Noncommunicable diseases - descriptive & analytical studies

Descriptive &Analyticalstudies
in Noncommunicable diseases
Dr. S. A. Rizwan, M.D.
Public Health Specialist
SBCM, Joint Program – Riyadh
Ministry of Health, Kingdom of Saudi Arabia

Learning
Objective
Identify the following for an NCD problem:
● the type of study to conduct (descriptive or analytic, case-control
or cohort)
● what data to collect
● the appropriate sources for that data
● the appropriate measure of association for the data
● and the sampling method
Descriptive and Analytic Studies 2

Lesson
Overview
 Principles of epidemiologic studies
 Definition and characteristics of descriptive studies
 Definition and characteristics of analytic studies
 Methods of sampling

Principles of
epidemiologic studies
4

WhyConduct
Studies?
 To learn what causes a disease
in order to prevent it
 To generate and test
hypotheses in order to learn
about disease cause-effect
relationships
 Causes are often called
exposures
 Effects are often called
outcomes

Exposures and
Outcomes
Exposure: Factors that increase a person’s risk for an
adverse health condition
Outcome: Effect of exposure (often, adverse health
condition)
 Different levels of exposure can be associated with different
outcomes
Exposure Outcome

Research
Questions
1. Does obesity increase the risk for diabetes?
What is the exposure?
What is the outcome?
2. Does diabetes increase the risk for cardiovascular
disease?
What is the exposure?
What is the outcome?
obesity
diabetes
diabetes
cardiovascular disease

Risk Factors
andCauses
 Risk factors may not be the same
as causes
 Men are at higher risk for lung
cancer, but being male is not a
cause of lung cancer
 Sleepiness, intoxication, and
dangerous road conditions are
risk factors for car crashes and
can also be the cause

Establishing
Temporality
 Causes must occur before the disease
 Causes must occur within a certain time frame
 Lung cancer often takes years to develop

Study Designs
toAssess
Cause and
Effect
Two types of epidemiologic studies:
 Experimental: Investigators expose study subjects to a
potential risk factor (or cure) and document outcomes
 Observational: Investigators observe the effect of
exposures on study subjects

Classification
of studies

Descriptive or
Analytic
Studies
Epidemiologic studies can be descriptive or analytic.
 Descriptive studies
 Ask who, where, when, and what
 Generate hypotheses
 Analytic studies
 Test hypotheses
 Ask why and how

Research
Question
Your office is interested in assessing the relationship between
number of years spent in school and age at first sexual intercourse.
1. What is the exposure?
2. What is the outcome?
3. Descriptive or analytic study?
4. Experimental or observational study?
Years of schooling
Age at first intercourse
Descriptive
Observational

Definition and
characteristics of
descriptive studies
17

Descriptive
Studies
●Characterize who, where, and when in relation to what (outcome)
 Person
 Characteristics (age, sex, occupation) of the individuals affected by the
outcome
 Place
 Geography (residence, work, hospital) of the affected individuals
 Time
 When events (diagnosis, reporting; testing) occurred

Types of
Descriptive
Studies
Aggregate
Ecological
Study
Individual
Case Studies
Case Series
Cross-
sectional
Study
 Often only cases included
 Cannot draw conclusions
about exposure-outcome
relationships
 Generate frequencies and
proportions

Cross-
Sectional
Study
Purpose: To learn about the knowledge, attitudes, and practices of
a population at one point in time (like a photo “snap shot”)
Design: No comparison group
Population: All members of a small, defined group or a sample from
a large group
Results: Produces estimates of the prevalence of the outcome of
interest

When to
Conduct a
Cross-
Sectional
Study
 To estimate burden of a health
condition in a population
 To learn knowledge, attitude
and practices of individuals in
a population
 To monitor trends over time
with serial cross-sectional
studies

Cross-
Sectional
Study
Measures
 Prevalence
= (# existing cases) / population

Practice:
Cross-
Sectional
Studies
CHD
No
CHD Total
Total 100 900 1000
Prevalence of CHD = ?

Practice:
Cross-
Sectional
Studies
CHD
No
CHD Total
Total 100 900 1000
Prevalence of CHD =
100 / 1000 = 0.10 or 10%

Example:
Cross-
Sectional
Study
Objective
 To estimate the magnitude and patterns of violence against
pregnant women
Study
 Population-based, household, cross-sectional study in Mbeya and
Dar es Salaam,Tanzania, 2001-2002
Result
 Violence experienced by 7% in Dar es Salaam and 12% in Mbeya

Example:
Cross-
Sectional
Study
Trends in diabetic death rates among US youth 1968-2010

Example:
Cross-
Sectional
Study
Prevalence of diabetes in Saudi Arabia, 1982-2009.

