Epidemiological study designs

Overview of the various epidemiologic study designs and hierarchy of epidemiologic strength

1. Epidemiologic
Study designs
Kayode Afolabi

2. Outline
•Introduction
•Descriptive Studies
•Analytical Studies
•Experimental Studies
•Summary

3. Introduction
•Epidemiology is the studyof the distribution and determinants of
health-related states or events in specified populations and the
application of this study to the control of health problems.
-John Last
•Objectives may be to determine/identify disease occurrence, course,
burden, risk factors/cause or effective/efficient prevention or
treatment options and overall influence control.
•Study: scientific data collection –observations or experimentation
and interpretation

4. Study Designs
Epidemiologic
Study
Observational
Descriptive
Case report
Case series
Ecological or
Correlational
Cross sectional
Analytical
Case –Control
Cohort
Experimental or
Interventional
Uncontrolled
Trials
Non-randomized
Controlled Trials
Randomized
Controlled Trials

5. Case report
•A detailed report of a SINGLE clinical subject/event
•Commonly used for unusual/unexpected and rare events
•Retrospective data collection
•No comparison group
•No statistical analysis

6. Case report
•Cannot be used to formulate hypothesis but is usually the first clue to
formulation of hypothesis
•Not generalizable
•e.g.
•Use of contraceptive and risk of thromboembolism
•Treatment of Recurrent Aphthous Ulcers in an HIV Patient –An Emerging Use
for Pentoxifylline
•Rib Exostoses Presenting as Mediastinal Masses: A Rare Presentation

7. Case Series
•A report of clinical characteristics or events of a group with similar
diagnosis
•From single or multiple source
•Also for unusual and rare conditions
•Retrospective or Prospective data collection
•Descriptive statistics can be used
•No comparison group
•Cheap and usually over a short period

8. Case series
•May be used to generate hypothesis
•Not generalizable
•Cannot study cause and effect
•Cannot assess disease frequency
•e.g.
•A new gram negative organism in the meningeal pus of 6 patients who died of
meningitis
•Atypical Causes of Urinary Tract Obstruction
•Nivolumab and Ipilimumab in the Treatment of Metastatic Uveal Melanoma:
A Single-CenterExperience

9. Ecological or Correlational Studies
•Variables are measured with a whole population as the unit of inquiry
•Data for these variables are usually from secondary data
•Cross sectional (single point) vs Longitudinal (repeated)
•Relatively cheaper, easier and quicker
•Can be used to generate hypothesis about possible exposure-
outcome relationship
•Useful for making large-scale comparisons

10. Ecological or Correlational Studies
•Findings may not be representative of individual characteristics ---
Ecological fallacy
•e.g.
•Percentage of children with no caries and mean annual exposure to
fluoridation
•per-capita sugar consumption andprostate cancer mortality

11. Ecological or Correlational Studies

12. Cross sectional study
•aka Prevalence study
•Studies a cross-section or Snap-shot
•Exposure and/or Outcome variables are observed at a point in time
•Often used to study frequent non-fatal and chronic diseases
•Can only measure prevalence not incidence
•Relatively cheap and quick

13. Cross sectional study
•Provides weak evidence of association
•Cannot establish causality
•Does not show temporal sequence
Time
Outcome
Exposure

14. Case Control Study
•aka Retrospective Study
•Seek to support or refute an hypothesis of association between an
outcome and an exposure
•Study subjects are selected based on presence of outcome (cases) or
absence of outcome (controls)
•Data collection goes backward in time to detect presence of exposure
(exposed) or absence of exposure (not exposed)
•Exposure odds amongst cases and controls are analysed to detect
association between exposure and outcome (Odds ratio)

15. Case Control Study
•Can be used to study several exposures
•Quicker and cheaper than Cohort study
•More useful for rare diseases than Cohort Study
•No issues with attrition
•Prone to bias from subject response or quality of records
•Cannot measure incidence
•Temporal relationship is difficult to determine

16. Case Control Study
ab
cd
Cases Controls
Exposed
Not
Exposed
Odds Ratio = Odds of a case being exposed
Odds of a control being exposed
a
a+c
÷
c
a+c
b
b+d
÷
d
b+d
OR = ÷
a
c
b
d
OR =
OR =
ad
bc

17. Cohort Study
•Study subjects are selected based on known exposure to a factor
(exposed) or non-exposure(not exposed or cohorts)
•These groups are then followed up prospectively for occurrence of
outcome
•Incidence rates of the outcome in the exposed and not exposed
groups are analysed to detect association (Relative Risk)

18. Cohort Study
•Can be used to study several effects of a single exposure
•Can measure incidence
•Dose response can also be studied
•Temporal relationship can be determined (cause --effect)
•More useful for rare exposures
•Time and resource consuming
•Prone to follow-up loss and attrition

19. Cohort Study
ab
cd
Disease
Present
Disease
Absent
Exposed
Not
Exposed
RelatuveRisk = Odds of a case being exposed
Odds of a control being exposed
a
a+c
÷
c
a+c
b
b+d
÷
d
b+d
RR = ÷
a
c
b
d
RR =
RR =
ad
bc

20. Time
Exposure
Outcome
Exposure
Outcome
Case Control Study
Cohort Study

21. Experimental Studies
•Involve deliberate intervention to manipulate onset or course of
disease
•Can establish cause of disease
•Can identify effective prevention or treatment options
•Incidence rate can be calculated
•Dose response can be studied
•Direct evidence of relative risk can be obtained
•Several outcomes can be studied

22. Experimental Studies
•Selection bias is minimal
•Time-consuming and expensive
•Various ethical issues
•Based on study population can be classified into clinical, field and
community
•Based on method of selection can be classified into
•Uncontrolled
•Non-randomized controlled
•Randomized controlled

23. Summary
Descriptive Studies
Experimental Studies
Ability to detect
association/cause
Ethical issues
Time and resources
Ease of
conducting study
Strength
Generalizability Bias

24. Thank
you