The professor asked the class what a “confounder” was. I raised my hand and answered, “A confounder is something that is associated with both the exposure and the disease, but is not in the causal pathway between an exposure and a disease.” That is how I learned what a confounder is. That is how I explain it to people.
The professor smiled and then mentioned that we needed to be more “nuanced” with the language we used to describe a confounder. He then proceeded to give us three rules for identifying a confounder.
So what was so different between my answer and the professor’s definition of a confounder?
I said “associated” instead of saying that a confounder “causes” the disease. I guess I should have said that the confounder is “causally associated” with the disease. However, I used “associated” because I was explaining the relationship between the confounder and the exposure and the disease. I wasn’t explaining the relationship of just the confounder and the disease, or just the confounder and the exposure.
The reason I’m writing about this is because this has been a continuing issue with the way that questions are asked by some professors, especially in the context of exams. (I’ve written before about my issues with how professors are not particularly good teachers.) When the instructions on an exam state that there is “one best correct answer,” then that implies that there are other answers which may be correct, though not to the satisfaction of the professor. It turns a question like “What is two plus two” into a riddle where we have to read the minds of the professor and wonder if the answer is “four,” “six minus two,” “one half of eight,” or “three.” It confuses things.
The real world is even more confusing. Consider a study from back in the day which stated that coffee causes pancreatic cancer. This is from The New York Times:
“A statistical link between the drinking of coffee and cancer of the pancreas, the fourth most common cause of cancer deaths among Americans, was reported yesterday by scientists of the Harvard School of Public Health. The discovery was unexpected, and its significance is not yet clear.
”If it reflects a causal relation between coffee drinking and pancreatic cancer,” the report said, ”coffee use might account for a substantial proportion of the cases of this disease in the United States.””
That’s from 1981. Here we are, almost 35 years later and Starbucks is booming selling the stuff. What gives? Isn’t coffee dangerous?
You see, the subjects of the study were being drawn from a GI (gastrointestinal) medical practice, so they were more likely to be in treatment for things like pancreatic cancer. Also, many of them were smokers, so they were more likely to have pancreatic cancer. When you “adjusted” (aka “controlled,” “accounted”) for smoking, it turns out that the smokers who drank coffee were the ones contributing to the elevated risk of cancer, not the non-smokers who drank coffee.
This is the same mistake (among many) that we saw Brian S. Hooker commit when he re-analyzed vaccine and autism data. He thought he saw an association between African American children at 36 months and vaccines and autism, when the actual association was only between their age and autism… Not their race nor their vaccine status. (Children are diagnosed the most at around that age.) Once those confounders are adjusted for, the association he saw disappears, and we get a picture closer to the truth. In his defense, Brian S. Hooker is quoted as saying that he prefers simpler statistics because he is “not really that smart.”
Simple statistics don’t always tell the whole truth, as you can see.
Anyway, back to language…
I’m not planning on going into academia, though I may — from time to time — give a lecture here and there on epidemiology and other stuff that interests me. Instead, I’m going into public health practice. I’m going to be using the evidence that exists, and is yet to exist, to address public health problems. A lot of that comes with using language, so, yes, I have to be careful of what language I use. However, my language is not going to be the language of epidemiologists. Instead, I’ll be translating the language of epidemiologists into plain English (or maybe even Spanish).
That is where the courses in the DrPH program are failing me. We do a lot of journal club reviews of articles, but they’re all done in “epi speak.” When I took the comprehensive exams last year, someone mentioned that it was clear that my approach to the essay questions on policy were not those of an epidemiologist but those of a public health practitioner. That was a good thing. Because I’m going to be that practitioner while being an epidemiologist, and I’d rather my language be that which translates evidence into action, not the language that leaves people confused.
And certainly not the language that confuses people on purpose.