identifying information.
• Demographic information.
• Clinical information.
• Risk-factor / exposure information .

Identifying information includes name, address, phone number, and other contact information.  Demographic information can include age, gender, race, occupation, place of employment, etc.  Clinical information can include symptoms, onset and duration of symptoms, lab findings, and severity of illness.  Risk-factor information is usually based either on well-established risk factors (e.g., in an outbreak of E. coli O157:H7, risk factors would include consuming certain foods such as ground beef, lettuce, alfalfa sprouts, recreational water exposure, and child-care attendance) or potential risk factors (e.g., items served at a reception, menu items at a restaurant, etc.).

5.  Develop a hypothesis

A tentative hypothesis is constructed from time, place, and person associations and is the basis for the initial outbreak definition and case definition.  The hypothesis should be written as soon as enough information is available. 

A hypothesis is a theory or speculation to explain how an event occurred.  It is a statement that can be tested and refuted or accepted.

A hypothesis should address:

• Agent – What is the organism or chemical causing disease?
• Source – What is the source of the outbreak?
• Mode of transmission
• Exposure periods – Who is at risk of becoming ill?
• Possible contributing factors that led to illness – food items, travel, water/environmental exposures, contact with ill persons

The hypothesis should be based on information gathered during initial “quick and dirty” interviews.

Investigators can use the knowledge of the disease – reservoirs, transmission, common vehicles, known risk factors – to develop a hypothesis.

Example:

• A physician calls the health department about two patients with diarrhea, fever, and headaches.  Both patients had attended a dinner at a lodge on March 6th.  They also knew of others who had attended the dinner that were ill.  Considering the onset dates of the two patients and the date of the dinner, the possible incubation period ranged from 2 to 6 days.  With this preliminary information, an initial hypothesis may be: The lodge dinner on March 6th caused people to become ill with diarrhea and/or fever within 6 days of the event.

6.  Plan a detailed investigation

Select a study design

Two main types of studies are used in foodborne-outbreak investigations – cohort and case-control.

Cohort study

A cohort study has a well-defined population, working from exposure (ate the salad bar or not) to outcome (ill or not ill).  Cohort can also used to describe the group in the investigation as well as the type of investigational study.  A majority of the foodborne investigations are cohort studies because they are related to group gatherings, which are more likely to be reported.  Cohorts would include people attending weddings, funerals, schools, office parties, etc.

Case-control study

Case-control studies involve working with an unknown population. They start from an outcome (has salmonella) and attempt to discern the exposure(s) (ate undercooked chicken). Cases are usually identified through surveillance or astute clinicians. To determine the source of the infection, people who are similar to the cases (e.g., family members, friends, neighbors) but are not ill are selected as controls.  The cases and controls are interviewed about possible exposures to identify an exposure that is associated with being ill.   Calculations can be done to determine the appropriate sample size to reach statistical significance.

7.  Develop a Questionnaire

A foodborne outbreak questionnaire contains three sections: demographic information, illness status/clinical information, and exposure information.

Demographic information

• Name
• Address
• Phone number
• Age/date of birth
• Gender
• Race/ethnicity
• Grade/teacher (if a school outbreak)
• Room number (if a hospital or long term care facility)

Illness status/clinical information

• Ill – yes/no
• Symptoms (diarrhea, vomiting, fever (highest measured temperature), abdominal cramps, etc.)
• Onset of symptoms (date and time – specify a.m. or p.m.)
• Resolution of symptoms (date and time – specify a.m. or p.m.)
• Willingness to provide a clinical specimen
• First symptom
• Worst symptom
• Health-care provider visit (Include provider name and contact information.)
• Hospitalization (Include length of stay and facility.)
• Specimens collected (Include date, type, and where it was sent.) and result

Exposure information

• Attendance at event in question
• Time ate at event
• Food history (Obtain menu and list all foods and beverages served.  Make sure to include condiments, ice, sugar, cream, etc.), include yes, no and unknown options
• Serving size (none, bite, half, one, two, three or more) of food and beverages
• Location of other meals consumed during incubation
• Attendance at any other gatherings (parties, sporting events, church events, etc.)
• Their opinion of what made them ill
• Their knowledge of others who are ill  (Get names & contact information.)

