IHP 330 Module 4 Assignment: sample paper, in real form

Investigating a Salmonella Enteritidis Outbreak After a Catered Banquet: A Retrospective Cohort Analysis of 248 Attendees

Student Name

Department of Health Professions, Southern New Hampshire University

IHP 330: Principles of Epidemiology

Instructor Name

Month Day, Year

Case Definition and Descriptive Epidemiology

On October 14, 2023, a county health department received three reports of acute gastroenteritis among people who had attended a catered awards banquet held on October 11 at a community conference center. The banquet had a fixed roster of 312 attendees who ate from a single buffet service between 6:30 p.m. and 8:00 p.m. A case was defined as three or more loose stools within any 24-hour period, with onset between 6 and 72 hours after the meal, in a person on the attendee roster, accompanied by at least one of fever, abdominal cramping, or vomiting. Applying that definition to 248 completed interviews identified 96 cases, an overall attack rate of 38.7 percent.

Onset times clustered tightly. The median incubation period was 26 hours with a range of 9 to 61 hours, and the epidemic curve rose to a single peak between 20 and 32 hours after the meal, then fell. A point-source pattern of that shape argues for one common exposure at a single sitting rather than person-to-person transmission, which would produce successive waves spaced roughly one incubation period apart. Cases were distributed across the seating chart rather than concentrated at adjacent tables, which weakens a contact explanation. Stool specimens were submitted by 14 ill attendees, and 11 yielded Salmonella enterica serotype Enteritidis with an indistinguishable whole genome sequencing profile, establishing the etiology of the cluster.

The denominator matters as much as the numerator here. Of the 312 people on the roster, 248 completed a structured telephone interview between October 19 and October 23, a response of 79.5 percent, and every attack rate below uses interviewed attendees rather than the full roster as its denominator. The 64 attendees who could not be reached did not differ meaningfully from respondents by age band or seating section, although that comparison rests on roster data alone. Reporting the interviewed cohort as the denominator is the conservative choice: it neither assumes that unreached attendees were well nor inflates the population at risk with people whose outcome is unknown.

Study Design and Choice of Measure

This investigation used a retrospective cohort design. The design followed from the setting rather than from preference: a complete roster defined a closed population, every member of that population shared the same exposure opportunity within a 90-minute service, and both exposure and outcome could be ascertained for each member after the fact. Those conditions make the population enumerable, which is the practical test for a cohort analysis. Had the meal been served to an anonymous walk-in crowd, no roster would exist, the population at risk would be unknowable, and a case-control design comparing reported exposures between cases and well controls drawn from the same source population would have been the only workable alternative (Centers for Disease Control and Prevention, 2012).

The measure reported is the risk ratio, computed from food-specific attack rates. Risk is the appropriate frequency measure because the cohort is closed, the risk period is a fixed 72 hours, and no one was lost to follow-up in a way that would require person-time in the denominator, so an incidence rate would add complexity without adding information. The risk ratio is preferred over the odds ratio for the reason it is always preferred when risks are directly calculable: it is interpretable as a multiple of risk. That preference is not cosmetic here, because with an overall attack rate of 38.7 percent the outcome is common, and an odds ratio would overstate the risk ratio noticeably (Rothman et al., 2021).

Of the 15 items served, chicken salad showed the strongest association. Among the 141 attendees who reported eating it, 82 became ill, an attack rate of 58.2 percent, compared with 14 of 107 among those who did not, an attack rate of 13.1 percent. The risk ratio is 4.45 with a 95 percent confidence interval of 2.68 to 7.39, and 82 of the 96 cases, or 85.4 percent, reported eating the item. The attributable fraction among the exposed is 77.5 percent, meaning that if the association is causal, roughly three quarters of the illness among those who ate chicken salad is attributable to it. Tossed green salad produced a crude risk ratio of 1.35 with an interval of 0.93 to 1.96.

Confounding, Bias, and Interpretation

The green salad result illustrates confounding rather than a second vehicle. Attendees who took chicken salad were substantially more likely to take green salad from the adjacent station, so the crude comparison mixes the two exposures. Stratifying on chicken salad consumption collapses the effect: among attendees who ate chicken salad the risk ratio for green salad is 1.05, among those who did not it is 1.12, and the Mantel-Haenszel estimate adjusted for chicken salad is 1.07 with a 95 percent confidence interval of 0.79 to 1.45. A crude estimate of 1.35 falling to 1.07 after adjustment, with stratum-specific estimates close to each other, is the signature of confounding without effect modification.

Three threats to validity remain, and stratification does not touch any of them. Interviews took place 8 to 12 days after the meal, and ill attendees who had already heard that chicken salad was suspected may have recalled eating it more readily than well attendees, which is differential recall bias and would inflate the risk ratio. Non-response is the second threat: 20.5 percent of the roster was never reached, and if illness made a person more willing to answer the telephone, the attack rate is overstated. Outcome misclassification is the third, because the case definition is syndromic, only 14 specimens were cultured, and unrelated diarrheal illness almost certainly entered the case count.

Taken together, the evidence supports chicken salad as the vehicle. The association is strong, the interval excludes the null by a wide margin, exposure preceded illness by a period consistent with the organism, 11 isolates share a sequencing profile, and the descriptive pattern is that of a point source (World Health Organization, 2008). What a single observational investigation cannot do is establish causation on its own, and no leftover food remained for testing, so the microbiological link between the item and the isolates rests on inference rather than culture. The defensible conclusion is that chicken salad is the probable vehicle, with residual uncertainty coming from the recall, non-response, and misclassification problems named above rather than left implicit.

References

Celentano, D. D., & Szklo, M. (2019). Gordis epidemiology (6th ed.). Elsevier.

Centers for Disease Control and Prevention. (2012). Principles of epidemiology in public health practice: An introduction to applied epidemiology and biostatistics (3rd ed.). U.S. Department of Health and Human Services. https://www.cdc.gov/csels/dsepd/ss1978/index.html

Centers for Disease Control and Prevention. (2023). Surveillance for foodborne disease outbreaks, United States: Annual report. U.S. Department of Health and Human Services. https://www.cdc.gov/foodsafety/fdoss

Council to Improve Foodborne Outbreak Response. (2020). Guidelines for foodborne disease outbreak response (3rd ed.). Council of State and Territorial Epidemiologists. https://cifor.us

Rothman, K. J., Lash, T. L., & VanderWeele, T. J. (2021). Modern epidemiology (4th ed.). Wolters Kluwer.

World Health Organization. (2008). Foodborne disease outbreaks: Guidelines for investigation and control. https://www.who.int/publications/i/item/9789241547222