Objectives

At the end of this session, students will be able to:

CHAPTER-IV • Categorize epidemiologic study designs
• Differentiate between descriptive and analytic study designs

Epidemiological study designs • Differentiate between observational and interventional study designs
• Describe different descriptive study designs
and
• Explain the strength and limitations of descriptive studies
measures of association and effect • Describe different kinds of analytic study designs
• Explain the merits and demerits of each analytical study designs
• Apply specific descriptive and analytic epidemiologic study designs
• Calculate and use different measures of association and effect (impact)
2

Contents Study design

• Study designs
• Descriptive study designs • Study design refers to a specific plan or protocol for
– Case report conducting the study, which allows the investigator to
– Case series
– Cross-sectional translate the conceptual hypothesis into operational
– Ecological/Correlational hypothesis.
• Analytic study designs
• It’s the procedures and methods, predetermined by an
– Observational (comparative cross-sectional, case-control,
cohort) investigator, chosen and to be adhered in conducting a
– Interventional (experimental, quasi-experimental) research project.
• Measures of association and effect (impact)
3 4

Classification of study designs

Cont…
Study Designs
• Generally epidemiological studies are used to provide
information on three areas:
Descriptive Analytic
1. Distribution and frequency of diseases, and known
and possible causes of diseases in populations
2. Strength of associations between diseases and other Correlational Cross-sectional Observational Interventional
factors (e.g. smoking, diet or socio-economic status),
Case-report/series Comparative
with particular emphasis on whether such Cross-sectional Experimental
associations are causal
3. Whether interventions aimed at preventing a disease Case-control
Quasi-
or improving its outcome actually do so Experimental
Cohort
5 6
Cont… A) Descriptive study designs
Study Designs
Does the study test hypothesis
or have comparison groups • Observational study designs which describe patterns
No Yes
of disease occurrence in relation to person, place and
Descriptive studies Analytical studies

Is the study unit individual?

time.
No Yes
• Often the first step or initial inquiry into a new topic,
Ecological/ Case report
correlational Case series Does the researcher intervene event, disease or condition
Cross-sectional the natural course of action?
No Yes • Can be divided into two roles:
Comparative cross-sectional Interventional – Designs that emphasize features of a new condition
Caes-control - Experimental – Designs that describe the health status of communities or
Cohort - Quasi-experimental
populations
7 8

Cont… (time, place, person) Cont…

Time – reflects seasonal and secular (long-term) trends of • Do not have a comparison (control) group
diseases
Place – provides information on geographic distribution of – do not allow for inferences to be drawn about
the disease
Person – indicates disease occurrence by personal
associations, casual or otherwise.
characteristics such as:
– but can suggest hypotheses that can be tested in
• Inherent characteristics (age, ethnic group, gender)
analytical observational studies.
• Acquired characteristics (educational, marital, immune,
or nutritional status),
• Activities (occupation, leisure activities, use of alcohol,
tobacco, or medications), or
• Living conditions (socioeconomic status, access to
health care)
9 10

Types and unit of study Types of descriptive study designs

1. Case report
– Case report
– Describes experience of a single patient.
– case-series deal with individuals
– Case reports are quite common in medical journals (account
– cross-sectional
for over 1/3 of all articles published).
– Typically depict an observant clinician identifying an unusual
– Ecological/correlational - examines population feature of a disease or a patient's history.
E.g. 40 years old premenopausal woman who developed
pulmonary embolism 5 wks after beginning to use OCP
to treat endometriosis.
11 12
Cont… Cont…
Explanations
Advantages
• Since pulmonary embolism is far more common in older,
postmenopausal women, drug may have been responsible for – Can represent the first clues in the identification of new
this rare occurrence. diseases or adverse effects of an exposure
• Since the use of OCP is not an unusual occurrence among the – Can prompt further investigations with more rigorous study
women in this age group, it was, of course, also possible that
designs
some other characteristics of the patient (e.g. medical Hx) could
have accounted for this occurrence in which case the use of OCP – Useful to public health as they provide an interface between
is merely coincidental. clinical medicine and epidemiology
⇒ The crucial question: whether women who develop PE are more
likely to have used OCP than who do not use it.
13 14

