Selecting a Study Design

Goals
Describe the cohort study design.
Describe the case-control study design.
Compare situations in which cohort and
case-control study designs should be
used.
About Analytic Studies
We can use analytic studies to test hypotheses.
We want to know:
Whether there is an association between hypothesized
exposure and disease,
How strong the association is,
What proportion of cases are due to exposure, and
Whether there is an increased risk of disease with increased
exposure (a dose-response relationship).
Two common types of analytic studies are cohort
study and case-control study.
What is a Cohort?
A cohort is a group of people who have
something in common.
Can represent the source populationthe
population from which cases of disease arise.
Examples of cohorts:
All employees in an office building
Everyone who attended a football game
All the residents of a neighborhood
Cohort Studies
Tend to be retrospective (exposures in the
past in relation to disease that has already
happened).
Occurrence of disease in exposed group
compared to occurrence of disease in
unexposed group = risk ratio.
Risk ratio tells whether disease is associated
with exposure and strength of association.
Identifying a Cohort
To use a cohort study, you must identify
every person in the cohort.
Possible when the group is small and well
defined (e.g., wedding reception, cruise
ship, school, prison).
Option to interview every member of
the cohort or a sample of the cohort.
Identifying a Cohort
Sometimes it may be difficult to define
a suitable cohort.
Can you find every single person who ate
at the Main Street Deli on January 10-20?
How would you locate every person buying
and/or eating contaminated lunch meat
from a local supermarket chain?
An alternative: the case-control study.
Case-Control Studies
The most frequently used type of study in
outbreaks.
Can be quickly implemented.
Can be used when cohort study might be large
and time-consuming.
Identify people with disease (case-patients)
and people without disease (controls), then
ask everyone about past exposures.
You already know who is sick through doctor
diagnosis, lab culture, or health department.
Case-Control Studies
Calculate odds ratio to measure
strength of association between illness
and exposure.
Compare odds of exposure among case-
patients to odds of exposure among
controls.
Cannot calculate risk ratio in case-
control study.
Selecting Cases and Controls
Defining the source population may help
narrow down potential controls.
Do the cases live in the same city or attend the
same event?
Are they of a particular race or ethnicity?
Understanding where cases came from will
help select your controls.
Controls are a sample of people from the
source population.
Selecting Cases and Controls
Example: Outbreak of gastrointestinal illness linked to
eating at the Main Street Deli during January 10-20.
Cases recruited from people who ate at the Deli and
experienced vomiting. Controls recruited from people who
ate at the Deli but did not experience vomiting.
All cases recruited into study; only a portion of healthy
controls contacted because could not identify every person
who ate at the restaurant during these 10 days.
Want to know what case-patients and controls ate.
Controls selected from customers who ate at the Deli
during the time period of interest.
Case-Control or Cohort:
Which one is right?
The choice depends on the situation.
Always think about the source
population:
Are members of the group easily
identifiable? Can you interview all or a
sample of them?
Use a cohort study.
Is the cohort difficult to identify or too
large to contact all members?
Use a case-control study design.
 Case-Control or Cohort:
Which one is right?
Retrospective cohort study is the most appropriate
study design here.
If fewer than 200 people involved, should consider
interviewing everyone.

Figure 1: Easily identifiable cohort (e.g., church picnic, wedding, luncheon)

Not ill
n=81
Ill
n=34

Total N = 115
 Case-Control or Cohort:
Which one is right?
A case-control study could be used for efficiency
here.
Or capture entire cohort using e-mail or mail surveys.
Or identify cohorts within the larger cohort (e.g., a single
dormatory on a college campus).
Figure 2: Easily identifiable but large cohort (e.g., cruise ship, college campus)

Not ill
n=2354
Ill
n=21

Total N = 2375
Case-Control or Cohort:
Which one is right?
Figure 3: Selecting controls for a rare disease in a large cohort: case-control
design and efficient selection of controls.

Hispanic residents in a county

Female Hispanic residents in a

county
Female Hispanic residents of
child-bearing age in a county
Pregnant Hispanic residents in
a county
12 cases of listeriosis among
pregnant Hispanic women
Case Studies:
Yersinia and chitterlings
11/15/20012/15/2002: 12 cases of Yersinia
enterocolitica identified at large urban pediatric
emergency department in Tennessee.
Source population: black infants with access to
medical care from population served by hospital.
Controls chosen: black infants who presented to the
emergency department of same hospital with chief
complaint other than gastroenteritis.
Case-control study implicated source of outbreak:
Chitterlings prepared in 100% of case households but only
35% of control households.
Parents able to identify ways kitchen might have become
contaminated (e.g., chitterlings cleaned in sink).
Case Studies:
Pseudomonas from ear piercing
September 2000: Oregon physician treating 2 patients on 2
consecutive days with infections of the cartilage of the ear; both
patients received ear piercings at same kiosk.
Investigators could contact all patrons of kiosk; used a cohort
study design:
118 people received 186 piercings August 1 through September 15.
7 piercings (4%): laboratory-confirmed Pseudomonas aeruginosa.
18 piercings (10%): suspected case.
Risk of infection increased if piercing in cartilage rather than
earlobe.
The investigators were able to:
Determine the risk of infection among the entire population,
Determine that the risk was different based on site of piercing, and
Identify practices that might have led to contamination of
equipment and subsequent infection.
Conclusion
Cohort and case-control studies are both
options for determining cause of an outbreak.
Both study the source population.
Cohort uses entire population or representative
sample.
Case-control uses all cases of disease and
sampled controls.
Both types of studies are effective; your
choice will depend on the circumstances of
the outbreak you are investigating.
References
1. Dwyer DM, Strickler H, Goodman RA, Armenian HK. Use of
case-control studies in outbreak investigations. Epidemiol Rev.
1994;16(1):109-123.
2. MacDonald PM, Whitwam RE, Boggs JD, et al. Outbreak of
Listeriosis among Mexican Immigrants as a Result of
Consumption of Illicitly Produced Mexican-Style Cheese. Clin
Infect Dis. 2005; 40:677-682.
3. Jones TF. From pig to pacifier: chitterling-associated yersiniosis
outbreak among black infants. Emerg Infect Dis.
2003;9(8):1007-1009.
4. Keene WE, Markum AC, Samadpour M. Outbreak of
Pseudomonas aeruginosa infections caused by commercial
piercing of upper ear cartilage. Jama. 2004;291(8):981-985.