RESEARCH DESIGNS

MA. CARMEN C. TOLABING

Professor, Department of Epidemiology and Biostatistics
College of Public Health
University of the Philippines
At the end of the session on, the participants
should be able to:

1. Given a research objective, Identify the following:

1.1 study unit
1.2 study variable/s
1.2 appropriate study design
2. Define and differentiate the different research designs
3. Describe the steps in carrying out the different
study designs
4. Describe the elements of a research design
 Descriptive Analytic
Describe a disease/condition/ Examine association ( test of hypothesis)
phenomenon (treatment/program)
Observational Experimental
variables are Exposure variables
observed are assigned

Types
• Case report/ Case series
• Prevalence study
• Ecologic study Types Types

1) Cross-sectional 1) RCT
2) Case-control 2) Quasi Expt
3) Cohort

Risk/protective factors IDd

Hypothesis
Effect identified

Fig. 1 Summary of Research Designs

 Descriptive vs. Analytic

Descriptive
Assess frequency and distribution of
disease/health outcomes

Analytic
Assess presence of association between
two variables; cause-effect relationship
DESCRIPTIVE STUDY DESIGN
 Descriptive Studies
Inquiry into:

-- the amount and distribution of disease (health

condition/event) within a population or across
population groups

-- the pattern of occurrence of disease or

condition by person, place, time
 Types of descriptive studies
1. Case report /case series
- document unusual medical occurrences

- study of a case

study of a set of cases

Guidelines in doing case series
1. Case definition
- use diagnostic criteria

2. Obtain desired information from all cases

uniformly

3. Describe cases according to person,

place, and time
 Steps: case series
Decide whether to use prevalent or incident cases

Define a case and select the cases for inclusion in the study

Define the variables operationally and collect the data

Describe the cases according to person, place and time distribution

Figure 2. Procedure in carrying out a case series study

 Examples:
AIDS

5 cases of pneumocystis carini pneumonia (1980-1981)

among young, healthy, homosexuals in LA

Hypothesis regarding risk factors for development of AIDS

hepatic angiosarcoma

3 patients with angiosarcoma of the liver

males working in a vinyl chloride plant

Hypothesis: occupational exposure to vinyl

chloride is associated with hepatic
angiosarcoma
 2. Prevalence study
• Provide info on prevalence of disease or
other health conditions among individuals
in a defined population, subpopulations

(defined by geographic area, age,

occupation, etc)
 Steps: prevalence study
Define the study population and the study variables operationally

Draw an adequate random sample

Collect data

Calculate overall prevalence & subgroup prevalence of disease.

Figure 3. Procedure in carrying out a prevalence study

 Prevalence study
Used for studying

• prevalence of disease, health problem,

health service utilization, preferences, etc

• Assessing knowledge, attitude and

practices
 Schema of prevalence study
Choice of population

Study population
sampling

Study partcipants
Assessment of variables

Examined subjects

Subgrouping

Analysis of data (prevalence)

Examples: Prevalence study
Pop Disease

1 +
2 +
3 -
4 -

…100
Total 30
 3. Ecological Study
• Comparison of D and E among groups rather than
individuals

• Unit of observation is a group/unit of analysis is a

group
- E and D not known at the individual level
 Unit of observation
Table 1 (individual) Table 2 (group)
Pop Pop
Disease Exposure Disease Exposure

1 + + 1 ? ?
2 + - 2 ? ?
3 - + 3 ? ?
4 - - 4 ? ?
…100 …100
Total 30 50 30 50
Total
Group Disease rate or Exposure
death rate Rate
(per capita)

A 10 350 packs

B 20 700 packs

…J … …
 ex

http://www.euro.who.int/document/e81384.pdf
Colon Cancer Incidence and Meat Consumption

21
Advantages
- can be done quickly and inexpensively

Disadvantages
- cannot link exposure with disease at the
individual level (ecologic fallacy)

Ecologic fallacy
When conclusions about a relationship bet E and D are made at the
individual level based on ecologic data
 Summary: Descriptive studies
§ first step in risk factor ID
§ types:
1. case series – study of a population with
similar diagnosis or health condition

2. prevalence/survey – study of a population defined by

geographic location, occupation, etc.