Cont… Cont…

Limitations 2. Case Series

– No appropriate comparison group – Report on a series of patients with common outcome of
– Cannot be used to test for presence of a valid statistical interest (disease).
association – Collections of individual case reports which may occur within
– Since based on the experience of one person, presence of
a fairly short period of time and aggregated into one
any risk factor may be purely coincidental
publication
– Prone to atomistic fallacy
Example
– Have historical importance in epidemiology

At an individual level a high income may be associated with • It was often used as an early means to identify the
lower rate of suicide but this does not mean that societies beginning or presence of an epidemic.
which are rich have a lower rate of suicide. 15 16

Cont… Cont…

Advantages Limitations
• Used as an early means to identify the beginning or presence of • No appropriate comparison group
an epidemic • Cannot be used to test for presence of a valid statistical
• Very useful for hypothesis generation
association
• Can suggest the emergence of a new disease
• Prone to atomistic fallacy
E.g. 5 young previously healthy homosexual men were found to have
PCP at 3 Los Angeles Hospitals during a 6-month period in 1980-1981.
This clustering of cases was striking in that, until then, PCP had been
seen almost exclusively among older men and women whose immune
systems were suppressed. This unusual circumstances led to the
diagnosis of new disease, subsequently called AIDS.
17 18
Cont… Cont…

3. Cross-sectional (Prevalence/survey) study design Design of cross-sectional study

– Observation of a defined population at a single point in time
or over time interval
– Exposure and outcome are determined simultaneously
– Because exposure and outcome are measured at the same
time point, the temporal sequence is often impossible to
determine (“Chicken or Egg” Dilemma)
– Since it considers prevalence than incidence, data will reflect
determinants of survival as well as etiology.
19 20

Cont… Cont…

Steps in the conduct of cross-sectional studies Analysis of cross-sectional data

1. Define a population of interest (reference population) Numerators - two probabilities:
2. Recruiting a representative sample (adequate size, random • Referring to the outcome - prevalence rate of the
outcome is compared between exposed & non-exposed
selection) subgroups
3. Measure the variables of interest (disease &/or exposure) at the
• Referring to the exposure - the prevalence rate of the
exposure is compared in those with & without the
same point in time outcome
Denominator - total population in the sample, based on
4. Sorting (if necessary) & Analysis of the data the comparison being made (exposure or
outcome prevalence)
21 22

Cont… Cont…
Example Advantages
Outcome (COPD)
– Are a one-step, one-time collection of data
Exposure Yes No Total
(Smoking) Yes 70 50 120 ⇒ Less expensive & more expedient to conduct
No 30 70 100 – Useful for planning health services and medical programs
Total 100 120 220
– Evaluate medical care and health service delivery
Prevalence of smoking among COPD patients? – Show relative distribution of conditions, disease, injury and
= 70 x100 =70%
disability in groups and populations
100
Prevalence of COPD among smokers? – Based on a sample of a major population rather than on
= 70 x100 =58.3% individuals that present themselves for medical treatment.
120 23 24
Cont… Cont…

Disadvantages…
Disadvantages
– Can identify only prevalent cases rather than incident
– Cannot establish whether the exposure preceded disease or
disease influenced the exposure (potential exposure bias) cases (i.e. does not yield incidence or true relative risk)
⇒ ‘chicken or egg’ dilemma
– Potential sampling bias since only survivors are available
Example
for study (may under represent diseases with short
Lower levels of β-carotene among cancer patients than healthy duration)
individuals of same age and sex was found by cross-sectional
study. – Not feasible for rare conditions
– However, for factors that remain unaltered over time, such as – It may not show strong cause-effect relationship if sample
sex, race or blood group, it can provide valid statistical
association though such instances are rare. size is small.
25 26

Cont… Cont…

4. Ecological/Correlational study – The aim is to see relationship between disease and exposure
– Uses data from entire population to compare disease as they occur among groups of people.
frequencies: Examples

between different groups during the same period of time, or • Average per capita fat consumption and breast cancer rates

in the same group (population) at different points in time compared between countries

– Focuses on groups of people (rather than individuals) as the • Comparing incidence of dental caries in relation to fluoride
units of analysis content of the water among towns in the rift valley
⇒ Does not provide individual data, rather presents average • Mortality from CHD in relation to per capita cigarette sales
exposure level in the community
27
among the regions of Ethiopia 28

