M3-E
Modeling the attribution of foodborne illness in the United StatesMaryland E 3:30 PM
|
| Chair(s): Carl Schroeder, Elaine Scallan carl.schroeder@fsis.usda.gov
|
| Foodborne attribution models partition the human health burden of foodborne illness to specific food commodities. Model results assist risk managers and policy makers in formulating public health goals, prioritizing interventions, and documenting the effectiveness of prevention efforts for reducing foodborne illness. In the United States, the approach to food attribution has been multi-faceted, involving public health, regulatory, and academic organizations. As foodborne illnesses may be attributed to many foods at any point from consumption to the pathogen reservoir, there are several different yet complementary attribution models currently under development. This session provides an overview of and results from different types of foodborne attribution studies to survey the landscape of methods available for attributing illness to food and discusses the strengths and limitations of the different approaches. |
|
M3-E.1 3:30 PM Adaptation of a model for attributing human Salmonellosis to consumption of meat, poultry, and eggs in the United States. Guo C*; Federal Government chuanfa.guo@fsis.usda.gov
Abstract: Researchers with the Foodborne Diseases Active Surveillance Network (FoodNet) have developed a project to quantify attribution of meat, poultry, and eggs as sources of human salmonellosis in the US The project builds on an approach developed by the Danish Veterinary Institute, in which the prevalence and characteristics of Salmonella isolated from selected foods-of-animal-origin is compared with the prevalence and characteristics of Salmonella isolated from humans. The model described here, uses human surveillance data from the US CDC as well as data from the USDA for Salmonella isolated in meat, poultry, and eggs. This presentation will include an overview of the model structure, review of model results, and discussion of key assumptions. The output of this attribution model will likely reflect attribution of salmonellosis to the original source of food animal reservoirs of Salmonella. As such, model results will enable regulators to more effectively direct resources towards those food commodities implicated in foodborne salmonellosis.
|
M3-E.2 3:50 PM Estimating attribution of illnesses to food vehicle from reports of foodborne outbreak investigations. Painter JA*; US CDC jpainter@cdc.gov
Abstract: For most foodborne pathogens, outbreaks provide robust information on the sources of infections. To the extent that outbreaks represent the sources of all foodborne illness, analysis of outbreaks is likely to provide a comprehensive estimate of the proportion of all foodborne illness that can be attributed to specific foods. Based on data reported by state and local health Departments, the US CDC's Outbreak Response and Surveillance Unit is currently finalizing a model that estimates the percentage of foodborne illness attributable to various foods. Foods implicated in outbreaks were categorized into major food commodities that are meaningful for regulatory agencies, industry, and consumers. A model was developed to allocate the proportion of illnesses due to various food vehicles for each etiologic agent. This information can be applied to estimates of the number of persons with illness due to each pathogen. Thus, the estimated number of illnesses due to each food vehicle can be determined. Information derived from foodborne outbreak investigations incorporates illness resulting from contamination at any point from farm to fork; the analysis attributes the illness to the food consumed. The percentage of foodborne illnesses, hospitalizations, and deaths attributed to major food categories will assist public health decision-makers in setting food-safety priorities.
|
M3-E.3 4:10 PM A method for point-of-exposure attribution of diarrheal illness. Maldonado G, Ayers TL*; University of Minnesota, US CDC GMPhD@umn.edu
Abstract: Of the many cases of diarrheal illness that occur in the US each year, how many are caused by the food we eat, the water we are exposed to, or the animals we come into contact with? An accurate answer to this question would enable us to intervene to minimize the incidence of diarrheal disease. Information we could use to answer this question, however, is not collected in a uniform and representative manner. Diarrheal illness that is not associated with a foodborne outbreak ("sporadic" illness) is typically studied using a case-control design, while illness that is associated with a foodborne outbreak ("outbreak" illness) is typically studied using an outbreak-investigation design. Moreover, case-control studies typically study a relatively small, nonrandom subset of the US population, and not all foodborne outbreaks are reported and investigated. Can we blend together data from case-control studies and outbreak investigations in a way that overcomes these challenges to get an accurate answer to our question above? Here we describe such a method.
|
M3-E.4 4:30 PM Eliciting information on uncertainty from heterogeneous expert panels: Attributing US foodborne pathogen illness to food consumption. Hoffmann S*, Fischbeck P, Krupnick A, McWilliams M; Resources for the Future hoffmann@rff.org
Abstract: US food safety policy is organized around particular pathogens on specific foods, for example Salmonella in poultry. Food safety policy makers need a picture of how foodborne illnesses caused by particular pathogens are associated with exposure from consumption of different foods in order to prioritize risk management activities. Two fundamental approaches are being used to understand the relationship between fooborne pathogens and health outcomes: risk assessment and attribution of foodborne illnesses to causal sources. Each of these approaches confronts severe data constraints. Expert elicitation provides an additional measure. It provides formal methods of capturing experts' knowledge of microbial ecology, food handling practices, food consumption, and other factors that influence the likelihood that illness caused by a particular pathogen is associated with a particular food. This study presents the results of an expert elicitation of 44 US food safety experts on the association between food consumption foodborne illness associated with 11 major foodborne pathogens. The study elicited individual experts' triangular distributions of the proportion of foodborne illnesses associated with a particular pathogen attributable to one of 11 categories of food. The results of the study are used to attribute estimates of foodborne illness incidence by pathogen from Mead et al. and FoodNet to food consumption. Two measures of uncertainty are developed to provide estimates of the degree of uncertainty associated with each attribution estimate. Expert elicitation attribution estimates are compared to food attribution estimates based on outbreak data. The study concludes with a discussion of how multiple food attribution measures can be used both to provide uncertainty bounds on attribution estimates and to help identify where further investigation is most needed.
|
M3-E.5 4:50 PM Ranking foodborne risks under uncertainty: Comparing outbreak and expert attribution of illnesses to foods. Batz MB*, Morris JG, Taylor MR, Krupnick AJ, Hoffmann SA; Resources for the Future, University of Maryland School of Medicine misterbatz@gmail.com
Abstract: For the prioritization of public and private resources to prevent microbial foodborne illness, it is critical to understand the relative impact of different pathogens and their associated food vectors. We have developed the Foodborne Illness Risk Ranking Model (FIRRM), a quantitative, empirical decision tool to rank pathogen-food combinations by five measures of public health impact. We estimate annual illnesses, hospitalizations, and fatalities due to 28 pathogens, based on recent surveillance data, and attribute these illnesses to a set of food categories. For major pathogens, mortality and morbidity health states associated with illness are valued in dollars and in QALYs. We use Monte Carlo simulation to model parameter uncertainty through random input distributions, and use scenario analysis to test model assumptions. Within FIRRM, we apply food attribution by percentage: total illnesses due to a pathogen are portioned to a set of food categories. We compare two sources of attribution data - foodborne outbreaks and expert judgment - using the same dozen food categories organized by commodity. We analyze CDC data on foodborne outbreaks from 1990-2004 by binning food vehicles into food categories under various decision rules. For each pathogen, we compute the portion of outbreak cases due to each food. Expert attribution is based on an elicitation of 44 experts for attribution percentages for eleven major pathogens, including low and high estimates for each food category. We combine expert data, maintaining inter-expert and intra-expert uncertainty. Outbreak and expert attribution percentages are similar for some pathogens, but differ markedly for others. These similarities and differences speak to the advantages and disadvantages of each data set, and indicate for which pathogens improved attribution data is most needed. Rankings of food-pathogen combinations differ according to which attribution method is employed, but remain relatively robust.
|
M3-E.6 5:10 PM Discussion
|
