Masters in Hospital Management · Epidemiology for managers
Forty staff are sick by Tuesday evening. What do you do first?
Outbreak investigation, followed one step at a time through a single hospital outbreak
AIIMS Bhopal · 60 minutes
An outbreak is a management problem before it is a lab result
When staff or patients fall sick together, the manager feels it first.
Beds
Wards close or stop admitting. Surgery lists get cancelled.
Staff
Rosters collapse overnight. Sick staff must stay away from patients.
Patients
Vulnerable patients can be infected by staff, food or equipment.
Trust
Media, regulators and families ask one question: who knew, and when?
The epidemiologist finds the cause. The manager makes the response possible.
When does "a few sick people" become an outbreak?
Ask the room first. Then compare with this.
Outbreak
More cases than expected, in a given place, population and time.
Baseline
The usual number of cases. You cannot say "more" without knowing "usual".
Cluster
Cases grouped in place or time. It may or may not be above baseline.
Two cases of a disease you never see can be an outbreak. Twenty cases of the common cold may not be.
Tuesday, 9 am: twenty-nine staff have called in sick
29staff off sick with diarrhoea
7seen in Emergency overnight
2 to 3usual staff sick calls in a week
Nurses, residents, housekeeping, security and office staff. Nine different departments. Most say it started last night.
The nursing superintendent wants to know by 10 am whether to cancel today's surgery lists.
Illustrative figures, constructed for teaching.
Before anything else: is this outbreak real?
Twenty-nine against a usual two or three looks real. What else could make the numbers jump?
Check the diagnosis
Talk to the sick. See the clinical notes. Send stool samples today. Is it one illness or several?
Rule out a pseudo-outbreak
A new lab test, a new reporting rule, a festival leave pattern, or lab contamination can all create cases on paper.
Here: same symptoms, same night, far above baseline. Treat it as real and keep checking.
The steps are a checklist, not a queue
1 Prepare
Team, authority, lab contact
2 Confirm
Above baseline?
3 Verify
Diagnosis, samples
4 Define
Who counts as a case
5 Find cases
Line list
6 Describe
Time, place, person
7 Suspect
Form a hypothesis
8 Test
Compare ill and well
9 Confirm source
Lab, environment
10 Control
From day one
11 Watch
Keep counting
12 Tell
Report, communicate
Adapted from CDC, Principles of Epidemiology in Public Health Practice, 3rd ed., Lesson 6.
Who counts as a case? Write it before you count
A case definition has four parts. Try drafting ours.
Part
Our working definition
Person
Any member of hospital staff
Place
Working at this hospital
Time
Onset between Monday 12:00 and Wednesday 12:00
Clinical
Three or more loose stools in 24 hours
Add a lab tier later: confirmed (stool grows the organism), probable (meets the clinical definition, no lab result).
Do not put the suspected food in the definition. You cannot then test whether the food caused it.
Everything starts from a line list
One row per case, one column per fact. Here are the first 8 of 48 rows.
ID
Age
Sex
Job
Department
Onset
Stools
Fever
Admitted
Stool culture
C001
33
M
Resident doctor
Surgery
Mon 18:30
4
Yes
No
Not sent
C002
53
M
Housekeeping
NICU
Mon 20:15
8
Yes
Yes
Not sent
C003
45
F
Staff nurse
OT complex
Mon 20:45
4
Yes
No
No growth
C004
38
M
Lab technician
Laboratory
Mon 23:30
8
Yes
Yes
Salmonella
C005
54
F
Staff nurse
OT complex
Mon 23:30
6
No
No
Salmonella
C006
24
F
Resident doctor
Emergency
Tue 01:15
6
Yes
No
Salmonella
C007
51
M
Admin staff
Admin block
Tue 01:30
4
Yes
No
Not sent
C008
22
F
Admin staff
Admin block
Tue 01:45
7
Yes
No
No growth
Every table and graph in this investigation is built from these rows.
Illustrative figures, constructed for teaching.
What does this shape tell you?
Illustrative figures, constructed for teaching. 48 cases.
One sharp peak points to one shared exposure
Salmonella usually takes 12 to 36 hours to cause symptoms (range 6 to 72). Count back from the median onset: the exposure falls between Sunday evening and Monday evening. Monday lunch sits inside.
Incubation period: Heymann DL, ed. Control of Communicable Diseases Manual, 21st ed. APHA Press, 2022.
The wards are not the link. What the staff share is
Department
Cases
Surgery
10
Medicine
9
Admin block
7
Emergency
6
OT complex
5
Laboratory, Main gate
4 each
NICU, Paediatrics
3
Place: cases in nine departments, with no single ward standing out.
Person: all jobs, ages 22 to 56, day shift workers.
Night staff and people who bring lunch from home are almost untouched.
Illustrative figures, constructed for teaching.
Our suspect is Monday lunch. How do we test it?
A hypothesis is a testable guess about source and route. Ours: something served at Monday lunch in the staff canteen.
Cohort study, used here
The canteen's meal tokens list 240 people who ate. Interview them all about what they ate, ill or well. We reached 180.
Case-control study
When you cannot list everyone exposed, compare cases with similar people who stayed well.
Both designs do the same thing: compare the ill with the well.
Illustrative figures, constructed for teaching.
Attack rate: the share of exposed people who fell ill
Attack rate=people who ate the dish and fell illall people who ate the dish× 100
It is a proportion over the whole outbreak, not a rate per unit of time. The name is historical.
Chicken biryani=4296=43.8%
Compute it twice for every dish: once for those who ate it, once for those who did not.
Illustrative figures, constructed for teaching.
Which dish would you take off the menu?
