Generating Hypotheses about Likely Sources

Pathogens that cause foodborne illness can also spread by contaminated water or by direct contact with an ill person or direct or indirect contact with an infected animal. When looking for the source of the illness, investigators first need to decide on the likely mode(s) of transmission. The pathogen causing illness, where ill persons live or how old they are may suggest the mode of transmission, and even a specific source. Hypothesis generation should be thought of as an iterative process in which possible explanations are continually refined or refuted.

When exposure to a food is suspected, the investigators next must consider the large number of foods that may be the source or vehicle of infection. The number of different food items is vast, so the investigation needs to narrow the list to the foods that the ill people actually ate before they got sick, and then further narrow it to the specific foods that many of the ill people remember eating. Health officials interview persons who are ill to find out where and what they ate in the days or weeks before they got sick. These interviews are called "hypothesis-generating interviews."

The time period they ask about depends on the pathogen's incubation period—the time it takes to get sick after eating the contaminated food. This varies for different pathogens. Which foods they ask about depends on what investigators already know about the exposure. If several cases have occurred at a restaurant, hotel, or catered event, for instance, interviews will focus on the menu items prepared, served, or sold there. If there is no obvious place of exposure or subcluster of cases identified, investigators may use a standardized questionnaire, also known as a "shotgun" questionnaire.

A shotgun questionnaire may include:

• Questions that ask whether a person ate any of a long list of food items
• Open-ended questions that review each meal a person ate in the days before illness began
• Questions about food shopping habits, travel, restaurant dining, and attendance at events where food was served

From the interviews, investigators create a short list of the foods and drinks that many ill persons had in common. Foods that none or very few of the sick people reported eating are considered as less likely to be the source. Investigators then look at other information, such as the results of any food testing, past experience with the suspected pathogen, and the age or ethnicities of ill persons. Based on all the information they gather, the investigators make a hypothesis about the likely source of the outbreak. However, shotgun interviews can only suggest hypotheses that are contained on the questionnaire. This approach may not lead to any refined testable hypothesis. Intensive open-ended interviews can help in this situation.

Coming up with a hypothesis is often challenging and may take time for several reasons. First, interviews of ill persons are highly dependent on their memories. The time from the start of illness to knowing that the ill person was part of an outbreak is typically about 2–3 weeks. Ill persons may not remember in detail what they ate that long ago. Also, when the contaminated food is an ingredient (such as eggs, spices or herbs, or produce in a salsa), the task becomes even harder. People often don't remember or know the ingredients of the foods they ate. These challenges may prevent a hypothesis from quickly appearing. In some cases, ill persons may be interviewed multiple times as new ideas arise about possible sources. It can sometimes be helpful to visit someone's home and look at the foods in their pantry and refrigerator, or to get their permission to review the information from their shopper cards.

A useful method for generating hypotheses in large, multistate outbreaks includes rapid and thorough investigation of restaurant clusters; these cluster investigations are critical to identifying specific food vehicles and provide detailed ingredient content and information on sources of food items for traceback investigations. However, delays inherent in the current system of surveillance for investigation of foodborne disease outbreaks contribute to the time it takes to recognize clusters. The median time from illness onset to submission of the PFGE pattern of patients' Salmonella isolates to PulseNet should be decreased, improved interview formats should be used, and interviews that identify restaurant clusters should be conducted sooner. Conducting and completing interviews with a standardized questionnaire in a timely manner remains a critical deficiency during multistate outbreak investigations, partly because such interviews can lead to identification of clusters. Questionnaire data are typically submitted by facsimile to CDC for lack of a flexible and rapidly configurable electronic data gathering platform.