PowerPoint 簡報

Cross-Sectional Studies
Narges Khanjani, MD, PhD, Fellowship in Environ Epi
Research
methods
Observational
Descriptive
Case series,
case reports,
CS, cohort
Experimental
Analytical
Ecological
Crosssectional
Controlled
Cohort
Case control
Uncontrolled
Definition


A cross-sectional studies

a type of observational or descriptive
study

the research has no control over the
exposure of interest (e.q. diet).
It involves

identifying a defined population at a
particular point in time

measuring a range of variables on an
individual basis
Definition


Cross-sectional studies are studies of prevalence.
Proportion with an attribute or disease / Number of
subjects = Prevalence.

a type of observational or descriptive study

the research has no control over the exposure of interest (e.q. diet).
3 important questions to consider:

Definition of Case

Definition of the Population

Are cases and non-cases from an unbiased sample of the population?
Definition

“Snapshot Studies” (Paffenbarger, 1988)

Observations at a single hypothetical point
in time

Each subject assessed once at point in time.

Point Prevalence Studies
Definition

also called a Prevalence survey

A study that is quick and inexpensive to
complete.

Designed to determine “ what is happening ?
right now”
Basic features

“Snapshot” of a population, a “still life”

Assesses both the exposure and outcome
simultaneously, at a single point in time

Calculates prevalence, but not incidence

A study that is quick and inexpensive to
complete.

The first step in testing associations
Uses

Prevalence used in planning

Individual: Pre-treament probability for Dx

Population: Health care services
Examine associations among variables
 Hypothesis generating for causal links

Uses

Identify and describe a problem

Collect information for planning e.g. surveys of
immunisation, antenatal care, coverage

Evaluate utilisation rates of services

Monitoring health status of a community by
regular repeated surveys
Uses

Hypothesis generating for causal links

Method of Difference: If frequency of a disease is
markedly different between two groups then it is likely to
be caused by a particular factor that differs between
them.

Method of Agreement: If a factor commonly occurs in
which a disease occurs with high frequency then the
factor is very likely associated with the disease.

Concomitant variation: Frequency of a factor varies in
proportion to frequency of disease.
Uses
Prevalence survey: The studies are commonly
used to describe the burden of disease in the
community and its distribution.
Describe population characteristics: They are
also commonly used to describe population
characteristics, often in terms of person (who?)
and place (where?)



The British National Diet and Nutrition Survey

Nutrition and Health Survey in Taiwan

To describe various age groups in the population in
terms of food and nutrient intake and range of other
personal and lifestyle characteristics.

Migrant study : Some migrant studies may
full into the classification of cross-sectional
studies. These studies give clues as to
association between genetic background and
environmental exposures on the risk of
disease.
 e.q.
A study of the prevalence (percentage) of
coronary heart disease
 among
men of Japanese ancestry living in Japan,
Honolulu and the San Francisco Bay area
 showed the highest rates among those who had
migrated to the United States.

KAP (knowledges, attitudes, and practices )
study:



KAP studies are purely descriptive and help to build
up a better understanding of the behavior of the
population, without necessarily relating this to any
disease or health outcome.
Management tool: health service managers
and planners may make use of cross-sectional
survey to assess utilization and effectiveness
of service.
Development of hypothesis: Hypotheses on
the causes of disease may be developed using
data from cross-sectional study survey.
Design of cross-sectional
survey

The problem to be studied must be clearly
described and a thorough literature review
undertaken before starting the data collection.

Specific objectives need to be formulated.

The information has to be collected and data
collection techniques need to be decided.

Sampling is a particularly important issue to
ensure that the objectives can be met in the
most efficient way.

Fieldwork needs planning:

Who is available to collect the data ?

Do they need training ?

If more than one is to collect the data then it is
necessary to assess between-observer variation.

The collection, coding and entry of data need
planning.

A pilot study is essential to test the proposed
methods and make any alternations as
necessary.
Measure: Prevalence

Measure exposure and outcome variables at one point in time.

Main outcome measure is prevalence
P = Number of people with disease x at time t
Number of people at risk for disease x at time t
Prevalence=k x Incidence x Duration
Measure: Prevalence
Example:
RQ: What is the prevalence of chronic pain after hernia surgery?
Exposure of interest: Hernia surgery
Outcome of interest: Chronic pain (lasting for more than 3 months)
Methods: questionnaire survey
Sample:
All patients who had a hernia procedure between 1995-1997
n=350
Results:
Period prevalence chronic pain = 30% (CI 95% 24 - 36%)
Point prevalence chronic pain = 25% (on day of survey)
Interpretation

Measures prevalence – if incidence is our real
interest, prevalence is often not a good surrogate
measure

Studies only “survivors” and “stayers”

May be difficult to determine whether a “cause”
came before an “effect” (exception: genetic
factors)
Study Design
Disease
(Outcome)
_
+
Exposure
(Risk Factor)
+
_
Things to consider when designing a cross-sectional
study (survey)

What is your research question?

