Module 7 Flashcards
Cross-sectional study- how to best look at it
Snapshot of the population you wish to learn about
A small replica of all parts of the population: often proportional to size
Key features of cross-sectional studies
-A type of prevalence study
-Single period of observation. No f/u.
-Exposure and dz data collected simultaneously.
-Exposure and dz measures obtained at the individual level.
-Both probability and non-probability sampling used.
-Best done as a probability sample.
Characteristics of a cross-sectional survey
-Conducted over a short period of time
—For small studies- a few days or weeks
—For large studies with thousands of participants- a few years
-The unit of analysis is the individual
-There is no f/u period.
a in the cross-section table
Was exposed and has the dz
b in the cross-section table
Was exposed, did not get the dz
c in the cross-section table
Was not exposed, got the dz
d in the cross-section table
Was not exposed, did not get the dz
Familiar examples of cross-sectional studies
Opinion polls of all kinds:
-All political polls
-Telephone surveys
-Surveys done by companies about their products
Definition of serial surveys
Cross-sectional surveys that are routinely (repeatedly) conducted
Each serial survey recruits different people in each cycle
Examples of serial surveys
Behavioral Risk Factor Surveillance System
National Health Interview Survey
National Hospital Discharge Survey
National Health and Nutrition Examination Survey
Probability sample definition
Every element in the population has a quantifiable probability of being included in the sample
Types of probability samples
Simple random sample
Systematic random sample
Stratified probability sample
Non-probability sample definition
It is impossible to quantify the probability of being included in the sample
What is a type of sample that is a non-probability sample?
Convenience sample
Simple random sample characteristics
Everyone in the sources pop has an equal chance of getting into the study
Participants are chosen completely at random
Systematic random sample characteristics
Like a simple random sample, but the investigators choose every nth person to invite into the study, from a list, called the sample frame, representing the source population
Stratified probability sample characteristics
Identify subgroups in the pop, such as racial/ethnic groups
And take a random sample of each subgroup
Most large surveys, like NHANES are this type
Characteristics of convenience sample
Not a probability sample
Only interview whomever you can get
Not much scientific validity
Cross-sectional studies collect only which type of data?
Prevalence data
Uses of prevalence data
Describe health burden of a pop
Describe status of dz in a pop
Estimate the frequ (prevalence) of exposure
Project healthcare needs of affected individuals
Can cross-sectional studies be both descriptive and analytic?
Yes
Type of question addressed in a descriptive cross-sectional study?
What is the prevalence of disease x or exposure y in a given pop?
Type of question addressed in an analytic cross-sectional study
Is exposure A associated with outcome B?
Characteristics of descriptive cross-sectional studies
Have no “a priori” hypothesis
Person, place, and time
Often exist just to discover or describe the prevalence of various factors in the pop
Characteristics of analytic cross-sectional studies
Answer a research question about the association between an exposure and outcome variable
There is an “a priori” hypothesis.
Target pop definition
The pop about which information is desired
Source pop definition
The larger pop from which the sample was drawn
All members of the source pop had some opportunity to be in the study
Sample or study pop definition
The pop that is actually included in the analyses
What analysis is used when using categorical data?
Prevalence ratios
Prevalence odds ratios
a + c in the 2x2 table
all with the outcome
b + d in the 2x2 table
all without the outcome
a + b in the 2x2 table
all exposed
c + d in the 2x2 table
all not exposed
Prevalence ratios definition
The ratios of the prevalence of dz in the exposed compared to the prevalence of dz in the non-exposed
Prevalence ratio formula
(a/(a + b)/ (c//c+d))
What does prevalence ratio (PR) = 1 indicate?
The outcome is equally common in those who have and do not have the exposure (implies no relationship between exposure and outcome)
What does PR>1 indicate (assuming statistical significance?)
Means the outcome is more common in those who have the exposure (possible risk factor)
What does PR <1 indicate (assuming statistical signficance?)
Means the outcome is less common in those who have the exposure (possible protective factor)
Example of wording for a precise interpretation of a prevalence ratio
The prevalence of the dz (or outcome) in the exposed was XX times the prevalence of the dz in the not exposed
What can cross-sectional studies generate?
Hypotheses
What is one limitation to an analytic cross-sectional study?
Time sequence impossible to establish with certainty
-There has been no actual lapse of time between measurement of exposure and dz
-This presents difficulties in distinguishing the exposures from the outcomes of the dz
Advantages of cross-sectional studies
-Can estimate the prevalence of a health problem, which can indicate its magnitude and distribution, a source of descriptive statistics.
-Useful for program planning and justification
-Quicker and cheaper than some other study designs
-Involve the collection of data on individuals, which can later be aggregated, if desired.
