Attribution of Foodborne Illness: Methods and Data Sources

There is no single data source or analytic method that is best for estimating source attribution for all agents. CDC and its partners use several data sources and combinations of methods to estimate the number of illnesses associated with each type of food. Some are highly specific for one type of infection or one type of food, and others are more general. For example, by comparing the characteristics of Salmonella found in foods and animals with those found in people, we can attribute Salmonella infections back to specific animal sources. The best method may differ for different agents and foods.

Analysts use models to estimate the major food sources for illnesses. They use data from outbreaks, from sporadic illnesses, and from subtyping of pathogens:

Outbreaks: Outbreak data provide information about the role of various foods in causing illnesses. However, outbreaks contribute only a small proportion (less than 5 percent) of lab-confirmed foodborne illnesses. Only using outbreak data assumes that the food sources are the same for outbreak-related illnesses and the more common sporadic (non-outbreak) illnesses, which is more accurate for some food-pathogen combinations than others.

Sporadic illnesses: Most health departments do not collect detailed data about exposures of individual people with illnesses, such as salmonellosis, that are often foodborne. Because most people do not know what made them sick, illnesses reported to a health department usually do not reveal the likely source. CDC and health departments do intensive studies that compare the exposures of people with sporadic cases of a particular infection, such as E. coli O157 1, with the exposures of other people who are not ill.

  • By comparing exposures reported by the two groups, we can determine which foods likely are responsible for some of the illnesses. This approach is called a case-control study.

Pathogen subtype: Pathogens of the same type are not exactly alike. In the laboratory, small differences between pathogens can be used to subtype them. The subtype is a marker, like a fingerprint, that sometimes can help distinguish strains from different sources.

  • Investigators can compare the subtypes of pathogens isolated from animals and foods with the subtypes isolated from infected humans. They can combine this information with data on how often people consume certain foods, and then use mathematical models to estimate the number of illnesses associated with each source.
  • CDC and the U.S. Department of Agriculture’s Food Safety and Inspection Service (FSIS) recently collaborated to adapt a pathogen-subtype model for Salmonella infections to US data 2.

1FoodNet 2011 Publications

2 Guo C, Hoekstra RM, Schroeder CM, Pires SM, Ong KL, Hartnett E, Naugle A, Harman J, Bennett P, Cieslak P, Scallan E, Rose B, Holt KG, Kissler B, Mbandi E, Roodsari R, Angulo FJ, Cole D. Application of Bayesian techniques to model the burden of human salmonellosis attributable to U.S. food commodities at the point of processing: adaptation of a Danish model. Foodborne Pathog Dis. 2011 Apr;8(4):509-16. doi: 10.1089/fpd.2010.0714. Epub 2011 Jan 16.

 Top of Page