3. ecologic – study of several population groups and

unit of observation is a group/unit of analysis is a
group

§ generate hypothesis regarding causal factors/ disease

transmission
Analytic Designs
 Analytic Studies

• Examine relationship between variables

Cause - Effect
Factor - Outcome
Exposure - Effect
Intervention - Effect
Treatment - Outcome
 Analytic studies
• Objective
To test a hypothesis
- a statement about the relationship
between 2 variables
(cause-effect relationship)

• Classification
observational or experimental
cross sectional or longitudinal
prospective or retrospective
 Analytic studies

• Key strategy:

Use of a comparison (or control) group

- to approximate the counterfactual scenario

Groups to be compared should be

comparable in terms of other risk factors of the
disease
- to control confounding (biased result)
 Classification of analytic studies
• Classification according to the presence of manipulation of the exposure
variable.
– observational
– experimental
• Classification according to the direction of inquiry on the presence of
exposure and the disease
– prospective
– retrospective
• Classification according to whether or not the assessment of the factor and
the outcome relate to two time points in the life of the study participant
– longitudinal
– cross-sectional
Observational vs. experimenal

Observational
observing the population for the
presence of the factor and the outcome

Experimental
manipulation of the factor
- deliberate application of the factor
- deliberate withholding of the factor
 Prospective vs retrospective
Factor Disease

?
Prospective

Factor Disease

?
Retrospective

Past Present Future

Longitudinal vs. cross-sectional
? factor ?outcome

Longitudinal

? factor ?outcome

Longitudinal

? Factor
?outcom e

Cross-sectional

past present future

Data layout for analytic studies
– 2x2 contingency table
Factor Disease Total

+ -

+ a b a+b

- c d c+d

Total a+c b+d n

Classification of analytic studies
• Classification according to the presence of manipulation of the
exposure variable.
– observational
– experimental
• Classification according to the direction of inquiry on the presence
of exposure and the disease
– prospective
– retrospective
• Classification according to whether or not the assessment of the
factor and the outcome relate to two time points in the life of the
study participant
– longitudinal
– cross-sectional
•
Observational vs. experimenal

Observational
observing the factor and the outcome

Experimental
manipulation of the factor
- deliberate application of the factor
- deliberate withholding of the factor
Prospective vs retrospective
Factor Disease

?
Prospective

Factor Disease

?
Retrospective

Past Present Future

Longitudinal vs. cross-sectional
? f actor ?outcome

Longitudinal

? f actor ?outcome

Longitudinal

? Factor
?outcome

Cross-sectional

past present future

COHORT DESIGN
 Cohort
Classification
§ Analytic (test hypothesis)
§ Observational ((-)manipulation of E)
§ Prospective (observe occurrence of D in the future)
§ Longitudinal (E D)

Study population (2 groups, FREE of disease of interest)

§ exposed (+ exposure, E )
§ unexposed (- exposure, E) compare proportion of outcome

Objective
§ To show that the probability of disease is greater in exposed than
the unexposed

Measure of Association
Risk Ratio = Risk in Exposed/Risk in Unexposed
 Factor ? Disease
+
+
-
+
-
- Factor

-
Present Future

Figure 8. Cohort study design

Cohort: Planning and execution
1. Defining and assembling the cohort
• should be susceptible or at risk of the
outcome of interest
(exclude: those with the outcome,
not at risk, immune)
• Select cohort with higher proportion
exposed to the factor of interest
2. Determining exposure status
• observation
e.g biochemical levels, envi/workplace exposure,
pollutants in ambient air
• query
• records review
3. Classifying the cohort into various
exposure categories
• may consider extent of exposure, length
of exposure
4. Selecting comparison group of unexposed
individuals
• internal comparison, those from the
cohort without the exposure
• external comparison, general
population where the exposed group is
obtained
5. Measurement of Outcome
• use diagnostic criteria
• method should be similar for both
exposed and unexposed groups
-observation (clinical or laboratory exams)
-review of records
med record – impression/discharge dx
death certificate – cause of death
 Factor ? Disease
+
+
-
+
-
- Factor

-
Present Future

Figure 8. Cohort study design

Example
Hypothesis:

? Study population:

? Study groups :

?Variable to be collected:
? Measure of Association
Factor Outcome Total

+ -
+ a b a+b
- c d c+d
≠

Total a+c b+d n

≠
Interpretation

RR = 1, no association
RR = 1, association : > 1 risk factor
< 1 protective factor

The risk of disease is RR X greater for those

exposed to the factor compared to the unexposed
? Interpretation and conclusion
CASE CONTROL STUDY
 Case control
Classification
§ analytic
§ observational
§ retrospective

Study population (study base)

§ Cases ( + disease, D)
compare proportion exposed (+factor)
§ Control ( - disease, D)

Objective
§ To show that the probability of exposure is greater in those with D
than in those without D

Measure of Association
• ODDS RATIO (= Odds of exposure in D+/Odds of exposure in D-)
Case Control

