Tuesday, May 20, 2014

Lies, Damn Lies, and STD Risk Statistics, Part 1

Every time I read anything about STD risks, I tend to get mightily annoyed at how difficult it is to get any useful information from the medical profession, at least in the popular press, about the actual magnitude of different types of risks. I remember talking about this problem in the case of cancer risks and smoking. Smoking causes cancer, living under power lines causes cancer, and eating burnt steak causes cancer, but they do not all cause cancer at anything like the same rate. Same thing with STDs. I sometimes find it hard to tell how much of this is because the people writing it are morons when it comes to causal inference, and how much is due to them knowing the right answer but spinning nonsense for public consumption, assuming that everyone is a child unable to make their own risk assessments. 

Let's hear from the Mayo Clinic, they're a famous hospital, surely they'll have top quality medical advice about what big ticket items to avoid. And their list of risk factors is ...(drumroll).... :
Having unprotected sex.
Having sexual contact with multiple partners.
Abusing alcohol or using recreational drugs.
Injecting drugs.
Being an adolescent female

The first thing you know is that what people mostly want to know are estimated treatment effects of particular actions. If I do X, my chance of an infection go up by Y%. Instead, what you get are a mish-mash of treatment effects, correlations with prevalence, correlations with transmission rates, and absolutely nothing on relative magnitudes, all leading to answers that are just laughable.

'Abusing alcohol or using recreational drugs' is hilariously stupid, because it doesn't map to anything directly. It could be correlation, it could be treatment, it could be both, who knows. They explain it as if it's mostly a treatment effect - "Substance abuse can inhibit your judgment, making you more willing to participate in risky behaviors.". In other words, the whole of their advice is that once you're drunk, you might do other stupid stuff. So just list that stuff! Of course, there's a strong correlation between people who get drunk all the time and people who do other stupid things. At a minimum, any treatment effects are going to be wildly heterogeneous. I'm pretty sure if your Aunty Gladys has a few too many sneaky shandies, the increase in her STD risk is zero. If you're a normally sensible person and you get drunk once, the chance of you picking up an STD are similarly low, because I'm guessing that most people will be unlikely to rush out and have anal sex with strangers just because they got drunk, though obviously some will. Most of the effect that makes this a risk factor has to be straight correlation with omitted factors, namely a tendency for reckless and risky behaviour. This is marginally actionable, if it tells you to avoid sleeping with perpetual drunks, but that's about it.

'Being an adolescent female' is even more stupid. The actionable interpretation of the previous statement was that perhaps we were being given correlations with overall prevalence. But how the hell do you interpret this one then? Do you really think that 'adolescent females' have high STD rates? Of course not. They may have higher transmission rates of certain diseases relating to cervical cancer, but this is a very different proposition. In what sane ordering is this among the five biggest STD risks for the general population to worry about? What adolescent females do have is a high rate of unplanned pregnancies, and it would be greatly in their interest to start using condoms regularly. So just say that! Stop trying to sell us a bunch of bull$#!& about how they also have massively high STD risks.

Since this post is already turning into a monster, I'll be back with Part 2 in a few days.

No comments:

Post a Comment