Example:
Cross-
Sectional
Study

Developing
Hypotheses
 A hypothesis is an educated guess
about an association that is testable
in a scientific investigation
 Descriptive data (Who?What?
Where?When?) provide
information to develop hypotheses
 Hypotheses tend to be broad
initially and are then refined to have
a narrower focus

Developing
Hypotheses
Example
 Hypothesis: People who ate at the church picnic were more likely
to become ill
 Exposure is eating at the church picnic
 Outcome is illness – this would need to be defined, for example, ill
persons are those who have diarrhea and fever
 Hypothesis: People who ate the egg salad at the church picnic
were more likely to have laboratory-confirmed Salmonella
 Exposure is eating egg salad at the church picnic
 Outcome is laboratory confirmation of Salmonella

Definition and
characteristics of
analytic studies
32

Analytic
Studies
 Analytic studies test
hypotheses about exposure-
outcome relationships
 Measure the association
between exposure and
outcome
 Include a comparison group

Analytic
Studies
Experimental
Studies
Randomized
Control
(Intervention)
Trials
Observational
Studies
Cohort
Case-control
Cross -
sectional
 Comparison group
 Inference about
exposure-outcome
relationship

CohortStudies
What is a cohort?
 A well-defined group of individuals who
share a common characteristic or
experience
 Example: Individuals born in the
same year
What are other examples of cohorts?

CohortStudy
(longitudinal
study, follow-up
study)
● Participants classified according to exposure
status and followed-up over time to ascertain
outcome
● Ensures temporality (exposure occurs before
observed outcome)
 Can be used to find multiple outcomes from a single exposure
 Appropriate for rare exposures or defined cohorts

CohortStudy
Design
Study Population
Exposed
Disease
Unexposed
No Disease
Disease
No Disease
Exposure is
self-selected
Follow
over time

Types of
CohortStudies
 Prospective cohort studies
 Group participants according to past or current exposure and
follow-up into the future to determine if outcome occurs
 Retrospective cohort studies
 At the time that the study is conducted, potential exposure
and outcomes have already occurred in the past

Prospective
CohortStudies
Study
Population
Exposed
Disease
Unexposed
No Disease
Disease
No Disease
Start of
study
(Present) (Future)

Retrospective
CohortStudies
Study
Population
Exposed
Disease
Unexposed
No Disease
Disease
No Disease
Start of
study
(Present)
(Past)

When to
Conduct a
CohortStudy?
 When the exposure is rare and the outcome is common
 Agricultural pesticide use and cancer events
 To learn about multiple outcomes due to a single exposure
 Health effects of a nuclear power plant accident

Analysis of
CohortStudies
Risk:
 Quantifies probability of experiencing the outcome of
interest in a given population
 Calculation: Number of new occurrences of outcome /
population at risk
Example:
 29 new cases of diabetes in a community
 100,000 people in the community at risk for diabetes
What is the risk of diabetes? 29/100,000

Risk Ratio
 Can also be called Relative Risk or RR
 Quantifies a population’s risk of disease from a particular
exposure
 Calculation:
= Risk in the exposed group / Risk in the unexposed group
 When using Person-Time as denominator
 In one group - Rate
 Comparison of two rate – Rate ratio

RRStrength
Scales
RR Strength RR
0.71 – 0.99 Weak 1.01 – 1.50
0.41 – 0.70 Moderate 1.51 – 3.00
0.00 – 0.40 Very strong >3.00
OlecknoWA. Essential epidemiology: principles and applications. Prospect
Heights, IL 2002;108.

2x2Tables
Typical set-up:
Cases Non-cases
Exposed a b
Not Exposed c d
Exposed Not Exposed
Cases a b
Non-cases c d
Alternate set-up:
RR = a/(a+c)
b/(b+d)
RR = a/(a+b)
c/(c+d)

Example:
Risk Ratio
Descriptive and Analytic Studies
500 150,000 200,000
200 780,000 800,000
700 930,000 1,000,000
Risk (Smokers) = 500/200,000 = 0.0025 or 25/10,000
Risk (Non-smokers) = 200/800,000 = 0.00025 or 25/100,000
Risk Ratio = 0.0025/0.00025 = 10
Interpretation: The one-year risk of lung cancer among those who
smoke is 10 times the risk among those who do not smoke.
Smokers Lung Cancer No Lung Cancer Total
Yes
No
Total
47

Example:
Risk Ratio
Person time
 NHANES – Follow-up Study (male diabetics subset)
 Original enrollment 1971- 1975
 Follow-up 1982 – 1984
 Complete follow-up on:
 Mortality Rate Ratio: 100/1414.7 ÷ 811/28,029.8 = 70.7/1000 ÷
28.9/1000= 2.5
48

Practice:
Risk Ratio
Example: In a given community…
 19 new cases of diabetes among obese individuals
 10 new cases of diabetes among non-obese
individuals
 100,000 people in the community at-risk for
diabetes (no one already has diabetes)
 25% of the community are estimated to be obese
What impact does obesity have on the risk of diabetes?

