Confounding occurs when another factor (or factors) influences the association of interest. This occurs when this other factor is associated with both the risk factor of interest and the outcome of interest. For example, if examining the association between alcohol consumption and lung cancer mortality, it might be that an association is found. However, smoking is a confounder. There is an association between smoking and alcohol consumption as people who tend to smoke also tend to drink more alcohol. There is also an association between smoking and lung cancer mortality, therefore, it is possible that there is no real association between alcohol consumption and lung cancer mortality and smoking is acting as a confounder. Failure to take into account or consider smoking when examining this association can lead to biased results – known as confounding bias. Age, gender and deprivation are frequently related to the prevalence of behavioural risk factors, and poor health and mortality are also associated with age, gender and deprivation. Therefore, any of these factors can act as confounders when examining the relationship between risk factors and poor health. Therefore, examining the relationship between two factors is not straightforward, and can be further complicated by effect modification and interaction.
Also see: Bias, Causality, Confidence Intervals, Effect Modification, Interaction and Small Numbers.