Exposure Disease

Past Present
? Factor Disease

+
-
+
-
-
past Present

Figure 9. Case control study design

Case control: Planning and Execution

1. Defining and selecting the cases

- those with the disease of interest
- use diagnostic criteria to define case

2. Defining and selecting the controls

- ensure absence of disease
- should be comparable to the cases
3. Assessment of exposure to risk factor
• Exposure should precede the disease
• Method should be the same for both
cases and controls
-Query
-Records review
-Observation (may not reflect levels
before/at the time the disease
developed
Example
Hypothesis:

? Study population:

? Study groups :

?Variable to be collected:
? Measure of Association
Factor Outcome Total

+ -
+ a b a+b
- c d c+d
Total a+c b+d n

≠
Interpretation

OR = 1, no association
OR = 1, association : > 1 risk factor
< 1 protective factor

The odds of exposure is OR X greater among those

with the disease compared to those who do not have
the disease
Or
The estimated risk of disease is OR X greater for
those exposed to the factor compared to the non-
exposed
CROSS SECTIONAL STUDY
 Cross sectional
Classification
§ Analytic (test hypothesis based on the coexistence of
both factor and outcome)
§ Observational ((-)manipulation of E)

Study population
Defined population (one group)

Objective
§ To show that the proportion of those with disease is
greater in exposed than the unexposed

Measure of association
• Prevalence ratio = Prev in Exposed/Prev in Unexposed
 -
+factor + factor - Factor - Factor
- disease + disease + disease - disease

Figure 10. Cross-sectional study design

Example
Hypothesis
? Study population:

? Study groups :

?Variables to be collected:
? Measure of Association
Factor Outcome Total

+ -
+ a b a+b
- c d c+d
Total a+c b+d n

≠
Interpretation

PR = 1, no association
≠
PR not = 1, association

Those exposed to the factor are PR X more likely

with disease compared to the unexposed
 Exposure Disease

?
Case Control
?

Exposure Disease

?
Cohort
?

Cross-sectional ? Disease
? Exposure

Past Present Future

 Experiment
Classification
§ Analytic (test hypothesis)
§ Experimental (+manipulation of E)
§ Prospective (observe occurrence of outcome in the future)
§ Longitudinal (E D)

Study population (2 groups)

§ Expt’l grp (+ exposure)
compare proportion of outcome
§ Control grp (- exposure)

Objective
§ To show that the probability of disease is greater in exposed than
the unexposed

Measure of Association
% Effectiveness Incidence Control – Incidence Experiment
Incidence Control
 Experimental Study
• Follows the cohort design except
that the two study groups are
randomized
• Limited use in studying disease
etiology because of ethical issues
• Utilized more for studying effects
of intervention
 Experiment

(+) outcome

• Experimental
group
Eligible (-)outcome
participant
- Indiv
- Pop RANDOMIZE
(+) outcome
Control
group

(-) outcome

Direction
of inquiry
 Exposure Disease

Cohort
?

Exposure Outcome

?
Experiment R
?

Past Present Future

 EXPERIMENT: PROCESS AND EXECUTION

Follows the cohort design except (-) randomization and (-) intervention
1.Formulate a testable hypothesis (specific)
= 5 elements
2. Define study variables operationally
2.1 factor/intervention (type and quantity)
2.2 outcome
3. Define study population (individuals/groups)
4. Randomize study population into 2 grps (>2)
4.1 Experimental group
4.2 Control group
5. Administer the intervention to experimental group
5.1 experimental group : (+) intervention
5.2 control group : (- ) intervention

6. Follow-up to determine incidence of outcome

7. Analyze data
7.1 Assess comparability of groups (?)

7.2 Determine if the intervention has any effect (beneficial)

= % effectiveness

Incidence(Control) – incidence(Expt)
Incidence (Control)

1- RR
True experiment:
random assignment of people or groups to the study groups

• Randomization creates two or more groups that are

equivalent , on the average, on just about any characteristic
= method to control confounding

BUT randomization not always possible .

1. TRUE EXPERIMENT (+ randomization)

= Randomized Controlled Trial (RCT)

2. QUASI EXPERIMENT (- randomization)

Example
Hypothesis
? Study population:

? Study groups :

?Variables to be collected:
 Summary
Factor Outcome Total
+ -
+ a b a+b Cohort
- c d c+d Experiment

Total a+c b+d n

Cross-sectional
Case control
 References
Kelsey, J. et al. Methods in Observational
Epidemiology (2nd ed). 1996. Oxford
University Press NY

Detels, R. et al (ed). Public Health. The

Methods of Public Health. 2002. Oxford
University Press. NY