FDOSS is a passive surveillance system that collects reports of foodborne disease outbreaks from federal, state, local, and territorial health departments in the United States. It is the primary source of data for outbreak-associated illnesses, hospitalizations, and deaths; etiologic agents; implicated foods; contributing factors; and preparation and consumption settings. Foodborne outbreaks are nationally notifiable and defined as ≥2 cases of a similar illness resulting from ingestion of the same food (8). When exposure occurs in 1 state, the outbreak is classified as a single-state outbreak; when exposure occurs in ≥2 states, the outbreak is classified as a multistate outbreak. Foods or specific ingredients are identified as sources by using epidemiologic, laboratory, traceback, and environmental assessment data. On the basis of this evidence, a sole food (e.g., apple) or a specific ingredient that is part of a complex food (e.g., beef in a sandwich) is reported as the source of an outbreak. A complex food is reported as the source when no specific ingredient is implicated. When an investigation does not identify a source, the food vehicle is reported as undetermined. CDC uses a hierarchical scheme to categorize reported foods (9,10). For simplicity and ease of interpretation, for this analysis we collapsed some categories (e.g., seeded and row crop vegetables are reported generically as vegetables).

Using FDOSS data from 1973–2016 (accessed December 11, 2017), we reviewed all 14,216 reported outbreaks with an implicated food or ingredient (henceforth, collectively referred to as food). We compared reported foods from outbreaks with a year of first illness onset during 2007–2016 with those during 1973–2006 by using a 3-stage process to identify novel outbreak-associated food vehicles. First, foods that were reported identically in both time frames were identified and removed by using SAS version 9.4 (SAS Institute Inc., https://www.sas.com). Second, 2 independent reviewers manually compared the remaining 878 food items and flagged any foods that seemed to be novel in 2007–2016. Third, a 5-member panel reviewed all foods initially identified as novel. For the second and third steps, we excluded foods if they were reported in a prior outbreak by using a different term (e.g., rice, wild vs. wild rice), a more general term (e.g., cheddar vs. white cheddar), or a different spelling or an abbreviation (e.g., brat vs. bratwurst) or if a specific contaminated ingredient(s) was not implicated for a newly reported complex food. We adopted this final exclusion criterion when reviewing complex foods in which no specific ingredient was implicated because it could not be determined what ingredients were included in the food itself, much less which ingredient was actually contaminated and could be novel (i.e., direct comparison between 1973–2006 and 2007–2016 was impossible).

We then conducted a secondary check of additional sources for all foods initially identified as novel (PubMed, online forums [e.g., Food Safety News, Food Poison Journal, and MarlerClark], and media reports). This check served to identify false-positive results from 2 scenarios: 1) the food had been implicated in an outbreak during 1973–2006, but the outbreak had not been reported to FDOSS; or 2) the food had been reported as part of an outbreak occurring during 1973–2006 with a more generic term. We reclassified foods only if the available information was sufficient to follow our criteria (i.e., there were >2 confirmed cases and an identified implicated food).

Novel food vehicles are presented along with key outbreak characteristics, including etiology and various measures of burden (e.g., case counts, deaths) in addition to statistical comparisons of these characteristics for outbreaks associated with food vehicles that are novel or not novel. Specifically, we used nonparametric Wilcoxon tests to assess difference in means and χ² tests to assess differences in percentages. Statistical analyses were completed in R version 3.3.3 (R Foundation for Statistical Computing, https://www.r-project.org).

We identified 28 novel foods linked to outbreaks that occurred during 2007–2016 in the United States. Compared with other outbreaks, those linked to novel foods were more likely to involve illnesses in multiple states; result in a food recall; and to be associated with, on average, larger numbers of cases, hospitalizations, and deaths. Investigating large and complex outbreaks requires considerable government resources and major costs to the public and industry. Moreover, two thirds of novel outbreak-associated food vehicles did not require cooking after purchase, and roughly half of novel foods did not require refrigeration. These factors highlight the need for targeted industry efforts to reduce contamination before point of purchase to protect consumers.