Cont… Cont…

Figure 1. Correlation between per capita meat consumption and colon Figure 2. Breast Cancer Mortality and Dietary Fat Intake in various
29 30
cancer among women in various countries countries
Cont… Cont…

Figure 4. Forecast of cancer deaths if present trends continue (Data from

Figure 3. Expected and observed deaths from coronary heart disease in
31 the American Cancer Society) 32
US from 1968-1977

Cont… Cont…

Advantages
Assumption
• Quick and inexpensive (use already available information)
– The constancy of the characteristics in all individuals in the
• May be superior to individual level studies (group averages
group.
versus difficulty of measuring individual level exposures
Example
because of large intra-person variations)
• If exposure is pollution of the air, all individuals have
• Data can be used from populations with widely differing
the same level of exposure as the other individual in
characteristics
the group

33 34

Cont… Uses of descriptive study designs

Disadvantages 1. Health care planning

• Ecologic bias/fallacy (fail to reflect the effect at the biological – provide knowledge about which populations or subgroups

or individual level) are most or least affected by disease.

• Potential confounding factors cannot be readily controlled – enable public health administrators to target particular

• Usually rely on data collected for other purpose segments of the population for education or prevention
programmes
• It may mask a non-linear relationship between exposure and
disease – can help to allocate resources more efficiently.

35 36
Cont… Common misuse
2. Hypothesis generation • Establishing a casual relationship not supported by the
– identify descriptive characteristics which frequently data
constitutes an important first step in the search for
– Can highlight associations between exposure and
determinants or risk factors
outcome variables, but cannot establish causality
3. Trend Analysis
– used to follow patterns of disease change over time.

E.g. malaria trends at different times.

37 38

Summary

• In general, descriptive study designs:

– Characterize disease occurrence by time, place and

person

– Generate testable hypothesis as to the cause of

diseases but can not establish the causality by testing

Thank You!
hypothesis

– Important for health care planning and trend analysis

39 40

B) Analytical study designs Types of analytic studies

• Analytical study attempts to establish causes or risk factors 1. Observational studies

A. Comparative cross-sectional study
for certain problems.
B. Case control study
• This is done by comparing two or more groups, some of which C. Cohort study
have or develop the problem and some of which not. 2. Interventional studies
• It answers the questions “why some segment of the population  Experimental:
A. Therapeutic trial (clinical trial)
are affected by disease while others are not”.
B. Preventive trial (field trial or community trial)
• Ultimate goal is judging whether a particular exposure causes or  Quasi-experimental
prevents disease. => hypothesis testing
• Data are more likely gathered for specific study.
41 42
 1. Comparative cross-sectional studies Comparative cross-sectional…

• Many cross-sectional surveys focus on describing as well as Example

comparing groups. • If you assume there is a difference between model and non model

Example families in KAP on family planning, you can separately interview

– By comparing malnourished and well-nourished children, mothers from both model families and non model families to

one can try to determine which socio-economic, behavioral determine the extent of difference in KAP on family planning.

and other independent variables may have contributed to ⇒ Two populations are compared in terms of:

malnutrition. • Magnitude (prevalence) of the problem under consideration

• In any comparative study, one has to watch out for • Factors associated with the problem under consideration
confounding or intervening variables.
43 44

Comparative cross-sectional… 2. Case-control study

Advantages
• Cases (those having a specific disease)
Early and and Controls
important (thosestudy:
case-control who
• Quick and inexpensive
do not have the disease)Doll
are and
compared for their
Hill (1950) exposure status.
to investigated the
• Better than descriptive cross-sectional study in assuring
temporal relation. relationship between cigarette smoking and
Example
lung cancer.
• No loss to follow up
Identify people with lung cancer (cases) and people with out
Disadvantages
• Causal associations can not be made (can not directly compute lung cancer (controls) and then ask both groups whether they
risk).
• Survival bias. are smokers (exposure)
• Recall bias (exposure may be difficult to remember) if no • It assesses retrospectively exposure status
record.

45 46

Case-control… Case-control…
Application (design) Design…
• Identify group of participants with disease (cases) and group of
participants without disease (controls)
• Then look back historically to identify differences in predictor
variables [exposure (s)]
• Then look at the level of the risk factor in a group of case
compared to a group of healthy subjects (control)
• If the level of the assumed risk factor is higher in the cases, it
implies an association between risk factor and disease.

