I’ve been meaning to blog on this editorial about ethnicity as a risk factor (or not!) for infection by Prof Leibovici in Clinical Microbiology and Infection for a while. The basic story is that “ethnicity” often falls out as associated with infection-related variables (e.g. colonisation or infection with resistant bacteria), but the editorial poses an important question: it is actually ethnicity, or something co-correlated with ethnicity (e.g. socio-economic status) that is the causal risk factor?

Firstly, what is ethnicity? How do we define it? Google says “the fact or state of belonging to a social group that has a common national or cultural tradition”, but I’m sure this isn’t the only valid definition.

An example of an ethnicity-related risk factor related to infection would be certain funeral practices performed by certain ethnic groups that increase the risk of acquiring Ebola virus. Whereas an example of “ethnicity” appearing to be a risk factor for an infection-related outcome is probably the link between community-associated MRSA and indigenous communities. (I say probably because there may be some cultural practices that pre-dispose MRSA in these communities, but the association is more likely to be driven by low-socioeconomic status which often = poor health = sub-standard housing = limited education opportunities = inadequate access to healthcare etc).

I have my own recent experience of grappling with this issue, in a paper published recently in CMI. In this study, we identified individual and community-level variables that were associated with ESBL carriage at the time of admission. Sure enough, in the analysis of individual-level associations with ESBL carriage including all patients in the study, certain ethnic groups were indeed associated with ESBL carriage. However, on further analysis, these same ethnic groups were correlated with other variables that may be more important and actually causal in influencing ESBL carriage rate (such as overseas travel and household overcrowding – although see the followup Reflections critique on the household overcrowding issue!).

The editorial sets out a sensible set of steps for handling ethnicity as a possible risk factor for infection-related variables:

• Clearly define what you mean by ethnicity. Write this definition in your study protocol – and stick to it!
• Start with a causal model. Is it biological plausible that ethnicity could be a risk factor for the infection-related outcome you’re interested in? If the answer is no, then you shouldn’t really include it as a variable in a risk factor analysis.
• Then, collect the right data to test whether ethnicity is a causal risk factor. If you don’t have access to the right data, then don’t include it as a variable.
• Finally, configure your statistical approach and methodology around your causal model.

It was helpful for me to go through this process of being a bit more careful in managing ‘ethnicity’ as a variable to avoid potentially spurious (and far reaching) conclusions.