Is the design appropriate for your study?

Who are you going to study?

How are you going to obtain your sample?


Everyone who is eligible should have an equal chance of being invited to take
part
Is there a risk of ‘selection bias’?

E.g. taking people attending a specialist clinic; might not be ‘representative’ of
all patients with that condition

Selection bias is a threat

How you will collect your exposure/outcome data

Think about analysis (proportion %, denominator)
Things to consider when designing a crosssectional study (survey)

In Cross-sectional studies think of:

Sampling Procedures.

Clear definition of Target Population.

Clear definition of outcome.

Clear definition of risk factors.

Remember Confounders.
Sampling
Sampling

A sample is a subset of the population

Can be random or non-random; can be representative or nonrepresentative

Different types of sampling

This is major challenge when doing cross-sectional studies
Methods for collecting data

face to face interview

mail questionnaire

telephone interview

Self-administrated questionnaire

Medical examination

Laboratory test
Issues in collecting data

To sure what data shall be obtained

To sure which index will be used

Methods for collecting data

Criteria of disease diagnosis

Definition of variables

Training investigators
Variable assessment in cross-sectional studies

assessment methods for cross-sectional studies

Measures an individual’s intake at one point in time.

Does not require long-term follow up or repeat
measures

Valid

Reproducible

Suitable

Cost within study budget
Dietary method application

Food records using household measures have been
used in cross-sectional studies.

The recall method attempts to quantify diet over a
defined period in the past usually 24 hours.

The most commonly used dietary assessment method
which attempts to measure usual intake is the food
frequency questionnaire (FFQ).
Analysis

Before starting any formal analysis, the data should
be checked for any errors and outlines.

Obvious error must be corrected.

The records of outliners should be examined excluded

Checking normality of data distribution.

e.q. using the Kolmogorov-Smirnov Goodness of Fit Test.
Analysis

Descriptive analyses

Analysis of differences

Analysis of association / relationship

Multivariable analysis
Analysis



Standard descriptive statistics can then be used:
mean, median, quartiles, and mode; measure of
dispersion or variability such as : standard
deviation; measure precision such as: standard
error, and confidence intervals.
Mean can be compared using t-tests or analysis
of variance (ANOVA).
More complex multivariate analysis can be
carried out such as multiple and logistic
regression.
Analysis
DZ = Rash
(+)
Grape
95
(–)
88
183
35
43
(52%)
Tomato
8
(19%)
Prevalence ratio = 52%/19% = 2.6
Analysis

Instead of looking at a ratio of prevalences, we can also
look at a ratio of odds.

Odds are not intuitively appealing: they are the
likelihood of an event occurring divided by the
likelihood of the event not occurring.
Analysis
DZ = Rash
(+)
Grape
Tomato
95
8
-
88
35
183 PR= 95/183
------- =2.6
8/43
43
Odds of grape work in rash pts: 95/8=11.9
Odds of grape work in healthy: 88/35=2.5
Odds ratio=(95/8)/(88/35)=11.9/2.5=4.7
Bias
Selection Bias
Is study population representative of target population?
Measurement Bias
Outcome
 Misclassified (dead, misdiagnosed, undiagnosed)
 Length-biased sampling

Cases overrepresented if illness has long duration and are underrepresented if
short duration.(Prev = k x I x duration)
Risk Factor
 Recall bias
 Prevalence-incidence bias

RF affects disease duration not incidence
Bias
The selection bias classic for cross-sectional
studies is “the healthy worker effect.” I.e., only
“healthy workers” are available for study,
distorting your findings.
 Example: Low asthma rates in animal handlers
(because persons contracting asthma quit and
are not available for study).

Advantages

Quick, cheap

Easy to obtain prevalence

Outcome

Exposure
Disadvantages

Prone to selection bias

Recall bias

Cannot measure disease onset

Problem of temporality (not a
problem if exposure is constant)

Not suitable for rare disease
Limitation of cross-sectional study

It is not possible to say exposure or
disease/outcome is cause and which effect

Confounding factors may not be equally
distributed between the groups being
compared and this unequal distribution may
lead to bias and subsequent misinterpretation.

Cross-sectional studies within dietary survey,
may measure current diet in a group of people
with a disease. Current diet may be altered by
the presence of disease.

A further limitation of cross-sectional studies
may be due to errors in reporting of the
exposure and possibly outcome.