Purpose: This study’s objective is to assess the reliability of multiple sources of food attribution information – namely foodborne outbreak data, expert elicitation, and case-control studies – and to use these results to create integrated attribution estimates for 14 major pathogens.
Methods: This study builds on prior work evaluating outbreak attribution based on multiple reliability measures: outbreak density, ratio of estimated incidence to reported outbreak cases, sum of mean differences squared between outbreak and expert attribution, and the mean standard deviation across experts. We develop a novel approach to integrating attribution data from multiple sources that uses expert elicitation results to weight primary sources.
Results: Outbreak attribution for Campylobacter, Toxoplasma, Cryptosporidium, and Yersinia are shown to be unreliable based on multiple metrics, while estimates for E. coli O157:H7, Vibrio spp., and Cyclospora are the most reliable. The mean standard deviation in expert results is highest for Toxoplasma (2.02) and norovirus (1.77) suggesting the need for improved attribution for these pathogens. We present previously unpublished integrated attribution estimates for 14 pathogens over 12 food categories.
Significance: Outbreak data is shown to have variable reliability across pathogens as a source of attribution information, while expert elicitation is shown to be a powerful tool for evaluating attribution from multiple data sources when designed for such comparisons. Our findings suggest that creating combined estimates of attribution of illnesses to foods are possible and may be more reliable for public health policy than estimates based on a single data source.