Food consumption patterns are dynamic, influenced by dietary trends, public health messaging, food accessibility, advertising, and affordability (6,7). Increased consumption of a food results in more opportunities for exposure and potentially larger outbreaks. Importing of foods and beverages into the United States has increased; average annual growth in economic value from 2007–2017 was 5.9% (12). As a result, access has expanded to a broader range of foods from diverse areas. One third of outbreaks linked to novel foods resulted from imported foods, whereas the overall percentage of outbreaks reported to CDC with an imported food implicated is relatively small (5% during the period 2009–2014) (13).

Identifying of novel outbreak-associated foods highlights the need for improvements in public health. The 3 key areas are outbreak investigation, prevention, and communications.

First, identifying the source of an outbreak can be difficult. Especially in multistate outbreaks, investigators typically use standardized food history questionnaires to identify common foods among a sample of patients. Investigators develop questionnaires largely on the basis of trends in previous outbreaks and are influenced by common—rather than novel—food vehicles. However, identifying novel foods may require a detailed investigation of everything consumed by patients during the exposure period. A targeted questionnaire can be developed on the basis of data from these in-depth, hypothesis-generating interviews and administered to a larger group of patients and controls. Barriers to this approach include limited resources, reluctance of investigators to allow investigators from other areas to interview patients in their jurisdiction, lack of training with regard to conducting hypothesis-generating interviews, and the assumption that more interviews with the standardized questionnaire will eventually reveal the underlying food vehicle(s). Efforts to further engage, fund, and support states in investigating and reporting outbreaks are needed to effectively identify novel food vehicles.

Second, identifying novel food vehicles provides opportunities for new prevention measures. The occurrence of >2 outbreaks linked to novel foods (we identified 7 instances) may serve as a warning signal for public health authorities indicating gaps in food safety practices, regulatory oversight, or both. Targeted outreach to industry may be pursued when a novel food is identified. Focused, collaborative prevention efforts undertaken by industry in collaboration with academic institutions, regulatory agencies, consumer advocacy groups, and other nonprofit public health organizations have been successful in the past and should become standard practice.

Third, communications regarding novel food vehicles to the public, industry, and regulatory agencies could be improved. Intensive public health messaging may be needed to notify the public, industry, and public health partners of newly discovered risks to prevent additional illnesses. This type of communication is regularly performed by public health and regulatory agencies, but greater emphasis on the novelty of the food source could attract additional media attention and, in turn, lead to greater awareness.

The first limitation of our analysis is that it encompasses reported outbreaks with data on implicated foods representing both a subset of foodborne outbreaks occurring in the United States and of outbreaks reported to FDOSS. As a result, these novel outbreak-associated food vehicles may not actually be novel; rather, previous outbreaks associated with these vehicles may not have been detected or reported. The second limitation is that only a small percentage of foodborne illnesses are linked to recognized outbreaks (outbreak-associated cases are estimated to range from 0.5% of laboratory-diagnosed Campylobacter cases to 19.0% of Shiga toxin–producing E. coli O157 cases) (2), and it is possible that outbreak-associated novel foods differ from those linked to sporadic foodborne illnesses. The third limitation is the possibility that novel food vehicles were misclassified. Considerable efforts were made to reduce false-positive identifications by searching for different spellings, alternative names, and shared ingredients, as well as by using a team of independent reviewers. We also checked secondary sources for foods identified as novel from FDOSS data. This effort identified 2 food vehicles (mamey fruit and jalapeño peppers) that incorrectly initially seemed to be novel because of the use of more generic food terms in reports before 2007.

In summary, identifying novel food vehicles for foodborne illness provides opportunities for early interventions that should not go unheeded. This analysis highlights the need for expanded food safety measures to reduce opportunities for contamination as foods are grown, harvested, and processed and to prevent illness from both novel and other food sources in the United States and abroad. Greater investment in public health outbreak investigation capacity may increase and expedite identification of novel food vehicles, which is vital because novel foods may serve as signals for emerging threats.