Food Source Prediction of Shiga Toxin–Producing Escherichia coli Outbreaks Using Demographic and Outbreak Characteristics, United States, 1998–2014

BACKGROUND: Foodborne illness is a continuing public health problem in the United States. Although outbreak-associated illnesses represent a fraction of all foodborne illnesses, foodborne outbreak investigations provide critical information on the pathogens, foods, and food-pathogen pairs causing illness. Therefore, identification of a food source in an outbreak investigation is key to impacting food safety. OBJECTIVE: The objective of this study was to systematically identify outbreak-associated case demographic and outbreak characteristics that are predictive of food sources using Shiga toxin-producing Escherichia coli (STEC) outbreaks reported to Centers for Disease Control and Prevention (CDC) from 1998 to 2014 with a single ingredient identified. MATERIALS AND METHODS: Differences between STEC food sources by all candidate predictors were assessed univariately. Multinomial logistic regression was used to build a prediction model, which was internally validated using a split-sample approach. RESULTS: There were 206 single-ingredient STEC outbreaks reported to CDC, including 125 (61%) beef outbreaks, 30 (14%) dairy outbreaks, and 51 (25%) vegetable outbreaks. The model differentiated food sources, with an overall sensitivity of 80% in the derivation set and 61% in the validation set. CONCLUSIONS: This study demonstrates the feasibility for a tool for public health professionals to rule out food sources during hypothesis generation in foodborne outbreak investigation and to improve efficiency while complementing existing methods.