02 April 2011

Unhealthful News 92 - A difference without a difference?

As regular readers of this blog and others of you who understand epidemiology know, throwing a variable into a statistical model just because you have it (and even worse, because you have it and it substantially changes your effect estimate) is, for most types of studies, bad practice.  Yet health researchers have been throwing in the "race" variable in a large portion of studies, usually without any good reason to believe that race was a confounder, nor even a theory as to which of the hundreds of variables that race is correlated with (and thus is a rough proxy for) might be a confounder.  Generally no one did anything with the race result, except use it to pretend they had interesting results when they really did not ("the study showed no elevated risk, but if we restrict the analysis to men whose skin is a light shade of brown, then we saw a statistically significant 37.24% increase").

Well good news for bad epidemiologists:  You are about to get a new variable.

Numerous news stories today reported that the "LGBT Community" is ignored by health researchers.  The recommendations, by the U.S. government panel looking at the topic, included making sure that studies collect data on L, G, B, T, etc. status, apparently regardless of whether it seems to matter for the matter being studied.  (T probably matters for most everything – at least the actual surgery or hormone-injection version, but the others not so much.)

Of course, these people are not really ignored – they are included in most all studies in proportion to their existence in the population.  And for the few health conditions where the specifics of someone's sexual behavior is critical, there are quite often studies of sexual minority groups, at least if they are believed to be at higher risk, and such studies would almost always collect data on LGB status.  There is currently an upsurge in attention to studying young gay people who are at higher risk of psychological problems and suicide – a lot of study of one specific way in which being different can be horribly stressful.

So how is it suggested they are ignored?  The news stories identified such problems as gay people are more likely to reach old age without children to help take care of them, which sometimes matters for health research.  Um, yeah, but the issue is not have children to take care of you, something that is far more widespread than LGBT status, and when it matters can be studied directly.  Someone's sexuality is a really lousy proxy for that. 

No doubt it is true, as the report said, that we do not know everything we might want to about the health experiences and risk factors that are different for LGBT people.  But we also do not know very much about how health outcomes and risks differ for people who are have two cats, prefer cold weather to warm, are fans of mixed martial arts, or scored much higher on the verbal portion of the SAT than the math.  All of these are minorities that I suspect are larger than the LGBT community, and might genuinely have some interesting health characteristics.  And like LGBT people, when we do come across something that seems different about a group like one of these, we tend to study it.  These other minorities only differ in that they are are not a socio-political special interest group.

It is not that I am irked that this special interest group has successfully demanded more special attention.  I am not even suggesting that they do not deserve it, nor that it might generate some interesting information.  But I do get annoyed by the notion that if something is different about someone and we are not actively focused on it, then we must be neglecting that person and should make sure to collect data on the characteristic.  Sometime a difference does not really make a difference, and we should be able to distinguish the relatively rare cases when it does.

Though if this means it is now open season for such claims, I call for more attention to be devoted to the minority who make it a practice to object to bad research and questionable methodology, and try to explain why those are wrong – we have been ignored for too long!  And they had better hurry, because with this post, 365 Days of Unhealthful News is more than 1/4 complete.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.