However, it may not be possible to identify and measure all possible confounders in an observational dataset. Only by considering all possible confounders and adjusting for them (by study design or analysis), can we confidently claim that joint trauma causes knee osteoarthritis (that’s if we can also rule out other biases). Causal diagram illustrating a distorted association between joint trauma and osteoarthritis through failure to control for the confounding factor, exposure to high impact sport. In this case, if the confounder (high impact sport) is not controlled for in the data analysis, we will observe a distorted association between joint trauma and knee osteoarthritis (Figure 2).įigure 2. That is, individuals involved in high impact sport may be more susceptible to both acute joint trauma and chronic knee osteoarthritis (through repeated use). In our example, it is plausible that joint trauma and knee osteoarthritis share a common cause – high impact sport (the confounder).

Causal diagram illustrating the structure of confounding. Failure to control for the confounder makes it appear that there is an association between the exposure and the outcome, whereas in fact both are caused by the confounder and are not related to each other at all (or as strongly).įigure 1. Only if we can rule out these biases can we begin to think that association may imply causation.Ĭonfounding occurs when an exposure and an outcome share a common cause (the confounder Figure 1). To claim that this association represents a causal effect, we need to first rule out two possible issues that lead to a non-causal association: Let’s say that, in this dataset, we detect an association between joint trauma and knee osteoarthritis. Because it would be unethical to conduct an experiment whereby we deliberately inflict joint trauma to assess its effects on chronic knee osteoarthritis, we decide to tackle this question by using observational data from a hospital registry. Suppose that we want to know if acute trauma to a joint (an exposure) causes chronic knee osteoarthritis (an outcome). Hence the mantra: “association is not causation.” It does not necessarily imply that one causes the other. Hopin Lee, Jeffrey K Aronson and David Nunan blog about how to tell when an association does and does not mean causation in health researchĪ statistical association between two variables merely implies that knowing the value of one variable provides information about the value of the other.

Neyo do you ever