Practice:
Risk Ratio
19 24,981 25,000
10 74,990 75,000
29 99,971 100,000
Risk (obese) = 19/25,000 = 0.00076 or 76/100,000
Risk (non-obese) = 10/75,000 = 0.00013 or 13/100,000
Risk Ratio = .00076 / .00013 = 5.6
Interpretation: The risk of diabetes among those who are obese is 5.6
times the risk among those who are not obese.
Obese Diabetes No diabetes Total
Yes
No
Total
50

Example:
Cohort study
Diabetes RiskAmong Overweight and Obese Metabolically HealthyYoung Adults

Case-Control
Study
Purpose:
 To study rare diseases (when <5% of population experience
the outcome)
 To study multiple exposures that may be related to a single
outcome
Study Subjects
 Participants selected based on outcome status:
 Case-subjects have outcome of interest
 Control-subjects do not have outcome of interest

Case-Control
Study Design
53
Source
Population
Cases (Diseased)
Exposed
Unexposed
Exposed
Unexposed
Identify
cases
and
select
controls
Assess
exposure
history
Controls
(No Disease)

When to
Conduct a
Case-Control
Study
 The outcome of interest is rare
 Multiple exposures may be associated with a single outcome
 Funding or time is limited

Case-Control
Study:
Analysis
Format
Exposure Cases Controls
Yes a b
No c d
Exposure odds ratio (OR) ≈ RR when disease is rare
Odds of being a case among the exposed = a/b
Odds of being a case among the unexposed = c/d
Exposure odds ratio = (a/b)/(c/d) = (a*d)/(b*c)
(Cross-product ratio)

ExampleOdds
Ratio
Hospital birth Fistula No Fistula
Yes 30 150
No 70 250
Odds of fistula among the exposed = 30/150 = 0.20
Odds of fistula among the unexposed = 70/250 = 0.28
Odds ratio = (30/150)/(70/250) = (30*250)/(70*150) = 0.71

Practice
Odds Ratio
Mine Worker Lead Poisoning No Lead
Poisoning
Yes 42 247
No 107 825
Odds of being a case among the exposed = ?
Odds of being a case among the unexposed = ?
Odds ratio
= (a/b)/(c/d) = ?

Practice
Odds Ratio
Mine Worker Lead Poisoning No Lead
Poisoning
Yes 42 247
No 107 825
Odds of being a case among the exposed = 42/247 = 0.170
Odds of being a case among the unexposed = 107/825 = 0.130
Odds ratio = [(a/b) / (c/d)] = [(42/247) / (107/825)] = 1.3
Interpretation: Mine workers had 1.3 times the odds of developing lead
poisoning than did people who did not work in a mine.

Prevalence
Ratio &
Prevalence
Odds Ratio
 Chronic disease – date of onset is unknown
 Measure prevalence rather than incidence
 RR -> PR (prevalence ratio)
 OR -> POR (prevalence odds ratio)

Prevalence
Ratio
 Chronic disease – date of onset is unknown
 Measure prevalence rather than incidence
 RR -> PR (prevalence ratio)
 OR -> POR (prevalence odds ratio)

Prevalence
Odds Ratio
 Usually from a cross-sectional study
 Similar to odds ratio from case control study
 Calculated same way as odds ratio:
 POR = a*d /c*b

Prevalence
Ratio &
Prevalence
Odds Ratio
 Prevalence of breast cysts among ever users = 124/3247 = .038
 Prevalence of breast cysts among never-users = 77/2644 = .029
 Prevalence ratio = .038/.029 = 1.3
 Prevalence odds ratio = 124 * 2557 / 3123 * 77
 = 1.3

Example:
Case-control
study
Colorectal cancer: A case control study of dietary factors

Practice
exercise 1
 Background:
 NCDs such as type 2 diabetes are poorly understood and under-
prioritized in many low-to-middle income countries.
 You want to determine the risk of type 2 diabetes associated with
cardiovascular risk factors such as obesity and abdominal fat mass
in your country.
 Questions:
 What type of study would you conduct and why?
 What is the measure of association to calculate for this study?

