1 - Case Control, and Longitudinal Cohort Studies Flashcards
Sample
The subset of the population that is chosen to be in the study
Cause
Exposure or Intervention
Effect
Disease or Outcome
2 x 2 table - A
Exposed + Outcome
2 x 2 table - B
Exposed + No Outcome
2 x 2 table - C
Not Exposed + Outcome
2 x 2 table - D
Not Exposed + No Outcome
Prospective Study
At the beginning, nobody has the outcome. Exposure happens, outcome happens. Cause is proven.
Ideal Experiment for Cause/Effect
All start out unexposed
Sample accurately reflects population
Random assignment to exposed/unexposed groups
None have the outcome at the beginning (prospective)
Outcome happens quickly and not rarely.
Exposure and outcome are not harmful
Interventional Studies
Researchers DO something to exposure group
Sample requires subjects/volunteers
First Step - Exclusion Criteria (Limit # of other causes, but may not reflect population)
Second Step - Randomization
Third Step - Test and monitor for some period of time
Result - Applicability to real life patients is limited
What can’t we examine with interventional trials?
Physiologic Phenomenon Social/Environmental Phenomenon Exposure is harmful, unethical Outcome is rare Outcome happens a lont time after exposure
Observational Studies
Cohort or Case Control
Cohort Study Types
Longitudinal (prospective, going forward, retrospective
Cohort Study
Start with exposure (selected/sampled based on their exposure)
Start without outcome
Groups “CHOOSE” what groups they are in. This means we see the effects of other exposures that “come with” the exposure in question. We must measure the confounding exposures.
Pros of a Cohort Study
Prospective (stronger relationship between cause & effect)
Real life populations
Multiple out comes can be studied
Cons of a Cohort study
People choose their own groups (lots of things “come with” the exposure)
Following people for a long time, lose people
Are people in one group more likely to drop out than others?
Measuring the effect of a Cohort Study - Questions
Is there a cause and effect relationship?
Use a Measure of Association
Cohort Studies - Risk
Risk of Disease (in Exposed)
(Exposed & Diseased)/(Total Exposed)
Risk of Disease (in Unexposed)
(Unexposed & Diseased)/(Total Unexposed)
Critiquing a Cohort Study - Questions
Is this amount clinically meaningful?
Could this difference be due to chance?
Could there have been other reasons that the exposed group had more outcomes?
Tools to determine how meaningful results of a cohort study are
Absolute Risk
Relative Risk (Risk Ratio)
Population Attributable Risk (Risk Difference)
Number Needed to Harm, Number Needed to Treat
Absolute Risk
AMOUNT of disease in the population
Relative Risk (Risk Ratio)
(Risk in Exposed)/(Risk in Unexposed)
RR = 1
There is no association of exposure to disease
RR
The exposure is protective against the outcome
RR > 1
The exposure is associated with the outcome
Attributable Risk (Risk Difference or Absolute Risk Increase)
The amount of the disease in the population that is “attributable” to the exposure
Attributable Risk Formula
Risk in Exposed - Risk in Unexposed
Number Needed to Harm
How many people would have to be exposed for 1 person to develop disease
Number Needed to Harm Formula
1/(Attributable Risk)
=1/(Risk in Exposed - Risk in Unexposed)
Problems with the Cohort Study
What else might “come with” exposure?
Who did we include in the study? What “comes with” our selection methods?
Who stayed in the study for the whole year?
Confounding Factor
Another factor that “comes with” the exposure and affects the outcome.
Can be controlled for
Bias
Problems with the structure of the study that affect who participates, how we measure, or how we analyze.
Can’t be controlled for
How do we control for confounding factors?
Stratification
Stratification
Separate out results into categories by confounding factors, see if there’s a difference.
Regression
Controlling for many possible confounders at the same time.
Selection Bias
Do we choose people in the exposed/unexposed differently and in a way that could be associated with the outcome?
Loss to Follow Up Bias (Attrition)
If people are less likely to stay in the study based on their exposure status, it can affect results
Limitations to Cohort Studies
Require a lot of people Require a long time Require a lot of money Need to have some idea there is an association before investing in it Very prone to confounding
Positives of a Cohort Study
Can study exposures that can’t be “done to” people
Can study multiple different kinds of outcome from an exposure
Can calculate RISK because exposure clearly came before outcome (prospective)
When do we use a Case-Control Study?
When a disease is RARE or takes a LONG TIME to develop (Cancers, etc)
Case-Control Study
Retrospective
Start with a group of people with the disease, and a group without the disease.
Look back to see if they had the exposure
Pros of a Case-Control Study
Quicker analysis
Fewer people required
Difficult part of Case-Control
Choosing Controls
Attempt to limit Bias and Confounding
Matching (by individual or by group)
Where do cases vs. controls come from? (Cases come from hospital/medical setting, does this generate bias or confounding?)
Case-Control Process - Were they exposed?
Collect data from past exposures (Medical records - was it recorded? Interviews - Memory?)
Needs to be at a point in time BEFORE DISEASE
Case Control - Measure of Association
Disease already exists, does not have temporal relationship of cause - effect
Can’t calculate risk
Can only calculate odds
Odds that a case was exposed
Exposed Cases / Unexposed Cases
Odds that a control was exposed
Exposed Controls / Unexposed Controles
Odds Ratio
(Exposed Cases x Unexposed Controls) / (Unexposed Cases x Exposed Controls)
OR = 1
No effect of the exposure on disease
OR > 1
The exposure has an association with disease
OR
The exposure is protective against disease
Problems with Case Control
Disease already exists
Have to look back and rely on memory/records
Bias in Case Control
Recall Bias
Selection Bias
Recall Bias
Those with disease are more likely to recall exposures than those without disease
Selection Bias
How controls are chosen compared to the cases
Prescription Bias
Confounding by Indication
For studies involving medications, the decision to give a certain medication can be affected by other factors that also affect the outcome
When can Odds estimate Risk
If disease is rare (less than 20% of population has the disease)
AND
The sample for the study is an unbiased representation of the population
Cross-Sectional Study
“Prevalence Study”
Define a population
Gather data on presence/absence of exposure and disease at that time for each individual.
Snapshot