Questionnaires must be tailored to the outbreak at hand.  Have people answer all questions, regardless of illness status. It is just as important to get information on those who are well and never got ill so there is a comparison group available for the ill in the analysis component of the investigation. Try to limit the number of blank answers.  If the person did not eat a particular food, be sure to select the ‘no’ option instead of leaving it blank. Typically, questions with set answers are better than open-ended questions because it makes data analysis easier.  Other questions can be added if the hypothesis is incorrect.  Therefore adding questions on where people ate during the entire incubation period is very important.  Sample questionnaires can be found in Appendices C Sample Forms. 

8.  Interview Ill and Well

Once the questionnaire has been completed, it is time to interview the cases. There are three main types of interviews – face-to-face, phone, and self-administered.  Each method has advantages and disadvantages and the type used will depend on the situation.

Face-to-face

Advantages:

• Higher response rates
• Can use a more complex questionnaire design (including skip patterns)
• Usually gives more accurate recording of responses

Disadvantages:

• Need to find subjects and arrange meetings – can be time consuming
• Less anonymous – subject may be uncomfortable
• Subjects may give answers they think the interviewer wants to hear (This happens with all types of questionnaires, though it’s probably more common in face-to-face interviews.)
• Potential for the interviewer to influence the process
• Most costly of the interview types

Telephone

Advantages:

• Less time required to track down subjects
• Higher response rate than with self-administered
• Can use a more complex questionnaire design
• More accurate recording of responses
• Less costly than face-to-face

Disadvantages:

• Less anonymous than self-administered
• Potential for the interviewer to influence the process
• More costly than self-administered

Self-administered

Advantages:

• Most anonymous
• Subjects may be more honest
• Less investigator time required
• Least expensive

Disadvantages:

• Questionnaire needs to be self-explanatory and easy to complete
• Usually results in poorer data quality
• Lower response rates

Group interviews can also be conducted with an outbreak has occurred in a school setting.  Having people raise their hands to the yes/no questions can also provide a quick analysis of what may have occurred in an outbreak. 

In face-to-face and telephone interviews, it is helpful to have a meeting (or conference call) with everyone who may be administering the questionnaire on how to administer it, what information is expected and which data are especially important to collect.  Having a written script for the interviewer can help increase consistency.  For example:

 Hello, my name is _______. I’m with_______. The Iowa Department of Public Health and the County X Health Department are investigating an outbreak of illness characterized by diarrhea and vomiting among persons who attended X event from Date to Date.  In order to determine if the cause of this illness was food-related and to prevent such an occurrence in the future, we need to get more information from those persons who were at the X event during those dates, whether ill or not.  All information will be kept confidential. 

It is important to conduct a pilot test with the questionnaires before contacting and completing interviews with all the cases. Have each interviewer interview one case and then have all the interviewers meet again to discuss any discrepancies in the questionnaire. Once the discrepancies have been fixed, proceed forward with all the interviews.  Also determine with all the interviews how to answers will be recorded including the “I don’t know” answers. 

9.  Compile and orient the data by person, place, and time

After collecting the information, systematically organize key information by person, place, and time. Often, compiling and orienting the data may occur in the field, but this typically occurs in-house.

Person

• Age
• Gender
• Symptoms
• Race/ethnicity
• Socio-economic status
• Behavioral risk factors

Place

• Geographic distribution (include a spot map if possible): Where cases live, work, attend school, hospital/LTC room
• Clusters or patterns of cases
• Common venues (concerts, sporting events, etc.)

Time

• Onset of symptoms
• Incubation period
• Infectious period
• Seasonality
• Baseline vs. epidemic

10.  Create an Epidemic curve

An epidemic curve is a graph of the number of cases meeting the case definition (y-axis) by date or time of onset of illness (x-axis).