47 48
Case-control… Case-control…

Assumption Sources of cases and controls

• Cases should represent all cases in the population  Population-Based study

• Controls should represent all persons without disease in the
Cases – from Registry (fed by population)

population
Controls – from General Population

 Hospital-Based study

Cases – selected group that made visit to hospital

Controls – people with no similar disease (cases unrelated

to the problem of interest) in the hospital

49 50

Case-control… Case-control…
Advantages Disadvantages

 Optimal for the evaluation of RARE disease  Inefficient for the evaluation of rare exposure

 Can examine multiple risk factors for a single disease  Can not directly compute risk

 Quick and inexpensive  Difficult to establish temporal relationship

 Relatively simple to carry out  Determining exposure will often relay on memory

⇒ Information bias (Recall bias) is possible

 Persons who die as a result of disease caused by the

determinant may not be known to the study

51 52

Case-control… 3. Cohort study

Disadvantages…
Definition of “cohort”
 Can’t determine incidence or prevalence (period prevalence)
Group of people who have something in common when
 Sampling bias:
they are assembled.

cases are not representative of all those with disease since
 A group of individuals that are all similar in some trait
missed, misdiagnosed, or dead wont’ be studied.
and move forward together as a unit.

also people with access to care are overrepresented.
E.g. Birth cohorts, cohort of smokers, cohort of

occupational exposures

53 54
 Cohort study…
Cohort study…
General design
Earlyare
 In cohort study, participants and important
classified cohort
on the basisstudy:
of their
specific exposure status and are followed for a specific time to
Framingham Heart Study (1950s) to
explore the relationships of a wide variety
determine for the development of a disease
of risk factors (outcomeheart
with coronary of interest).
disease.
 There is usually a follow up.

 Their primary purpose is to analyze associations between the

outcomes and risk factor (usual type)

⇒i.e. used in confirming hypothesized relationship explored by

descriptive study.
55

Cohort study… Cohort study…

Types of cohort studies Types…

 Closed vs. Open cohort  Incidence vs. Prognostic cohort

Closed/fixed cohort: exposure groups are defined at the
Incidence Cohort Study
start of follow up and no new members are added during • To assess incidence of disease
the follow up
• To identify risk factors for disease onset

Open/dynamic cohort: people move in and out during the
• Is incidence greater in exposed than non-exposed?
follow up

57 58

Cohort study… Cohort study…

Design of Incidence Cohort Types…

No Disease Exposure to Disease
Prognostic Cohort Study

Risk Factor – Follow diseased cohort to assess factors associated with
Yes
outcome (recovery, complication or death)
Exposed No – Goal is to identify explanatory/prognostic factors of those
Population
without factors contributed to the development of the disease
Sample Time
Disease
Not Exposed Yes

No 59 60
Cohort study… Cohort study…

Design of Prognostic Cohort  Prospective vs. Retrospective

Diseased Prognostic Recovery  This is depending on temporal relationship between initiation
Factor
of the study/timing of data collection/ and occurrence of the
Yes
disease.
Present No • The term “prospective” refers to the timing of data
Population
collection and not to the relationship between exposure
incident Sample Time
cases and effect.

Absent Yes ⇒ there can be both prospective and retrospective

No 61 cohort studies. 62

Cohort study… Cohort study…

Prospective cohort studies Design

 Exposure data is collected before start of the study whereas
outcome data is after start of the study

Conduct

• Cohorts are identified in the present

• Exposure status or possible explanatory/prognostic

factors determined in the present

• Cohorts followed up to identify outcome

Nov 2013 Nov 2013 Nov 2015
• Ascertainment of outcome done in future 63 64

Cohort study… Cohort study…

Retrospective (historical) cohort Design

 All the exposure and outcome of interest occurred before the
actual study begins.

Conduct

• Identify cohort in the past using records/databases

• Determine exposure or prognostic factors in the past

using again records or databases
• Identify outcome in past or present
Nov 2006 Nov 2006 Nov 2013
NB: outcome can also be identified in the future (in
case of mixed cohort) 65 66
Cohort study… Cohort study…

Retrospective cohort... Retrospective cohort...

