We defined unspecified agents as agents that cause acute gastroenteritis but that were not included in our estimate of foodborne illness caused by 31 major known pathogens (1). They include known agents with insufficient data for estimating agent-specific episodes of illness; known agents not yet recognized as causing foodborne illness; microbes, chemicals, or other substances known to be in food but for which pathogenicity is unproven; and agents not yet described. To estimate the extent of foodborne illness caused by unspecified agents, we estimated the number of acute gastroenteritis illnesses, hospitalizations, and deaths and subtracted the estimated number of acute gastroenteritis illnesses, hospitalizations, and deaths caused by 24 major known pathogens that typically or often cause diarrhea or vomiting (Figure 1). We refer to them as the 24 known gastroenteritis pathogens, although for a few, diarrhea or vomiting was not the main clinical sign. Estimates of illness were not made for unspecified agents that do not typically result in acute gastroenteritis.

We used data from the 24 known gastroenteritis pathogens to estimate the proportion of unspecified agents that were acquired in the United States (hereafter referred to as domestically acquired) and transmitted in food. Most of our data were from 2000 through 2007, and all estimates were based on the US population in 2006 (299 million persons) (9). To account for uncertainty, we used probability distributions to describe a range of plausible values for all model inputs. The modeling approach used and parameters of these probability distributions are detailed in the Technical Appendix. Our model outputs are in the form of probability distributions summarized by a mean point estimate with 90% credible intervals (CrIs).

Unspecified agents are major contributors to the total number of episodes of acute gastroenteritis and foodborne diseases. If distribution of domestically acquired and foodborne agents is similar to that of the 24 known gastroenteritis pathogens (1), then these agents cause 38.4 million episodes of foodborne gastroenteritis each year in the United States, resulting in 78,878 hospitalizations, and 1,686 deaths. Combining the estimates for unspecified agents and major known pathogens provides an estimate of the total effect of contaminated food consumed in the United States: 47.8 million episodes of illness, 127,839 hospitalizations, and 3,037 deaths (Table).

Our estimate of foodborne illness caused by unspecified agents is lower than that estimated by CDC in 1999 (38.4 million vs. 62 million, respectively) (19). A major reason for this decrease is our lower estimate of episodes of acute gastroenteritis, which probably resulted from changes in data sources and methods rather than a real decline in the rate of illness. Our estimate is derived from the 3 most recent FoodNet Population Surveys, which had a sample size 5× greater than that in the 1996–1997 FoodNet survey used for the 1999 estimates. Additionally, the 1999 estimates relied on respiratory symptom and vomiting data from US studies conducted before 1980 (20,21). The current and the 1999 estimates excluded persons reporting concurrent cough or sore throat, but the proportion of respondents reporting these signs was higher in the current than in the earlier surveys (38% vs. 25%), contributing to a lower estimated prevalence of acute gastroenteritis (0.60 vs. 0.79 episodes/person/year). In addition, the current study excluded persons with vomiting who had been ill for <1 day or whose illness did not result in restricted daily activities, whereas the 1999 estimate included all persons with vomiting. All these factors contributed to the current estimate of acute gastroenteritis being 24% lower than the 1999 estimate.

The proportion of illnesses estimated to be foodborne was also a major driver of the current lower estimate of illness caused by unspecified foodborne agents. Because no data existed with which to directly estimate the proportions of unspecified agents that were domestically acquired and foodborne, distributions of these proportions were estimated to be similar to those of the 24 known gastroenteritis pathogens (1). Because norovirus accounts for 59% of illnesses caused by the 24 known gastroenteritis pathogens, the foodborne proportion was driven largely by norovirus. The proportion of foodborne norovirus used for the current estimate is 26%, a marked decrease from 40% used for the 1999 estimates. Additionally, unlike the 1999 estimates, the current estimates exclude international travel-related illnesses. As a result of these newer data and revised methods, the mean proportion of unspecified agents that were estimated to be transmitted by food was 25%, which is lower than 36% used for the 1999 estimate.

Estimating the number of hospitalizations and deaths caused by unspecified foodborne agents is challenging. For an illness caused by a pathogen to be recorded, a physician must order the appropriate diagnostic tests and the pathogen must be detected. Without identification of a pathogen, infections producing signs and symptoms of gastroenteritis may be coded as nonspecific signs or symptoms or as noninfectious illnesses (22). Our overall estimates of hospitalizations and deaths from acute gastroenteritis, derived from national data sources, include codes for infectious and noninfectious gastroenteritis. To avoid overestimating the number of hospitalizations, we selected outpatient visits resulting in hospitalization and we selected hospital discharge records on the basis of only the first 3 listed diagnoses. This approach was a compromise between limiting the analysis to hospitalizations for which acute gastroenteritis was listed as the primary cause and including hospitalizations for which signs or symptoms of gastroenteritis may have been a manifestation of another illness. This approach has been taken in other studies (23,24). For deaths, we included all records in which acute gastroenteritis was listed as an underlying or a contributing cause.

