It is associated with the exposure.

It is associated with the outcome.

It is not on the causal pathway between the exposure and outcome.

As an example, look at Figure 20-1, which illustrates a study of patients with Type II diabetes where

there is a hypothesized causal relationship between the exposure of having served in the military and

the negative outcome of having an amputation due to diabetic complications.

As shown Figure 20-1, inability to exercise and low income are both seen as potential confounders.

That is because they are associated with both the exposure of military service and the outcome of

amputation, and they are not on the causal pathway between military service and amputation. In other

words, what is causing the outcome of amputation is not also causing the patient’s inability to exercise,

nor is it also causing the patient to have low income. But whatever is causing the patient’s amputation

is also causing the patient’s retinopathy. That’s because Type II diabetes causes poor circulation,

which causes both retinopathy and amputation. This means that retinopathy and amputation are on the

same causal pathway, and retinopathy cannot be considered a potential confounder.

© John Wiley & Sons, Inc.

FIGURE 20-1: Example of how confounders are associated with exposure and outcome but are not on the causal pathway

between exposure and outcome.

Avoiding overloading

You may think that choosing what covariates belong in a regression model is easy. You just put all the

confounders and the exposure in as covariates and you’re done, right? Well, unfortunately, it’s not that

simple. Each time you add a covariate to a regression model, you increase the amount of error in the

model by some amount — no matter what covariate you choose to add. Although there is no official

maximum to the number of covariates in a model, it is possible to add so many covariates that the