 Can be conducted more quickly and cheaply
 Costs can occasionally be reduced by using a historical cohort
(identified on the basis of records of previous exposure)
All relevant events have occurred

 This sort of design is relatively common for studies of cancer
Efficient for disease with long latency periods
related to occupational exposures (long latency period)  Depend on availability of routine data
Example

Incompleteness
Records of military personnel exposure to radioactive fall-out at

Lack of data on confounding variables
nuclear bomb testing sites have been used to examine the
possible causal role of fall-out in the development of cancer over
the past 30 years 67 68

Cohort study… Cohort study…

Selection of exposed group Selection of comparison group

 Depends on frequency of exposure – Subjects similar in all other factors except the determinant

Common exposures: general population factors

Rare exposures: selected groups – Internal comparison with in the cohorts

– Outcome must not be rare in exposed group – Special comparison group

– Attributable risk must be high (cause/effect relation) – General populations

 Depends on accessibility

69 70

Cohort study… Cohort study…

Advantages Disadvantages
• Valuable when the exposure is rare • Inefficient in evaluation of rare diseases
• Can examine multiple effects of a single exposure • Expensive (costs, time consuming)
• Can elucidate temporal relationship • Loss to follow up affects validity (the most important source
• Allows direct measurement of risk/incidence of bias)

• Minimizes bias in ascertainment of exposure • Retrospective cohort requires adequate records

71 72
 4. Experimental study Experimental…

• Experimental studies are investigations in which the researcher • The participants in the study who actually receive the
decides which participants (person, animal, etc) will be exposed treatment of interest are called the treatment (experimental)
to, or deprived of the factor. group.
• In an experimental study, we determine the exposure status for • Those who receive no treatment or a different treatment are
each individual (clinical trial) or community (community trial); called the comparison (control) group.
then follow the individuals or communities to detect the effects of
• Experimental design is the gold standard study design
the exposure.
compared to other designs because it can produce high quality
data.

73 74

Design outcome Types of experimental study

Intervention
no outcome
Study
1. Clinical trial
population outcome  Usually performed in clinical settings and the subjects are
No
Intervention patients
no outcome
 Randomized clinical Trials (RCTs) are special type of clinical
trials where the researcher randomly assigns participants into
population
Source

Direction of inquiry experimental & control groups.

Time
75 76

Types of experimental… Types of RCT (according to purpose)

A. Prophylactic trials - trials to test disease prevention.
Approach to RCTs
e.g. immunization, contraception, etc
• Define treatment and control groups. B. Therapeutic trials -trials to test established disease treatment.
• Administer exposure to treatment group, but not to controls. Used to test:
• Follow through time and compare rate of outcome in treatment  Efficacy – testing if therapeutic agent work in an ideal,
group with rate of outcome in control group. controlled situation.

⇒Investigator is involved during the entire time from  Effectiveness – testing if the therapy introduced to the

exposure to outcome development. population at large is effective when dealing with other
co-interventions, confounding, contamination, etc, after
77 having established efficacy. 78
Types of RCTs… RCT…
.
Advantages of RCT
C. Safety trials – trials to test safety of the agents
• Randomization balances prognostic factors across the groups
E.g. side effects of oral contraceptives and injectables

D. Risk factor trials – trials to test etiologic agent of diseases • Randomization minimizes selection bias and confounding

E.g. proving the etiologic agent of a disease by inducing it • Greater likelihood that patients, staffs, and assessors can be
with the putative agent in animals, or withdrawing the agent blinded
(e.g. smoking).

79 80

RCT… Types of experimental…

Disadvantages of RCT 2. Field trials (FTs)

• Expensive in terms of time and money – Conducted at fields rather than clinics on healthy subjects

– Used in testing medicine for preventive purpose

• Frequently, need a large number of patients
3. Community trials (CTs)
• Ethical issue (potential effective Rx may be withheld from some
patients, or some may be exposed to a dangerous one) – Unit of study is group of people in the community

• Subjects may not comply (cross-over) – The major difference between RCTs and CTs is that
randomization is done on communities rather than
• Rare or late adverse effects of intervention may not be picked up
individuals.

81 E.g. Fluoridation of water to prevent dental caries 82

5. Quasi-experimental study Quasi-experimental….

• In a quasi-experimental study, one characteristic of a true

experiment is missing, either;

Randomization before

or

Separate control group

• A quasi-experimental study, however, always includes the

manipulation of an independent variable (i.e. intervention).

83 84
Quasi-experimental….

Thank You!
85 86