The Graduation Dinner: Reading Food-Specific Attack Rates Before the Laboratory Speaks
A district hospital reports 60 patients with acute watery diarrhoea and vomiting over 36 hours. The first arrived 8 hours after a school graduation dinner attended by 200 people; stool cultures are pending. The public health team interviews all 200 attendees and takes food histories. Overall, 60 of 200 became ill (30%). Chicken salad: 120 ate it, 50 ill; 80 did not, 10 ill. Rice: 160 ate, 52 ill; 40 did not eat it, 8 ill. Baklava: 100 ate it, 30 ill; 100 did not, 30 ill. The chicken salad was prepared five hours before serving, held in a warm kitchen, and handled by a cook with a healing whitlow on her thumb. Two of the 60 cases are themselves food handlers. The school director wants the kitchen closed and the event publicly blamed today.
1. Calculate food-specific attack rates and risk ratios for the three foods. Which food is the most likely vehicle, and what are the limitations of this analysis at this stage?
Read the worked answer
Attack rates and risk ratios
| Food | Ill/exposed | AR | Ill/unexposed | AR | RR |
|---|---|---|---|---|---|
| Chicken salad | 50/120 | 41.7% | 10/80 | 12.5% | 3.33 |
| Rice | 52/160 | 32.5% | 8/40 | 20.0% | 1.63 |
| Baklava | 30/100 | 30.0% | 30/100 | 30.0% | 1.00 |
Interpretation. The chicken salad has both the highest attack rate among those exposed and the highest risk ratio; it is the leading candidate. Supporting features are biological plausibility (five hours in a warm kitchen allows toxin production or bacterial multiplication; a hand lesion allows staphylococcal contamination of a ready-to-eat food) and the epidemiological pattern: a single meal, all cases within about 36 hours, no evidence yet of person-to-person spread - a common-source, point-exposure outbreak with an overall attack rate of 30%.
Limitations before acting on it.
- Confounding by co-consumption: most rice eaters probably also ate chicken salad, which alone could produce the RR of 1.63 for rice. Stratify rice by chicken-salad status, or fit a stratified/adjusted estimate.
- Small numbers in the unexposed cells mean wide confidence intervals; a single case shifts the RR appreciably.
- The 12.5% attack rate among non-eaters of chicken salad needs explaining - cross-contamination of other items, recall error, or background gastroenteritis.
- Recall bias and non-differential misclassification of exposure, which would dilute rather than inflate the RR.
- A dose–response analysis (amount of chicken salad eaten) would strengthen or weaken the inference and should be run now.
2. Set out your case definition, immediate control measures, and the specimens and environmental investigations you would request. How does the incubation period inform the likely organism?
Read the worked answer
Case definition (provisional). A person who ate food prepared in the school kitchen on the dinner date AND developed three or more loose stools in 24 hours, or vomiting, with onset within 72 hours of the meal. Build a line list: name, age, food and quantity eaten, onset time, symptoms, stool results, contacts.
Incubation reasoning. The first presentation at 8 hours favours a preformed toxin - Staphylococcus aureus (typically 2–6 hours, vomiting predominant) or Bacillus cereus (1–6 hours emetic type). If the epidemic curve continues to rise at 24–72 hours, Salmonella becomes more likely, with a longer incubation and more fever and diarrhoea. Plot onset times as a histogram before committing to a hypothesis.
Immediate control. Close the kitchen pending investigation; exclude all symptomatic food handlers until 48 hours symptom-free and, for Salmonella or Shigella, until stool cultures are negative; reinforce hand hygiene and no bare-hand contact with ready-to-eat food; verify hot-holding above 60°C and cold-holding below 5°C; ensure a safe water supply at the school.
Specimens. Stool from at least 10 cases for culture and, where toxin is suspected, for staphylococcal enterotoxin or B. cereus; vomitus if available; swab of the cook's whitlow plus nasal swab for S. aureus typing; any retained food samples.
Environmental. Preparation timeline, temperature logs, water source, staff illness records, cleaning of utensils used for raw and cooked items. Notify the food safety authority and the national surveillance system.
3. The director asks whether the dinner 'caused' the outbreak and whether the school should be closed. Answer with your reasoning and your proposed communication.
Read the worked answer
On causation. The evidence is strong but not yet complete. Strength of association (RR 3.33), consistency across two analytic approaches, temporality (exposure preceded illness), biological plausibility, and the coherence of a single-meal, single-point epidemic curve all support the chicken salad. What is missing is identification of the agent and its source, and we should say so plainly. It is legitimate to state a strong working conclusion while flagging that laboratory results will confirm or modify it. Avoid naming the cook publicly: this is identifiable personal health information and would damage future cooperation with investigations.
On closing the school. Closure is not the appropriate control measure. There is no evidence of sustained person-to-person transmission, which is what school closure interrupts; the exposure was confined to a single meal and the kitchen is already closed. Transmission risk to pupils is now negligible. What is indicated instead is exclusion of symptomatic children and staff until 48 hours symptom-free, supervised handwashing, and cleaning of shared surfaces. Closing the school would be a visible but ineffective action, and it would impose real costs on families.
Communication. Give one clear, practical message: the school kitchen is closed while the investigation continues; anyone with vomiting or diarrhoea should seek care and stay away from school or work until 48 hours after recovery; results will be shared within days. Acknowledge uncertainty openly, correct the framing that a culprit must be named today, and nominate a single spokesperson so the message does not fragment.