Professional Documents
Culture Documents
Week 3 Epi Tool Kit
Week 3 Epi Tool Kit
Cross-sectional
Sample of
population;
exposure and
outcome
measured at
same point in
time
Exposed
persons or
Prevalent cases
Ecological
Groups
Comparison
group
Non-exposed
persons or
Persons without
disease
Non-exposed
groups
Measures of
occurrence
Prevalence
Prevalence
odds
Group-level
prevalence
Group-level
prevalence odds
Group-level
risk
Group-level
rate
Group-level
odds
Average or
trend in: risk,
rate, prevalence,
or odds
Measures of
association
Prevalence
odds ratio
Prevalence
ratio
Prevalence
difference
Study
population
Persons of
interest
Exposed groups
Temporal
relationship
Multiple
associations
Can be hard to
establish
Can assess
several
exposures and
outcomes
Prevalence
ratio
Prevalence
difference
Prevalence
odds ratio
Risk ratio
Risk difference
Rate ratio
Rate difference
Correlation
coefficients
Regression
coefficients
Can be hard to
establish
Can assess
several
exposures and
outcomes
Time
required for
study
Relatively short
Relatively short
Attribution-NonCommercial-NoDerivs
CC BY-NC-ND
View License Deed | View Legal Code
Experimental
Voluntary
participants at
risk for
developing the
outcome of
interest at
baseline
Persons
assigned to a
treatment or
other exposure
Persons
assigned to not
receive
treatment being
investigated
(control / placebo
group)
Risk
Rate
Odds
Hazard
Risk ratio
Risk difference
Rate ratio
Rate difference
Odds ratio
Odds ratio
Hazard ratio
Survival curves
Efficacy
Easy to establish
Multiple
interventions on
single outcome
or effect of single
intervention on
more than one
outcome
Usually short,
depends on
disease
progression
Retrospective
Cohort
Sample of
population at risk
for developing
the outcome of
interest at
baseline
Prospective
Cohort
Sample of
population at risk
for developing
the outcome of
interest at
baseline
Case-control
Sample of
population;
cases and
controls from the
same source
population
Exposed
persons
Exposed
persons
Persons with
disease (cases)
Non-exposed
persons
Non-exposed
persons
Persons without
disease
(controls)
Risk
Rate
Odds
Hazard
Risk
Rate
Odds
Hazard
Odds of
exposure
Risk ratio
Risk difference
Rate ratio
Rate difference
Odds ratio
Hazard ratio
Survival curves
Risk ratio
Risk difference
Rate ratio
Rate difference
Odds ratio
Hazard ratio
Survival curves
Odds ratio
Sometimes hard
to establish
Often one
exposure with
multiple
outcomes,
though there are
exceptions
Easy to establish
Moderate,
depends on
obtaining followup data
Long, depends
on length of
follow-up
Often one
exposure with
multiple
outcomes,
though there are
exceptions
Which,
depending on
sampling, can
approximate:
Risk ratio
Rate ratio
Hazard ratio
Sometimes hard
to establish
One outcome
with multiple
exposures
Relatively short,
unless real-time
case acquisition
Updated by: Ricchetti-Masterson K, Lopes B, Yeatts KB. Original created by former EPID600 TAs.
Comparison of Epidemiologic Study Designs. UNC Gillings School of Global Public Health. August 2013.
Retrospective
Cohort
Generally less
expensive than
prospective
cohort study
Relatively large
Prospective
Cohort
Expensive
Ecological
fallacy
Reverse
causation
Confounding
Assessment of
outcome
Selection bias
Confounding
Information
bias
Onset of
disease is
prolonged
Rapidresponse settings
Measuring
descriptive
information
Inexpensive
Fast
Can often be
done using
publicallyavailable data
Usually good
generalizability
Individual level
information is
unavailable
Studying a
community-level
exposure
Exposure is rare
Assessment of
outcome
Selection bias
Confounding
Information
bias
Loss to followup
Exposure is rare
Inexpensive
Fast
Can often be
done using
publicallyavailable data
Can draw
conclusions
about group-level
characteristics
Can directly
estimate risks
and rates of
disease
Usually good
generalizability
Can directly
estimate risks
and rates of
disease
Fewer
problems
establishing
temporality
Usually good
generalizability
Sometimes
temporality
cannot be
established
Cannot
measure
incidence of
disease
Prevalence
varies with
duration of
disease
Inappropriate
conclusions may
be drawn
regarding
relationships at
the individual
level based on
ecological data
(ecological
fallacy)
Selection of
non-exposed
comparison
group often
difficult.
Changes over
time in disease
diagnosis /
treatment criteria
and research
methods
Loss to followup
Selection of
non-exposed
comparison
group often
difficult.
Changes over
time in disease
diagnosis /
treatment criteria
and research
methods
Loss to followup
Experimental
Very expensive
Cross-sectional
Generally
inexpensive
Ecological
Generally
inexpensive
Population
size needed
Relatively small
Can be large or
small
Usually large
since entire
populations are
studied
Potential
biases
Assessment of
outcome
Information
bias
Loss to followup
Survival bias
Reverse
causation
Confounding
Information
bias
Best when
Evaluating
treatment options
(drug,
counseling)
Vaccine trials
Advantages
Provides clearest
evidence of
causality
Challenges
Ethical
problems
(equipoise): very
few exposures
can be assigned
Low
generalizability
Very expensive
Cost of
study
Attribution-NonCommercial-NoDerivs
CC BY-NC-ND
View License Deed | View Legal Code
Relatively large
Case-control
Relatively
inexpensive
Much smaller
than other
similarlypowered studies
(i.e. cohort)
Assessment
of exposure
Selection bias
Confounding
Information
bias
Outcome is rare
Relatively
inexpensive
Fast
Can estimate
risks and rates
of disease
(under specific
sampling
parameters)
Needs fewer
participants
than cohort
Selection of
appropriate
controls often
difficult
Incomplete
information on
exposure
Updated by: Ricchetti-Masterson K, Lopes B, Yeatts KB. Original created by former EPID600 TAs.
Comparison of Epidemiologic Study Designs. UNC Gillings School of Global Public Health. August 2013.