Epidemiology and causal inference · BEGINNER LESSON
Confounder
A confounder is a pre-exposure variable that is tied to both the exposure and the outcome in a way that can distort the estimated effect of the exposure. In causal analysis, it is usually defined by whether adjusting for it helps remove all confounding without relying on any smaller adjustment set. [0,1,2]
What it is
A confounder is a variable measured before the exposure being studied. It is important because it can make an exposure look more or less related to an outcome than it really is. In observational studies, people are not randomly assigned, so pre-existing differences can create this distortion. [0,1]
[1] On the definition of a confounder - PMC - NIH [2] Reference Guide on Epidemiology - Reference Manual on ...
How it works
If a factor influences both the exposure and the outcome, then the exposed and unexposed groups may differ for reasons other than the exposure itself. That shared factor can create confounding bias, meaning the observed association is partly due to the confounder. Formal causal inference treats this as a problem of blocking backdoor paths, which are routes from exposure to outcome that pass through common causes rather than through the exposure’s effect. [0,1,2]
[1] On the definition of a confounder - PMC - NIH [2] Reference Guide on Epidemiology - Reference Manual on ... [3] Confounding Fuels Misinterpretation in Human Genetics
How researchers handle it
Researchers try to account for confounders by stratifying, matching, restricting, or statistically adjusting for them. The formal literature emphasizes that a useful confounder is one that belongs to some minimally sufficient adjustment set: a set where adjusting for all its members makes the exposure-outcome effect unconfounded, and removing any one member would no longer be enough. [0,1,2]
[1] On the definition of a confounder - PMC - NIH [2] Reference Guide on Epidemiology - Reference Manual on ... [3] Confounding Fuels Misinterpretation in Human Genetics
Limitations and cautions
Not every pre-exposure variable associated with exposure and outcome should be treated as a confounder. Some variables may reduce bias without fully removing it, and some adjustment choices can even introduce bias if they involve the wrong variables. Confounding can also persist because relevant variables are unmeasured, because the data structure limits what can be separated, or because a study has residual confounding after attempted adjustment. [0,2]
[1] On the definition of a confounder - PMC - NIH [3] Confounding Fuels Misinterpretation in Human Genetics
SEE THE IDEA
How a confounder distorts an association
The factor whose effect is being studied.
A pre-exposure variable linked to both exposure and outcome.
The result or disease being measured.
Stratifying, matching, or conditioning to block distortion.
The confounder points to both exposure and outcome, creating a false or distorted link unless researchers adjust for it.
A confounder sits outside the main exposure-to-outcome pathway but can still change the observed association. [0,1,2]
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.