Lectures 11-12 Flashcards
Effect Modification
A 3rd variable that modifies the magnitude of effect of an association by varying it within different levels of that 3rd variable
What is difference in how you handle
- Confounder
- Effect modifier
- Control/adjust for variable
2. Evaluate and describe strata
(T/F) Confounder is associated with both the exposure and the outcome?
True
(T/F) Confounder can be in causal pathway between exposure and outcome.
False
Two effects confounder has on association between exposure and outcome?
- Change in Magnitude
2. Change in Direction
Residual Confounding
When confounder measured imperfectly so adjustment using this imperfect value does not completely remove effect
Testing for Confounding Steps
- Crude RR/OR
- Adjusted RR/OR
- If difference is greater than or equal to 20%, confounder is present
Testing for Effect Modifier Steps
- Crude RR/OR
- Crude RR/OR for each strata
- Difference between the highest and lowest strata is greater than or equal to 20%, variable is effect modifier
Ways to control confounding:
- Randomization
- Restriction
- Matching
- Stratification
- Multivariate Analysis
Randomization
Allocates equal number of subjects with known and unknown confounders in each intervention group
Randomization weaknesses
- Sample size may not be large enough to control ALL known and unknown confounders
- Doesn’t guarantee equal allocation for ALL confounders
- Interventional studied only
Restriction
Restrict specific subjects
Randomization Strengths
With large enough sample size, will likely serve purpose of making groups equal
Restriction strengths
Straight forward, convenient, inexpensive
Restriction weaknesses
- Difficult to enroll subjects
- If not narrow restriction may have residual confounders
- Eliminates evaluating levels of variable
- Generalizing
Matching
Study subjects selected in matched pairs related to confounder to equally distribute confounder among study groups
Matching Strengths
Intuitive; gives greater analytical efficiency
Matching weaknesses
Difficult to accomplish
Can only control confounders in matching
Over-matching masks findings
Stratification
Statistical analysis by evaluating association between the exposure disease within the various strata of confounding variable
Stratification strengths
Straight forward and enhances understanding of data
Stratification weaknesses
Impractical for simultaneous control of multiple confounders
Multivariate Analysis
Mathematically factor out effects of confounding variable(s)
Multivariate strengths
Simultaneous control of multiple confounders
Multivariate wekanesses
Some people may not understand data analysis
Confounding Variable
A 3rd variable that distorts an observed relationship between the exposure and the outcome
