confounding

Question 1

explain the three factors of a confounder

Answer 1

1: is the potentential confounder a risk factor for the outcome by itself
2: does the level of exposure to the potential confounder vary by different exposure groups
3: the confounder is not the mechanism by which the exposure affects the risk of the outcome

Question 2

what are the 4 things a confounder can do

Answer 2

can over estimate an association
can under estimate an association
can create an association when there isn’t one
can change the direction of a true association - simpsons paradox

Question 3

whats simpsons paradox

Answer 3

when a confounder changes the direction of a true association

Question 4

what must we do if we are looking to control a potential confounder

Answer 4

plan ahead, if we dont measure it, it is difficult to do anything about it later on

Question 5

what does confounder control in the design phase aim to do

Answer 5

they all attempt to make the groups alike in regards to a potential confounder

Question 6

what are the three methods to control for confounders in design stage

Answer 6

randomisation, matching and restriction

Question 7

cons of doing randomisation

Answer 7

can only be used in an RCT!!! but it works best with a large sample, needs intention to treat analysis, also must have equipoise

Question 8

what is restriction and its pros

Answer 8

it is when we restrict the study participants to one stratum of a potential confounder,
easily done and can be applied to all study designs

Question 9

negatives of restriction

Answer 9

can reduce generalisability, can leave residual confounding, can reduce possible sample for the study, usually only applies to one confounder

Question 10

what is matching and when would we use it

Answer 10

matching is when choose people into the study to make the control/comparison group have the same composition as the case/exposed group regarding the potential confounder
usually only case controls

Question 11

what are the two types of matching we can do, explain them

Answer 11

individual, each case is matched with one or more controls who share the same confounding variables.
frequency matching is when we match the frequency of confounding variables in the control and comparison groups

Question 12

positives of matching

Answer 12

useful for difficult to measure/complex potential
confounders. can also improve efficiency of small sample size case control studies

Question 13

negatives of matching

Answer 13

individual matching can be difficult and limit the number of possible participants
it requires matched analysis, otherwise the MoA will be under reported.

Question 14

negative for controlling for confounder in the design phase

Answer 14

we cannot assess if it is truely a confounder or not

Question 15

what must we do inorder to be able to control for confounder and evaluate if it had an impact in the analysis phase

Answer 15

to control for and evaluate if the confounder had impact we must have measured the potential confounder.

Question 16

what is stratification

Answer 16

calculating the measure of association for each strata of a confounder then comparing them

Question 17

when the stratum specific MoA are similar but different to the crude MoA what does this mean

Answer 17

this means that confounding is present, thus we use the strata specific calculations

Question 18

when the stratum specific MoA are similar and similar to the crude MoA, what does this mean

Answer 18

this means confounding is not present and we can just use the crude MoA.

Question 19

positives of stratification

Answer 19

easy to apply for a small number of confounders with limited strata,
can evaluate confounding impact
can identify effect modification

Question 20

cons of stratification

Answer 20

can leave residual confounding.
no feasible when there is a large number of confounders with many strata

Question 21

what is multivariable analysis

Answer 21

statistical method for estimating measure of association whilst controlling for multiple potential confounders. uses regression techniques

Question 22

pros of multivariable analysis

Answer 22

can work in situations where stratification wont, its the most efficient method for controlling for controlling many confounders

Question 23

cons of multivariable analysis

Answer 23

nah g

Question 24

what is standardisation

Answer 24

comparing populations where structures of a confounder differ and making them the same in regards to that potential confounder, for example age standardisation

Question 25

strengths and weaknesses of standardisation

Answer 25

isnt great for many confounders with many strata. multivariable analysis is more efficient in analytical studies

Question 26

whats the general guideline for if confounding occurred or not

Answer 26

if the control for confounding alters the MoA by 10% or more. but use judgement if it has changed interpretation or not

Question 27

whats the magic phrase

Answer 27

confounding has resulted in the MoA being over or underestimated, as the crude MoA is further numerically upward/downward from the null

Question 28

whats effect modification

Answer 28

when the association differs across different strata of the potential confounder

Question 29

strength of controlling in the analysis phase

Answer 29

can asssess if it was a confounder or effect modifier
can assess how strongly it was a confounding variable.