-May generate new etiologic hypotheses for further study
Disadvantages of cross-sectional studies
-Do not provide incidence data
-The impact of quickly emerging dzs may be missed
-Not efficient for the study of rare dzs or rare exposures
-Cannot determine time sequence of exposure and dz, and therefore, one cannot say that one variable “caused” another
-Prevalent cases of lethal dzs are all survivors
What are two synonyms for case-control studies?
Retrospective study
Case-comparison study
Can case-control studies be both descriptive and analytic?
No, just analytic
Definition of analytic studies for this module
Attempts to identify causes or risk factors that explain health-related states or events, and tests specific “a priori” hypotheses often developed in descriptive studies.
There is a comparison of some kinds.
There is at least one exposure and one outcome.
Distinguishing features of case-control studies
-Individuals with a particular dz or condition are selected for comparison with individuals in whom the dz or condition is absent.
-Looks for an association by comparing h/o exposure between a group of diseased individuals and a group of non-diseased individuals.
-Two separate samples
-Dz status is known at the beginning
-Retrospective
What is another characteristic of a case-control study?
The outcome is always identified prior to the exposure.
-Involves grouping people as cases and controls, and investigating whether the cases are more or less likely than controls to have had past experiences, lifestyle behaviors, or exposures.
From where can cases come?
Public health clinics
Physicians’ offices
Health maintenance organizations
Hospitals
Industrial and government sources (registries)
From where can data be obtained?
Medical records
Personal interviews
Telephone surveys
Written questionnaires
Physical examinations
Laboratory examinations or tests
What is the dependent or outcome variable in a case-control study?
Dz
Requirements about dz characteristics
-The dz spectrum should be known
-The case definition should be take into account the stage of dz and be established before data collection, as well as person, place, and time characteristics.
Requirements for controls
-Should look like the case participants with the exception of the dz
-Requires selection from the same source pop as the cases.
-Must be chosen independently of exposure status
3 main sources of controls
Population-based (best): Can be a probability (representative) sample of those without the dz from the same source pop as the cases.
Downside: expensive
Same hospital as cases but different dz
Downside: no healthy controls
Family, friends, neighbors, coworkers, relatives of the cases
Downside: may be too similar to the cases
How can bias occur with control selection?
- Selection of cases that are not representative
-Only those receiving medical attention
-Only cases that survive - Selecting controls that did not come from the same source pop as the cases
- Unknowingly selecting controls (or cases) on the basis of exposure
- Differential response rates among cases and controls, with the possibility that controls (or cases) who did not participate are systematically different than those who did.
Can data on more than one exposure be gathered in a case-control study?
Yes
How can information about exposure status be gathered?
Medical records
Interviews
Questionnaires
Surrogates such as spouses, parents, siblings, or employers
How can there be issues with recall biases?
Determining exposure status for cases and controls:
-Be aware of limitations in recall
-Both cases and controls my not recall accurately their past exposures
-Cases may have a vested interest in recalling more
Types of questions asked in case-control studies
Among the cases, is an attribute more or less common than among the controls
Is the mean level of factor Y greater among cases than among controls
A in the 2 x 2 case-control table
Exposed cases
B in the 2 x 2 case-control table
Exposed controls
C in the 2 x 2 case-control table
Unexposed cases
D in the 2 x 2 case-control table
Unexposed controls
A + C in the 2 x 2 case-control table
All cases
B + D in the 2x2 case- control table
All controls
A + B in the 2 x 2 case-control table
All exposed
C + D in the 2 x 2 case-control table
All unexposed
Cross product ratio:
AD/BC
The ratio of the odds of dz among the exposed to the odds of the dz among the unexposed
(A/B) / (C/D)
The ratios of the odds of exposure among the cases to the odds of exposure among the controls
(A/C) / (B/D)
What does OR = 1 indicate?
Implies no association
What does OR>1 indicate (assuming statistical significance)?
Suggests the exposure is a risk factor
What does OR <1 indicate (assuming statistical significance?)
Suggests a protective factor
Precise interpretation of odds ratio terminology
Cases with ______ had ____x the odds of _____ compared to the controls
Advantages of case-control studies
Useful for etiologic studies, to study the causes (determinants of a dz)
Tend to use smaller sample sizes than cross-sectional surveys or cohort studies
Are usually of short duration
Cost-effective
The only studies useful for studies of rare dzs or outcomes
Useful for studies of dzs with long latency periods where it is impractical to wait for the outcome to occur in healthy people
Many risk factors can be studied simultaneously
Disadvantages of case-control studies
Not useful for planning purposes
Cannot estimate dz frequency, such as prevalence or incidence
Difficult to study rare exposures
Can investigate only one dz outcome per study
Problem with possible recall bias
Problems with bias d/t the selection of the wrong cases or controls
Cannot establish sequence of events very well
-Therefore, not usually sufficient to establish cause-effect relationship between exposure and dz