Shrimp at the Reception

The overall attack rate among all 180 guests is 68/180 = 37.8%. Among the 160 guests who returned questionnaires, it is 68/160 = 42.5%. For shrimp, the attack rate among exposed guests is 52/95 = 54.7%; among unexposed guests it is 16/65 = 24.6%. The risk ratio is 54.7% / 24.6% = 2.22, and the risk difference is 30.1 percentage points. That is a strong association: shrimp eaters were about twice as likely to become ill. For potato salad, the attack rate is 36/80 = 45.0% among eaters and 32/80 = 40.0% among non-eaters, giving a risk ratio of 1.13, which is much weaker. The shrimp result is consistent with a foodborne vehicle, but the overall attack rate and the shrimp risk ratio should be interpreted alongside the questionnaire response and recall problems. The health department should also check whether the 20 non-responders differ from responders, because if cases were more likely to respond, the attack rate could be inflated.

A retrospective cohort is preferable because the population is defined and enumerable: 180 guests attended, and the outcome occurred after a known exposure period. The cohort design estimates incidence directly and allows calculation of attack rates and risk ratios for each food item. A case-control study would be less efficient here because the outcome is common, so the odds ratio would not approximate the risk ratio well. A case-control design would also require selecting controls from the same source population, which is possible but less direct. Poor recall among non-cases is a serious information bias. If ill guests remember the shrimp better because they were interviewed sooner or because illness sharpened their memory, the shrimp risk ratio will be inflated. The fact that many non-cases cannot recall what they ate means exposure misclassification is likely differential. The health department should use a standardized food-frequency questionnaire, show photographs of the buffet items, interview all guests with the same instrument and timing, and cross-check with serving records, seating charts, and kitchen invoices. Even with recall bias, a risk ratio of 2.22 with a dose-response pattern or a positive food-handler specimen would strengthen the conclusion.

The health department should first verify the case definition and interview all 180 guests, not just the 160 responders, using a standardized questionnaire that asks about timing, symptoms, and every buffet item. It should look for dose-response: did guests who ate more shrimp have higher attack rates? It should compare attack rates by table or serving line to see whether one batch or one server was linked to illness. Laboratory steps include stool cultures or molecular testing from ill guests, testing the symptomatic food handler, and testing retained food samples if available. Environmental steps include inspecting the kitchen for temperature abuse, cross-contamination, handwashing, and ill-food-handler policies. The health department should trace the shrimp to its supplier and check whether other events using the same lot had Illnesses. If the handler's stool grows the same pathogen as the guests, and the shrimp was handled without gloves or after a lapse in hand hygiene, the hypothesis is strengthened. Closure of the kitchen may be justified if there is ongoing risk, such as a symptomatic handler still working or unsafe food handling, but closure should be based on immediate hazard assessment, not on the epidemiologic association alone.