• It is a histogram, not a bar graph.
• It can identify outliers (i.e., cases that stand apart).
• The unit of time for the x-axis is based on the incubation period of the disease under investigation and the time over which the cases are distributed. It can be by hours, days, weeks (Time frames can be varied/adjusted to create the most appropriate curve.).
• The shape of the curve can suggest how the disease was introduced, how the disease spread – point source or continuing source – and can also identify secondary cases.

If the disease is known, or if the agent is highly suspected, but an exposure is unknown the epidemic curve and incubation period can be used to identify the most likely period of exposure:

• Identify the peak of the outbreak and count back on the x-axis the average incubation period.
• Identify the last case and count back the maximum incubation period.
• Identify the earliest case and count back the minimum incubation period.

Ideally, the three dates are similar and represent the most likely period of exposure.  This technique is not precise, however, and widening the period on either side of these dates is recommended.

Interpreting Epidemic Curves 

A point source outbreak should show a tight clustering of cases in time with a sharp upslope and a trailing downslope on the epidemic curve.   All cases originate from a single, common cause and should occur during one incubation period. 

In a point-source outbreak, infected people may be sources of infection to others.  Secondary cases may appear as a prominent wave following a point-source outbreak by one incubation period.

Continuing common source:  Outbreaks may arise from a common source that continues over time.  The epidemic curve will rise sharply as with a point source. However, rather than rise to a peak, it will plateau.  The downslope may be sudden, if the common source is removed, or gradual, if it exhausts itself.

11.  Organize the data

All questionnaires should be reviewed for completeness and data quality before data entry begins.  Attempts should be made to gather missing or incomplete data.  Questionnaires that are not properly completed should be excluded from the data analysis.

This sometimes can be done by hand. If not, then completed questionnaires can be entered into a statistical analysis program, such as Epi Info, SAS, SPSS, etc. If the number of people affected is small, data analysis by hand may be preferred and data entry into a statistical analysis program can be skipped. Data should be tabulated in some way (graph paper, etc.), however, to make analysis easier.  After data entry is complete, frequencies should be run on all variables to check for data entry errors.  

12.  Analyzing the data

Once the data has been collected, it is time to analyze the data using analysis programs, such as, Epi Info, SAS, SPSS, etc.  The analyzing and interpreting data can be difficult and confusing and is not meant for individuals who have never had a basic statistics course. The Center for Acute Disease Epidemiology (CADE) can help analyze and interpret the data from the questionnaires.  Individuals with the appropriate epidemiology and statistical training can refer to the Appendices of this Chapter for detailed information on data analysis of foodborne outbreak investigations.

Epidemiology Outbreak Investigation Conclusions

Epidemiology can implicate vehicle and guide appropriate public health action, but laboratory evidence can secure the findings. The epidemiological results should always be considered along with the laboratory and environmental investigations. If the epidemiological results support that the chicken was the vehicle of transmission in the Salmonella outbreak, verify this with the results of the environmental and laboratory investigation.

Conclusions about the source of disease should take into account these main ideas:

• Whether symptoms experienced by patients were consistent with those commonly produced by the etiological agent.
  • e.g. Patients had diarrhea (some had bloody), stomach cramps, fever, nausea, and some had vomiting – consistent with the symptoms of Salmonella.
• Whether the organism/toxin was isolated from the cases;
  • e.g. Salmonella was isolated from a clinical specimen of a person who became ill after eating the chicken at the event.Whether the implicated vehicle (eg. food, utensils, animals, etc) had the same organism or toxin as found in specimens from the cases; e.g. The same Salmonella species was isolated from both the chicken and in the ill persons.
• Information on methods of food processing, preparation and storage and whether these provided opportunities for contamination, survival and growth of the organism. 
  • e.g. The environmental investigation uncovered errors in heating preparation of the chicken which allowed the Salmonella organism to survive.  The environmental investigation also noted errors in the hot holding process of the chicken that provided an opportunity for growth of the Salmonella organism.