The estimated number of illnesses from shell egg–associated S.
Enteritidis in the United States for the year 2000 (IllSE
) was calculated as:
(equation 1), where
IllSE = number of S. Enteritidis illnesses from eating shell eggs in 2000.
F1 = number of culture-confirmed salmonellosis cases ascertained by FoodNet in 2000 = 4,330.
F2 = the proportion of culture-confirmed salmonellosis cases ascertained by FoodNet for which isolates were serotyped as S. Enteritidis. From the 4,330 culture-confirmed salmonellosis cases ascertained for 2000, 3,964 Salmonella isolates were serotyped, 585 of which were identified as S. Enteritidis (585/3,964 = 0.148).
F3 = the proportion of S. Enteritidis cases from eating shell eggs. In 2000, FoodNet ascertained 15 S. Enteritidis outbreaks in which food vehicles were identified: 12 were egg-associated (12/15 = 0.8). This proportion was used as a surrogate for the proportion of sporadic S. Enteritidis illnesses from eating shell eggs.
F4 = a multiplier to account for cases of salmonellosis that occurred in the FoodNet catchment area but were not confirmed by fecal culture, and subsequently, not ascertained by FoodNet. The value used for this multiplier was 38.6 (5
F5 = a multiplier to extrapolate from the FoodNet catchment area to the U.S. population. For 2000, the population in the 8 FoodNet catchment sites was 30,500,000 persons, thus representing 10.8% of the U.S. population at that time (6
). The multiplier was computed by taking the inverse of the proportion of the U.S. population represented by the catchment area (1/0.108 = 9.2).
Thus, based on equation 1 above, the IllSE point estimate was calculated as: 4,330 x (585/3,964) x (12/15) x 38.6 x (281,400,000/30,500,000) = 182,060.
Uncertainty for the estimate of IllSE was also determined. As illustrated in equation 1, multipliers F2, F3, F4, and F5 adjusted the number of culture-confirmed salmonella illnesses ascertained by FoodNet in 2000 (F1) to estimate the number of S. Enteritidis illnesses due to eating shell eggs. Uncertainty associated with each multiplier contributes to the overall uncertainty associated with the estimate of IllSE. The distributions described below were incorporated into a Monte Carlo simulation (@RISK, version 4.0, Palisade Corp., Newfield, NY) of 100,000 iterations to estimate the range of potential values for IllSE ().
Figure Estimated number of illnesses from Salmonella Enteritidis in shell eggs, United States, 2000. The point estimate of 182,060 illnesses is indicated by the filled box and solid vertical line. The open diamonds and attached line indicate the range of estimate (more ...)
F2 assumed the proportion of Salmonella illnesses attributable to S. Enteritidis ascertained by FoodNet was equal to the proportion of Salmonella illnesses attributable to S. Enteritidis throughout the United States. The β distribution (585 + 1, 3,964 – 585 + 1) was used to describe uncertainty around the F2 point value.
F3 assumed that the proportion of S. Enteritidis outbreaks and sporadic infections attributable to eating shell eggs was equivalent. The β distribution, (12 + 1, 15 – 12 + 1), was used to model the uncertainty around the proportion of S. Enteritidis cases assumed to have resulted from shell egg consumption.
F4 assumed that the impact of diarrheal illness, and the behavior of persons with diarrhea and their healthcare providers, was the same in the FoodNet catchment area as in the U.S. population. It also assumed that the proportions of case-patients who 1) sought medical attention, 2) provided a specimen for fecal culture, and 3) were confirmed as salmonellosis patients contributed equally to case ascertainment, but that these proportions differed for patients who experienced bloody diarrhea compared to those who experienced nonbloody diarrhea. A triangular distribution with a minimum value of 9.8 and a maximum value of 67.7 around the point estimate of 38.6 was specified to quantify uncertainty associated with F4.
F5 assumed that the population of the FoodNet catchment area in 2000 was representative of the U.S. population. Because this assumption was qualitative, uncertainty associated with the multiplier could not be modeled.