Application of grocery purchase data to understanding patterns of Salmonella enteritidis illnesses in the United States, 2004-2006

Introduction: Salmonellosis, caused by infection with Salmonella species, is a significant public health problem, both in the disease burden and economic costs. Over 1 million illnesses every year are due to Salmonella infections. Salmonella enteritidis accounts for approximately 20% of reported illnesses. It has been estimated that approximately 90% of illnesses are acquired from exposure to contaminated foods in the home. This is a study to determine the association between household-reported grocery purchases and Salmonella enteritidis. Methods: This is a retrospective, ecological, cross-sectional study that analyzed Homescan market as unique geographic areas. Food data comes from Homescan, and illness data comes from National Salmonella Surveillance System (NSSS). Food data was standardized and a per-capita annual Purchase-Weight was calculated for each Food Category. A risk score for each Homescan market was calculated which identifies the average relative risk of foods reported by households in the Homescan market. A negative binomial model was applied to the Homescan market risk score as the independent variable and the incidence rate of Salmonella enteritidis illnesses as the dependent variable with the size of each Homescan market population included as an offset. Results: From 2004 to 2006 there were 12,589 cases of Salmonella enteritidis reported in the United States. The Homescan market incidence rate varied from <1 to 7 cases per 100,000 persons. In the same time period, 21,124 households reported 19,152,019 food observations which were grouped into 1 of 62 Food Categories. The population-weighted risk score varied between Homescan markets from 1.47 (San Diego) to 2.36 (Birmingham). There was no association between rates of salmonellosis and relative high-risk food exposure in Homescan market areas. Discussion: This is the first attempt to utilize grocery purchases as a proxy for food exposure and sporadic salmonellosis. Differences in relative amounts of high-risk food exposure were found at a population-level, demonstrating variation in food exposure throughout the United States. While there was no association with rates of salmonellosis, this may have been because there was not sufficient heterogeneity between the geographic units due to their size. As a result, a replication of this approach among smaller geographic units should be considered.