Dish
Ate: ill / total
Attack rate
Did not eat: ill / total
Attack rate
Chicken biryani
42 / 96
43.8%
6 / 84
7.1%
Raita
38 / 110
34.5%
10 / 70
14.3%
Veg pulao
8 / 80
10.0%
40 / 100
40.0%
Dal
40 / 150
26.7%
8 / 30
26.7%
Salad
20 / 70
28.6%
28 / 110
25.5%
Gulab jamun
17 / 60
28.3%
31 / 120
25.8%
Look for a high attack rate among eaters and a low one among non-eaters.
Illustrative figures, constructed for teaching. 180 people interviewed, 48 cases.
Eating the biryani carried six times the risk
Relative risk=43.8%7.1%=6.1
Attack rate in eaters divided by attack rate in non-eaters. 1 means no difference.
Dish
Relative risk
Chicken biryani
6.1
suspect
Raita
2.4
check
Veg pulao
0.25
check
Dal
1.0
clear
Salad
1.1
clear
Gulab jamun
1.1
clear
Biryani: relative risk 6.1, 95% confidence interval 2.7 to 13.7. And 42 of the 48 cases ate it.
Illustrative figures, constructed for teaching.
Raita looks guilty too. Is it?
Most people who took raita also took biryani. Split the table by biryani and look again.
Group
Raita: ill / total
Attack rate
No raita: ill / total
Attack rate
Relative risk
Ate biryani
37 / 85
43.5%
5 / 11
45.5%
1.0
No biryani
1 / 25
4.0%
5 / 59
8.5%
0.5
Within each group, raita makes no difference. It rode along with the biryani.
This is confounding: a third factor linked to both the exposure and the illness creates a false association.
Illustrative figures, constructed for teaching.
Your turn: run the investigation on the full data
Lab 1: Case definition and epi curve
Loosen or tighten the case definition. Watch the case count, the curve, and the biryani relative risk change.
On your phone: drpakhare.github.io/mhm-outbreak-investigation
In pairs, five minutes. Find one case definition that makes the biryani look less guilty, and explain why.
Both labs use the same 180 interviews. Illustrative figures, constructed for teaching.
A strong association is not yet proof
The numbers point at the biryani. Three other lines of evidence must agree.
Laboratory
Salmonella in 19 of 23 stool samples. Same organism in the leftover biryani sample.
Kitchen
Chicken cooked on Sunday night, left at room temperature for five hours, reheated on Monday.
Food handlers
Stool samples from all kitchen staff. One cook had mild diarrhoea last week.
Epidemiology, lab and kitchen findings tell the same story. Now the conclusion is safe.
Illustrative figures, constructed for teaching.
You act on day one, not after the analysis
When
Action
Who decides
Tuesday, 10 am
Stop serving leftovers. Keep samples of every Monday dish.
Manager, canteen contractor
Tuesday, 10 am
Sick staff stay off patient care and food handling until 48 hours after symptoms stop.
Infection control, nursing
Tuesday, noon
Emergency staff roster. Postpone non-urgent surgery lists if staffing is short.
Manager, medical superintendent
Wednesday
Kitchen cleaned, cooking and holding temperatures logged, cook excluded until cleared.
Food safety officer
Waiting for proof while the source is still active is the most common and costly mistake.
Exclusion period as per the hospital infection control policy. Timeline illustrative.
Who needs to hear, and in what order?
Inside, first hours
Medical superintendent, Hospital Infection Control Committee, heads of affected departments.
Public health, same day
District Surveillance Unit under IDSP, the Integrated Disease Surveillance Programme.
Staff and patients
What happened, what to do if unwell, who to call. Plain and short.
Media
One spokesperson. Facts, actions taken, next update time.
Silence gets filled by rumour. Tell people early, tell them what you know and what you do not know yet.
Same steps, inside the NICU
A healthcare-associated infection is one that was not present or incubating when the patient was admitted.
Canteen outbreak
NICU cluster of bloodstream infections
Baseline
Staff sick leave records
Monthly infection surveillance rates
Curve shape
One sharp peak
Slow build over weeks, continuing source
Likely sources
A dish, a cook
Hands, fluids, equipment, a sink
Study design
Cohort, everyone listed
Often case-control, few cases
Control
Stop the dish
Cohort babies and staff, hand hygiene audit, environmental cultures
What must exist before the next outbreak?
Before
A named outbreak team with authority. Baseline surveillance data. A written outbreak policy. Lab and IDSP contacts.
During
Daily team meeting. One line list. Decision log with times. Staff roster backup. One spokesperson.
After
Written report. Fixes assigned with owners and dates. Follow-up audit.
The quality of your response on Tuesday was decided months earlier.
An outbreak ends with a report, not with relief
The outbreak is over when cases return to baseline for twice the longest incubation period. For Salmonella, about six days.
The report says
What happened, how we know, what we did and when, what it cost, and what we are changing.
What we changed
No next-day reheating of cooked meat. Temperature logs checked weekly. Food handler illness reporting with paid sick leave.
If nothing in the system changes, the report was only paperwork.
Three things to carry out of this room
An outbreak is more cases than expected. You need a baseline to know.
Compare the ill with the well. The line list, the epi curve and the attack rate table do this.
Act on a reasonable suspicion. Control and communication start on day one.
Sources and further reading
Centers for Disease Control and Prevention. Principles of Epidemiology in Public Health Practice, 3rd ed. Lesson 6: Investigating an outbreak. Atlanta: CDC; 2012.
Rasmussen SA, Goodman RA, eds. The CDC Field Epidemiology Manual. New York: Oxford University Press; 2019.