Lecture 4: Confounding & Effect Modification

Question 1

the best population to compare with, is themselves minus the event

Answer 1

counterfactual theory

Question 2

counterfactual theory requires the assumption of?

Answer 2

exchangeability

Question 3

lack of exchangeability in a population results in ?

Answer 3

confounding or

we are unable to compare

Question 4

when we are assessing the associations of an exposure and outcome between two groups, what question are we asking ourselves regarding a true association?

Answer 4

are there alternative explanations to the association?

Question 5

researchers evaluate 3 aspects of their study, before declaring a true association

Answer 5

1. check for confounding variables
2. check for bias
3. check for statistical significance

Question 6

define ‘confounding variables’

Answer 6

a 3rd variable that distorts an association between the exposure and outcome

an alternative explanation

Question 7

what associations do confounding variables effect?

Answer 7

OR
HR
RR

Question 8

simple definition of confounding

Answer 8

another explanation that appears to be the cause of the outcome, but it is not

Question 9

in order for a variable to be considered a confounder, it must have 3 things

Answer 9

1. independent association with exposure
2. independent association with outcome
3. not directly in the causal-pathway

Question 10

what does it mean to ‘not directly be in the causal pathway’ ?

Answer 10

the confounding variable is not a stepping stone between exposure and outcome

it is independent of both

Question 11

a measure of association that ignores all other factors

Answer 11

CRUDE measure of association
or
unadjusted

Question 12

DAG

Answer 12

direct acyclic graph

Question 13

if confounding is present, how do you go about knowing if it is present?

Answer 13

you must test and calculate for confounding

Question 14

step 1 in testing for confounding

Answer 14

calculate crude RR

also called unadjusted RR

Question 15

step 2 in testing for confounding

Answer 15

calculate OR for each individual strata of the 3rd variable

Question 16

step 3 in testing for confounding

Answer 16

compare the crude vs. adjusted measures

Question 17

when comparing calculations, what tells you if the 3rd variable is a confounder?

Answer 17

compare crude vs. adjusted

if >15% different === confounder

Question 18

3 steps in testing for confounding

Answer 18

1. find crude RR
2. fund adjusted RR for each strata
3. compare crude vs. adjusted

Question 19

what are the 2 main impacts of confounding variables?

Answer 19

magnitude of association

direction of association

Question 20

what is the purpose of controlling for confounders?

Answer 20

to get a more accurate measure of association

Question 21

at what 2 points in a research study can you control for confounding?

Answer 21

study design stage

analysis of data stage

Question 22

what are 3 ways to control for confounding in the study design stage?

Answer 22

randomization
restriction of population
matching of participants in each group

Question 23

what are 2 ways to control for confounding in the analysis of data stage?

Answer 23

this is after data has been collected

1. stratification
2. multivariate statistical analysis

Question 24

2 important weaknesses of using randomization to control for confounders

Answer 24

must have a large sample size

practical only for interventional studies

Question 25

weaknesses of using restriction for confounder controlling.

Answer 25

narrow restriction criteria limits who can be a subject | can negatively impact generalizeability

Question 26

strengths of randomization

Answer 26

will large N, it is successful | stratified version precisely assures equal-ness

Question 27

strengths of restriction

Answer 27

convenient, cheap | does not negatively impact internal validity

Question 28

weaknesses associated with matching

Answer 28

difficult to do, time, expensive | only controls for confounders that are matched for

Question 29

strengths of matching

Answer 29

greater analytic efficiency

Question 30

strengths of stratification

Answer 30

straight forward, enhances understanding of data

Question 31

weakness of stratification

Answer 31

impractical for controlling multiple confounders, especially ones with multiple strata

Question 32

statistical analysis of data, evaluating associations between exposure and outcome within the various strata of a confounding variable

Answer 32

stratification | data stage confounder control mechanism

Question 33

statistical analysis of data by mathematically factoring out the effects of the confounding variables

Answer 33

multivariate analyses | regression stats

Question 34

multivariate weaknesses

Answer 34

requires researchers who really understand math | time consuming for statistician

Question 35

multivariate strengths

Answer 35

control for multiple confounders at once | ORs can be obtained and interpreted

Question 36

if you are interpreting an adjusted ratio, you must also include.....?

Answer 36

in the interpretation you must include what has been adjusted for

Question 37

a 3rd variable capable of changing the magnitude of effect, of a true association, by changing it within different strata of that variable

Answer 37

effect modification

Question 38

compare effect modification and confounding: relate to association between talc use and cancer in women: crude = 1.3 adjusted for race = 1.3

Answer 38

race is not a confounder (<15%) but if we check each race strata..... we find that not every race has the same talc use so race is an effect modification

Question 39

effect modification is vital detail that we want to ____ within a study.

Answer 39

describe

Question 40

3 steps in testing for effect modification

Answer 40

1. calculate crude 2. calc strata ratios 3. compare each of the strata

Question 41

step 1 in testing for effect modification

Answer 41

calculate crude RR between exposure and outcome

Question 42

step 2 in testing for effect modification

Answer 42

calculate RR for each specific strata within a variable ex. race -- white, black, latino, asian

Question 43

step 3 in testing for effect modification

Answer 43

compare the highest and lowest RR of the strata | >15% difference = effect modification is present