Practice
exercise 2
 Background:
 The prevalence of prostate cancer has increased in your country
over the last 5 years.
 You want to examine the association between calcium intake and
prostate cancer risk.
 You have limited time and funding to conduct this study.
 Questions:

Practice
exercise 3
 Background:
 Cardiovascular disease (CVD) is of growing concern; however your
country has no recent data on the burden of this disease.
 You want to estimate the burden of cardiovascular disease in the
two main cities in your country.
 Questions:

Discussion
Question
Why do we use sampling?
 Can’t get information on everyone in a population
 Efficiently gets information on a large population
 Obtains a representative sample of a population
68

Sampling
Methods
Two main types of sampling methods:
 Probability sampling
 Non-probability sampling

Probability
Sampling
What are types of probability-based samples?
 Simple random sampling
 Systematic random sampling
 Stratified random sampling
 Cluster sampling

Simple
Random
Sample
 Principle
 Equal chance/probability of drawing each unit
 Procedure
 List all units (persons) in a population
 Assign a number to each unit
 Randomly select units
71

Method:
Simple
Random
Sampling
Each unit has the same
probability of selection
(1/48)

Example:
Simple
Random
Sample
Example: Calculate the prevalence of tooth decay among
1200 children attending a school
(sample size =100)
 List all children attending the school
 Each child assigned a number from 1 to 1200
 Randomly select 100 numbers between 1 and 1200

Advantages &
Disadvantages:
Simple Random
Sample
 Advantages
 Simple!
 Disadvantages
 Need complete list of units
 Units may be scattered and poorly accessible
74

Systematic
Random
Sample
 Principle
 Select sample at regular intervals based on sampling
fraction
 Procedure
 Assign a number to each unit
 Calculate sampling fraction (population size ÷ sample
size)
 Select first unit at random based on sampling fraction
 Subsequent units are chosen at equal intervals
75

Example:
Systematic
Random
Sample
1200 children attending a school
(sample size =100)
 Each child assigned a number from 1 to 1200
 Sampling fraction =1200/100 = 12
 Randomly select a number between 1 and 12
 Example: 8
 Select every 12th child, starting with child #8
 Example: 8, 20, 32, 44…
76

Advantages &
Disadvantages:
Systematic
Random
Sample
 Advantages
 Simple
 Can be implemented easily without software
 Disadvantages
 Need complete list of units
 Periodicity
77

Stratified
Random
Sample
 Principle
 Select random samples from within homogeneous
subgroups (strata)
 Procedure
 Divide the units into groups (called strata)
 Assign a number to each unit within each stratum
 Select a random sample from each stratum
 Combine the strata samples to form the full sample
78

Method:
Stratified
Random
Sample
 Sampling frame divided
into groups (age, sex,
socioeconomic status)
 Units in each group
have the same
probability of selection,
but probability differs
between groups
Men Women
Probability:1/32 Probability: 1/16

Example:
Stratified
Random
Sample
1200 children attending a school, with equal representation of
males and females
(sample size =100)
 Divide the children into two groups
 540 males and 660 females
 Assign each child a number
 Males: 1 to 540
 Females 1 to 660
 Randomly select 50 males and 50 females

Advantages &
Disadvantages:
Stratified
Random
Sample
 Advantages
 Can get separate estimates from the whole population
and from individual strata
 Precision increased if less variability within strata than
between strata
 Disadvantages
 Can be difficult to identify strata

Class
Discussion
Question
What are some examples of strata that you might sample
within?
 Race/ethnicity
 Age group
 Gender
 Geographic location
 Socioeconomic status
 Smoking status
 Occupation
 Education
 Many possibilities!

ClusterSample
 Principle
 Select all units within randomly selected geographic
clusters
 Procedure
 Divide population into geographic groups (clusters)
 Assign a number to each cluster
 Randomly select clusters
 Sample all units within selected clusters OR select a
random sample of units within selected clusters

ClusterSample

Advantages &
Disadvantages:
ClusterSample
 Advantages
 List of sampling units not required
 More efficient for face-to-face interviews when units
are dispersed over a large area
 Disadvantages
 Loss of precision due to correlation within clusters
 This correlation needs to be taken into account in
sample size calculations and analysis (“design effect”)

Non-
probability
Sampling
 Probability of selection is unknown or zero
 Inexpensive
 Results not generalizable
 Results often biased
Common types of non-probability sampling:
 Convenience sampling
 Snowball sampling
 Voluntary sampling

