Module 13 Flashcards
Definition of confounding bias
A situation in which the effects of two processes are not separated. The distortion of the apparent effect of an exposure on the dz is brought about by the association of the exposure with other variable(s) that can influence the outcome
3 criteria of confounders
- Be a risk factor (or protective factor) for the dz
- Be associated with the exposure, independently of the dz
- Not be an intermediate step in the causal pathway between exposure and dz (or not be the result of the exposure)
What does a confounder do?
Confuses our conclusions about the relationship between an exposure and an outcome
It distorts the odds ratio or relative risk
Role of confounding
Should always be considered as a possible explanation for an observed association, particularly in observational studies.
Confounding may over or underestimate a true association
Unlike selection and information bias, one may prevent confounding at the design stage or control for confounding at the analysis stage of a study.
What are the three methods of controlling confounding in the design of the study?
- Restriction (all study designs)
- Randomization (clinical trials)
- Matching (case-control studies and exposure-based cohort studies)
More details on restriction
May prohibit variation of the confounder in the study groups.
-For example, restricting participants to a narrow age category can eliminate age as a cofounder
Provides complete control of known confounders
Cannot control for unknown confounders
Can be done in any study designs
Disadvantages of restriction
This limits generalizability (external validity) but often improves feasibility and focus
Impractical to restrict on a large number of factors
More details about randomization and confounding
Random assignment makes intervention and control groups look as similar as possible
Chance is the only factor that determines group assignment
Controls for both known and unknown cofounders
Guarantees that tx assignment is not based on pt prognostic factors
Works best with large samples
More details about matching and confounding
A strategy for controlling at both the design and analysis levels of a study
Commonly used in case-control and exposure-based cohort studies
Matches participants in the comparison and study groups according to the value of the suspected or known confounding variable to ensure equal distributions
Controls only known confounders (the variables on which participants were matched)
Why is matching done?
So that the control (or unexposed participant) has identical (or at least, very similar) values of the confounding variable as the case (or exposed participant)
What are common matching variables?
Age and sex
What are the two types of matching?
Individual
Frequency
Individual matching
The pairing of one or more controls to each individual case (or one or more unexposed individuals to each exposed participant) based on similarity in sex, race, or other variables
Frequency matching
The proportion or percentage of cases and controls (or exposed and not exposed) with particular characteristics is matched
Advantages of matching
There is direct control of potential confounders, on which you matched
Fewer participants are required than in unmatched studies of the same hypothesis
Disadvantages of matching
- Data collection is more complex.
-Costly because extensive searching and recordkeeping are required to find matches
Data analysis must take account of the matching
-Different methods of analysis must be done
The effect (on dz) of the matching variable cannot be estimated
Matching cannot be removed. You cannot unmatch later.
One can “overmatch”
What is overmatching?
Matching on a variable that is associated with the exposure or outcome but not both
-Not a true confounder
-The confidence interval is widened
Matching on a variable that is in the causal pathway between exposure and outcome
-Not a true confounder (It is a mediator or intervening variable, instead of a confounder)
-Biases the odds ratio or relative risk
Matching on too many variables
-Too many confounders
-Cost and time to find suitable participants meeting all the matching criteria
What are the two analysis strategies to control confounding (after all the data is in)
Stratified analysis or stratification
Multivariate modeling or multivariate techniques
What is the result of collecting data on potential confounding variables at the beginning of the study
It makes it possible to adjust for these potential confounders at the analysis level through stratification and multiple regression techniques.
Definition of stratified analysis or stratification
Analyses performed to evaluate the effect of an exposure on an outcome within homogenous categories or strata (levels) of the confounder
When is there positive confounding?
If the crude is stronger than the stratified
When is there negative confounding?
If the crude is weaker than the stratified
When is there no confounding?
If the crude = the stratified
10% rule
Used when comparing adjusted back to the crude number.
Add and subtract 10% of the odds ratio back to the odds ratio or relative risk of the crude number
-i.e., 10% of 1.63 = 0.163
-1.63 + 0.163 = 1.47
-1.63 - 0.163 = 1.79
-Set the lower and upper as boundaries
If the adjusted (i.e, stratified) ORs or RRs are less than the lower boundary or greater than the upper boundary, then it’s not due to random error
Means it could be a confounder.