Parametric distributions of underdiagnosis parameters used to estimate annual burden of illness for five foodborne pathogens.

Abstract

Estimates of the burden of bacterial foodborne illness are used in applications ranging from determining economic losses due to a particular pathogenic organism to improving our understanding of the effects of antimicrobial resistance or changes in pathogen serotype. Estimates of the total number of illnesses can be derived by multiplying the number of observed illnesses, as reported by a specific active surveillance system, by an underdiagnosis factor that describes the relationship between observed and unobserved cases. The underdiagnosis factor can be a fixed value, but recent research efforts have focused on characterizing the inherent uncertainty in the surveillance system with a computer simulation. Although the inclusion of uncertainty is beneficial, re-creating the simulation results for every application can be burdensome. An alternative approach is to describe the underdiagnosis factor and its uncertainty with a parametric distribution. The use of such a distribution simplifies analyses by providing a closed-form definition of the underdiagnosis factor and allows this factor to be easily incorporated into Bayesian models. In this article, we propose and estimate parametric distributions for the underdiagnosis multipliers developed for the FoodNet surveillance systems in the United States. Distributions are provided for the five foodborne pathogens deemed most relevant to meat and poultry.