Non-
probability
Sampling

Choosing a
Sampling
Method
Consider:
 Population to be studied
 Size/geographic distribution
 Availability of list of units
 Heterogeneity with respect to variable
 Level of precision required
 Resources available

Practice
exercise 4
 Background: You will choose a sampling method for each of the
following studies.
 Questions:
 What sampling method would you use for:
1. The cross-sectional study on CVD described in Practice Exercise
#3?Why?
2. A one-time survey of citizens’ attitudes toward smoking and
second-hand smoke in response to proposed legislation to
impose a ban on smoking in restaurants.Why?
3. Serosurvey of blood lead levels (or urinary arsenic levels) of
prisoners entering the nation’s largest prison (or pregnant
women entering the nation’s largest maternity ward) to
determine average level of exposure in the population.

Descriptive vs.
Analytic
Epidemiology
 Descriptive epidemiology deals with the questions:Who,
What,When, andWhere
 Types of studies
 Aggregate: Ecological study.
 Individual: Case study, case series, cross-sectional
study
 Analytic epidemiology deals with the remaining questions:
Why and How
 Types of studies
 Experimental: Randomized control trial
 Observational: Cohort, case-control, cross-sectional

Cohort vs.
Case-Control
Studies
Cohort Study Case-Control Study
Preferred Study Design
When…
Members are easily
identifiable
Members are easily accessible
Exposure is rare
There may be multiple
diseases involved
Identifying entire cohort
would be too costly or time
consuming
Accessing entire cohort would
be too costly or time
consuming
Illness is rare
There may be multiple
exposures involved
Study Group Exposed persons Persons with illness (cases)
Comparison Group Unexposed persons Persons without illness
(controls)

Sampling
Advantages
and
Disadvantages
Probability Sampling
Advantages
 Results are
generalizable
 Representative
Disadvantages
 Expensive
 Logistically difficult
 Time-intensive
Non-Probability Sampling
Advantages
 Easy
 Quick access to
certain groups
Disadvantages
 Pertains only to those
in the study
 Results are not
generalizable

Review
1. What are 3 types of descriptive studies?
Answer: Case study, case series, and cross-sectional.
Ecologic studies are also descriptive.
2. What are 3 analytic study designs?
Answer: Cohort, case-control, and cross-sectional (if
appropriately designed). Randomized controlled trials are
also analytic.

Review
3. In a cross-sectional study, information about the exposure is
collected _________ information is collected about the
outcome.
a. before
b. after
c. at the same as

Review
4. In a prospective cohort study design , information about the
exposure is collected _________ information is collected about
the outcome.
a. before
b. after
c. at the same as

Review
5. In a case-control study, participants are selected to participate
in the study on the basis of ______.
Answer: Outcome (or disease) status
6. A ________ sample is a selection of the larger population at
risk, chosen to efficiently represent the disease and exposure
experience of the larger population.The goal of this type of
sampling is to minimize bias; however this sample can take
more time and resources to conduct.
Answer: random

Review
7. If you conduct a case-control study, which ratio measure of
association can you calculate with the data?
Answer:The odds ratio, which is the odds of the outcome in
the exposed group divided by the odds of the outcome in
the unexposed group.
8. What are the units for risk ratio?
Answer:There are no units.This is a ratio measure of
association and divides the risk of the outcome among the
exposed over the risk of the outcome among the
unexposed.

Group activity
#1
 What is the burden of hypertension in Riyadh or KSA?
1. Do a literature review to find out the latest study published
for this question.
2. Describe the study methods and sampling method used
3. Could this study have been conducted in different or better
way?

Group activity
#2
 What are the risk factors of breast cancer in Riyadh or KSA?
for this question.
2. Describe the study methods and sampling adopted used
way?

Group activity
#3
 What is prevalence and factors associated with COPD in
Riyadh or KSA?
for this question.
2. Describe the study methods and sampling adopted used
way?

103
Thankyou!
Emailyourqueriestosarizwan1986@outlook.com

Noncommunicable diseases - descriptive & analytical studies

Recommandé

Recommandé

Contenu connexe

Tendances

Tendances (20)

En vedette

En vedette (11)

Similaire à Noncommunicable diseases - descriptive & analytical studies

Similaire à Noncommunicable diseases - descriptive & analytical studies (20)

Plus de Rizwan S A

Plus de Rizwan S A (20)

Dernier

Dernier (20)

Noncommunicable diseases - descriptive & analytical studies