Our approach to estimating illness caused by unspecified agents has many limitations. First, the accuracy of our estimate of the number of acute episodes of gastroenteritis from the FoodNet Population Surveys has not been validated. This estimate was based on responses to questions about diarrhea and vomiting in the past month. These data provide a measure of prevalent cases; however, we lack sufficient data on duration of illness, onset date, and multiple episodes in the past month necessary to estimate incidence. An analysis of the FoodNet Population Surveys reported a 2-day median duration of acute gastroenteritis (10), suggesting that the increase in the estimate of illnesses based on incidence versus prevalence would probably be small and would be included within the range of sampling variability and uncertainty associated with our estimate of acute gastroenteritis. The accuracy of the 1-month recall period for acute gastroenteritis is also unknown. Some evidence suggests that shorter recall periods (e.g., past week) may result in higher reported prevalence (25). Which recall period is more accurate is not known. Our estimate of the number of episodes of acute gastroenteritis may be too high. Although our survey attempted to eliminate other causes of illness by asking about chronic diseases, some of the vomiting illnesses classified as acute gastroenteritis, for example, could have been caused by medications, alcohol withdrawal, or other causes, and some of the diarrheal illnesses could be caused by medications or other causes. However, our criteria for acute gastroenteritis were fairly strict, and foodborne illnesses probably occurred in some persons who were excluded because they had concurrent cough or sore throat or because their illness lasted for only 1 day. Second, we may have underestimated the episodes of illness caused by the 24 known gastroenteritis pathogens. When our method is used, any increase in the estimate for the major known pathogens will result in a decrease in the estimate for unspecified agents. Recent serologic data from European countries suggest that infection with Salmonella spp. is more common than estimated by other methods, including ours; however, many of these infections may be asymptomatic (26). Finally, the proportion of illnesses transmitted by food for unspecified agents is unknown and may differ from that for the 24 known gastroenteritis pathogens. Studies estimating foodborne disease in England and Wales and in Australia (which also attributed a large proportion of foodborne illness to unspecified agents: 73% in Australia and 48% in England and Wales vs. 80% in the United States) have estimated a similar proportion of acute gastroenteritis episodes to be transmitted by food (32%, and 26%, respectively vs. 25% in the United States) (27,28). A study of illness caused by known chemical agents, with estimates of the proportion that include acute gastroenteritis and that are foodborne, could help improve these estimates.

Combining the estimate for unspecified agents with that for the 31 major known pathogens to arrive at an estimate of overall foodborne illness has limitations. The method used for unspecified agents began with an estimate of acute gastroenteritis episodes, hospitalizations, and deaths and scaled down to a number for domestically acquired foodborne illnesses, hospitalizations, and deaths. Conversely, for most known pathogens, our estimate scaled counts of laboratory-confirmed illnesses up to an estimated number of ill persons, accounting for underreporting and underdiagnosis factors that contribute to an illness not being reported to public health agencies. Combining different approaches is not optimal because the methods themselves may affect the estimates derived. Also, our estimates do not include unspecified foodborne illnesses that do not typically cause signs of acute gastroenteritis. Most foodborne outbreak–associated illnesses caused by chemical agents reported to CDC during 2001–2006 (29) were not due to agents characterized by acute gastroenteritis and so would not be included in our estimates.

Although the number of episodes of foodborne disease caused by unspecified agents is substantial, the claim that 80% of foodborne illnesses are unspecified must be treated with caution. Illnesses caused by the 24 known gastroenteritis pathogens were, in most instances, estimated by using models that scaled counts of laboratory-confirmed illnesses up to an estimated number of illnesses with aggregate multipliers to adjust for underreporting and underdiagnosis factors that contribute to an illness not being reported to public health agencies (1). These multipliers are sensitive to the methods and modeling approaches used, and different choices could have increased estimates for the 24 known gastroenteritis pathogens, thus decreasing the estimate of foodborne illness caused by unspecified agents. For example, we took a conservative approach to estimating the underreporting multiplier for pathogens for which illness counts were derived from outbreak data (1); a less conservative approach would have increased estimated illnesses for these pathogens.

Future estimates might be improved by validating them by using other data on acute gastroenteritis episodes, hospitalizations, and deaths, such as by reviewing acute gastroenteritis coded as a secondary discharge diagnosis or assessing the accuracy of acute gastroenteritis coding on death certificates. The diagnostic gap might be narrowed by identifying additional agents linked to foodborne transmission. Systematic laboratory investigation of specimens from well-investigated outbreaks of foodborne disease of undetermined cause, and detailed investigations of specific syndromes, may identify new agents (4,5,30).