Addressing Limitations in Foodborne Outbreak Investigation: Recall Bias and the Feasibility of New Surveillance Strategies
Seitzinger, Patrick J 1992-
Accurate data on the incidence of foodborne illness and food histories for affected individuals represent two important barriers to enteric outbreak surveillance and response. Innovative tools to collect and analyze this type of public health intelligence will play an important role in research efforts to improve understanding of the extent, impact of and risk factors for foodborne disease in Canada and around the world. Ethica, a smartphone based application used to acquire, store, and analyze data on human behaviour, provided an opportunity to gather information on the occurrence of enteric illness and the food consumption behaviour of 96 university students over a 10-week period. Nausea or vomiting were reported by 34% of participants, and 29% reported diarrhea at least once during the study using at least one of the available reporting options, but only 7% reported they sought medical care. Real-time data collected through digital images, meal descriptions, and microsurveys were used as a reference to measure the sensitivity and specificity of traditional food history questionnaires administered through an email link after 7 or 18 days (2.5 weeks). The validity of food history data collected after 7 days was found to be consequentially low with sensitivities ranging from 14.3% for sprouts to 100% for leafy greens and specificities ranging from 30.4% for beef to 80.4% for peanuts. Similarly, the sensitivities of questions administered after 18 days ranged from 15.8% for sprouts to 77.8% for tomatoes, with specificities ranging from 21.2% for leafy greens to 92.1% for melons. The impact of recall bias on the accuracy of food history data was found to vary with food type. Bayesian latent class analysis was conducted to determine the sensitivities and specificities in the absence of a true gold standard – the results support those of frequentist approach. These findings serve as a first step in measuring the occurrence of self-reported foodborne illness and the implications of recall bias on outbreak investigations so that these biases can be accounted for research and public health practice.
DegreeMaster of Public Health (M.P.H.)
DepartmentSchool of Public Health
CommitteeMartin, Wanda; Tataryn, Joanne; Bharadwaj, Lalita; Schwandt, Michael
Copyright DateSeptember 2017
Field Epidemiology, Foodborne illness, outbreak, recall bias, surveillance, investigation strategies, smartphone, enteric illness, memory bias