Confounding & Effect Measure Modification Flashcards
(23 cards)
What is confounding?
The association between the exposure and outcome is distorted due to a mixing of effects between the exposure, outcome, and a third extraneous variable (confounder)
Confounding can lead to incorrect conclusions about the relationship between exposure and outcome.
What are the three a priori criteria to assess whether a third variable is a confounder?
- Associated (non-causally or causally) with the exposure in the population that produced the cases
- A causal risk factor (or surrogate/proxy measure) of the disease in the unexposed cohort
- Not an intermediate factor on the causal pathway between exposure and outcome
These criteria help identify potential confounders in epidemiological studies.
True or False: Confounding can only inflate the measure of association.
False
Confounding can be positive (inflate the measure of association) or negative (bring the measure of association towards the null).
Fill in the blank: Confounding can be _______ or negative.
positive
Positive confounding inflates the measure of association, while negative confounding brings it closer to the null.
What is a confounder?
A third extraneous variable that distorts the association between exposure and outcome
Identifying confounders is crucial for accurate epidemiological analysis.
What are the methods to control or adjust for confounding during the design stage?
Randomization, Restriction, Matching
These methods help to ensure that the groups being compared are similar except for the exposure of interest.
What methods can be used during the analysis stage to control or adjust for confounding?
Stratified analysis, Standardization, Multivariate analyses
These techniques help to account for confounding variables when interpreting the results.
What is the definition of Effect Measure Modification?
When the effect of an exposure on an outcome differs across (or is modified by) values/levels of a third variable (effect modifier)
This indicates that the relationship between exposure and outcome is not uniform across different conditions.
What are some examples of effect modifiers?
age, sex/gender, genetic factors, SES, comorbid conditions, environmental exposure
How do you identify effect modification?
Use a stratified analysis or include interaction terms in regression models
If there is confounding, the association is ________ across strata by ________ from crude estimate
similar; differs
In effect modification, the association is ____________ across strata.
Different (also known as heterogenous effect)
What is the first step in identifying confounding or effect modification using the collapsibility method?
Draw a 2 x 2 table
This table is essential for organizing data and calculating associations.
What must be calculated after drawing the 2 x 2 table to determine whether confounding or effect modification is present?
STep 2: Calculate the crude measure of association
This measure provides the initial association before stratification.
What is the next step after calculating the crude measure of association to assess for confounding/effect modification?
Step 3: Break the 2 x 2 table up into stratum-specific 2 x 2 tables
Each table corresponds to a category or level of a third variable.
What is an example of creating stratum-specific tables?
Make 2 tables for males and females or 3 tables for youth, middle-aged, and elderly
This allows for assessing the effect measure modifier.
What should be calculated for each stratum-specific table?
Step 4: Calculate the stratum-specific measures of association
This involves obtaining separate measures for each group.
What is the next step after calculating stratum-specific (step 4) measures?
Step 5: Compare the stratum-specific measures to each other and to the crude measure
This comparison can be visual or statistical.
What indicates that effect modification is present?
If stratum-specific estimates differ from each other
This shows that the effect of the exposure varies by the third variable.
What indicates that confounding is present?
If stratum-specific estimates are the same but differ from the crude estimate
This suggests that a third variable is obscuring the true association.
Note on some challenging cases of determining confounding or effect modification
*Note: these two stratum-specific ORs (ssORs) may not be equal to each other in some cases where confounding is present. A rule of thumb: if ssORs are both greater than, or both less than the crude OR then you would surmise that there is effect measure modification as WELL as confounding. If they are on opposite “sides” of the crude OR, then you know that there is effect modification, but you cannot be sure on confounding.
