Foodborne outbreak investigation: Effect of recall inaccuracies on food histories

Abstract

Recall inaccuracies are a key limitation in a foodborne outbreak investigation. Misclassifications in self-reported exposure status reduce the power of epidemiological studies to detect meaningful associations between exposures and the development of illness. The purpose of this study was to investigate the effect of recall inaccuracies on the validity of food history data in a context comparable to outbreak investigations. The food consumption of 96 university students was collected using Ethica, a smartphone-based data acquisition system. Comprehensive food histories were captured for 10 days through a combination of digital images, meal descriptions, and short food exposure surveys. These real-time data were used as a reference to measure the sensitivity and specificity of food history questionnaires administered 7 or 18 days (2.5 weeks) after consumption (n1/486). The questionnaires and time intervals used in this study were designed to resemble a range of plausible local, provincial, and national enteric outbreak investigations conducted by public health officials in Canada. Comparably low accuracy of dietary memory after both time intervals suggests there is a substantial potential for bias for most food types following the first week after consumption. The magnitude of recall inaccuracies was not uniform across food types. This study serves as a first step in quantifying recall inaccuracies in a context comparable to how cases and controls might be questioned for outbreak investigations so that recall inaccuracies can be accounted for and mitigated in public health practice. HIGHLIGHTS • Comparably low accuracy of dietary recall is present after 7 and 18 days. • Experiencing gastrointestinal symptoms did not affect accuracy of dietary recall. • Repeatedly recalling food exposures leads to a higher rate of false positives. • Prevalence of exposure was associated with higher sensitivity and lower specificity. • Recall inaccuracies can be quantified and accounted for in public health practice.