Food-borne illness subject to publication bias

Publication bias creates a skewed picture of the true prevalence of restaurant-associated food-borne disease outbreaks.

“Ten percent of the people who came in to see their doctor for diarrhea got antibiotics. About 4% of them are cultured. And of course it’s only a minuscule proportion of people who get cultured [in whom we] ever find a pathogen. So other are a ton of people out there getting unnecessary antibiotics.”

At the jointly held annual Interscience Conference on Antimicrobial Agents and Chemotherapy and the annual meeting of the Infectious Diseases Society of America. Presented data on a search of restaurant-associated outbreaks appearing in the medical literature from 1973 to the present. An “outbreak” was defined as a minimum of two cases of similar illness and common exposure.

Then compared the literature search results with outbreaks reported from 1998 through 2006 via the Center for Disease Control and Prevention’s Electronic Foodborne Reporting System, or EFORS, State health departments are required to report all food-borne outbreaks to the CDC using this system.

A total of 133 published restaurants-associated outbreaks were found, compared with 7,300 unpublished reports in the EFORS.

Some stark differences emerged between the groups of published and unpublished reports. Of the 133 published reports, 121, or 91%, had an etiology identified vs. only 55% of unpublished reports. Also, the median number of persons reported ill in published studies was 72 vs. 5 in unpublished reports.

Botulism outbreaks accounted for 5% of published outbreaks, compared with 0.05% of all of those reported. Hepatitis A was identified as the etiology in 15% of published outbreaks vs. 3% of unpublished outbreaks with known etiology. And Norovirus, “far and away” the most common cause of outbreaks with a known etiology, was cited as the cause of an outbreak in 14% of published reports vs. 26% of the unpublished outbreaks.

Thus, publication bias can determine physicians’ ability to interpret the publicly available data on food-borne disease, he said. Novel pathogens or unusual circumstances surrounding the outbreak are obviously more likely to be published. But the fear of negative publicity may also inhibit publication, especially when an outbreak is related to an infected food worker.

Among published outbreaks, one-third implicated food workers as a cause, compared with only 11% in unpublished outbreaks. When restricting the analysis to outbreaks with an identified etiology, the results were similar, with food workers implicated twice as often in published vs. unpublished studies.

Each year in the United States, approximately 76 million food-borne illnesses result in 5,000 deaths. From 1998 to 2006, a mean of about 1,250 outbreaks were reported annually, of which two-thirds were associated with food-preparing restaurants or delicatessens.