Confounding I

Question 1

What is confounding?

Answer 1

Confounding is a mixing or muddling of effects when the relationship we are interested in is confused by the effect of something else - the confounder
Risk factors party together (e.g. alcohol and smoking, people tend to ‘socially’ smoke when they drink)

Question 2

Why is confounding a problem?

Answer 2

Confounding is a problem as it means the exposed and comparison group may not be equivalent hence analytic epidemiology is not valid as groups may not be equivalent for everything associated with the outcome

Question 3

What are three properties of a potential confounder?

Answer 3

1. Independently associated with the outcome - A risk (or protective) factor for the outcome regardless of exposure statues
2. Independently associated with the exposure - Imbalance in distribution across exposure groups
3. Not on the causal pathway - Not how the exposure affects the outcome

Question 4

What can confounding do?

Answer 4

Confounding can over estimate the true association (further away from the null), under estimate the true association (closer to the null), change the direction of the true association (risk factor becomes protective factor) or give appearance of an association when there is not one (go from null to something else)

Question 5

How can we identify any potential confounder?

Answer 5

Collect information on all potential confounder
Use literature to identity known and suspected risk factors for the outcome
Collect information on factors strongly associated with exposure, regardless if known risk factor
If you don’t measure it/plan ahead, difficult to do anything about it later

Question 6

What is the study design of confounding?

Answer 6

Design study to minimise confounding by selection and allocation of participants, three ways of doing this: randomisation, restriction and matching. All ways attempt to make groups being compared alike with regard to potential confounder.
Can’t assess association between potential confounder and outcome and can’t assess whether truly a confounder

Question 7

What is randomisation in regards to confounders?

Answer 7

Randomisation works best with large sample sizes, only can be used in RCT’s but can apply to known and unknown confounders (don’t need to identify potential confounders). Need to equipoise and need intention to treat analysis

Question 8

What is restriction in regards to confounders?

Answer 8

Restriction is where you restrict the study to one stratum/value of potential confounder, this is easy and be applied to all study designs however is can reduce generalisability, reduce number of potential participants, has potential for residual confounding with imprecisely measure confounders and usually only has one potential confounder

Question 9

What is matching?

Answer 9

Matching is generally used in case-control studies, measure someone in the control groups potential confounder and then measure people in the comparison group who match the confounder

Question 10

What is individual and frequency matching?

Answer 10

Individual level matching: each case matched with one or more controls having the sam confounding variable characteristics e.g. match one 45 year old male to another
Frequency level matching: matching at aggregated level e.g. match some 45-50 year old males with other 45-50 year old males

Question 11

What are the positives and negatives of matching?

Answer 11

Matching is useful for difficult to measure/complex potential confounders and can improve efficiency of case-control studies with small numbers but individual matching can be difficult and limit number of potential participants and we need to use a special matched analysis for individual matching