Examining market and weather events as signals of an increased probability of Shiga toxin-producing Escherichia coli outbreaks linked to romaine lettuce

Abstract

Romaine lettuce has become a recurring source of foodborne illness outbreaks over the past decade. Most lettuce consumed in the United States originates from two primary sources: California and Arizona. Federal, state, and local authorities that investigate outbreaks linked to the consumption of romaine lettuce encounter many challenges during traceback and on-farm investigations that hinder their ability to determine the root cause of contamination. In this study, we investigate market signals (including price and shipment volumes) and weather parameters and their correlation with past outbreaks of Shiga Toxin-producing Escherichia coli illnesses linked to the consumption of romaine lettuce. We construct a daily panel of economic and weather data for eight outbreaks linked to romaine growing regions in California and Arizona from 2011 to 2019 and estimate the relative importance of these indicators on the probability of identifying each outbreak. Results suggest that prior increases in price and shipments are positively associated with outbreaks. In addition, some weather variables, most consistently daily maximum temperature, are positively associated with the probability of identifying an outbreak. Monitoring data for market and weather signals, coupled with a deeper understanding of regional supply shifts could help investigators quickly identify conditions that may signal an increased probability of outbreaks linked to romaine lettuce. Early detection can lead to improved education and outreach as well as targeted inspectional and sampling efforts. • Romaine lettuce has become a recurring source of foodborne illness outbreaks. • Increased prices and shipments are positively associated with outbreaks. • Market and weather signals may help identify conditions that could lead to outbreaks. • Lagged values seem more consistent predictors for market variables. • Defined cutoff values seem more consistent predictors for weather variables.