Something that really, really annoys me is reading a paper discussing a rich and complex data set in which the authors squink their eyes tightly and use statistics to zoom in and stare fixedly at one parameter. It happens all the time. It’s as if some scientists think it’s a triumph to reduce a phenomenon to one single simple cause, rather than appreciating the diversity of inputs.
The latest example is a study pegging yet another medical procedure as the cause for autism, in this case, early induction of labor and augmented delivery. Autism is probably a perfect target for these kinds of silly approaches; it almost certainly has a wide range of contributory causes, and it’s always a mystery to the parents of affected children, who look for answers. It’s the vaccines, they say. No, it’s the drugs we took during pregnancy. No, it’s the doctors who did funny business in the delivery room.
Fortunately, we’ve got Emily Willingham to actually look at the forest.
When she looks at the data, she finds that the authors are right, that there’s a correlation: if a mother gets both induction and augmentation, there’s a 27% increase in the chance that the child will later be identified as autistic.
What they don’t tell you is that the same data set shows that having a college-educated mother increases the odds of autism by 30-33%. And that smoking during pregnancy decreases the chance of an autism diagnosis by 14%.
Wait, stop! If you’re pregnant, don’t take these numbers as an indication that you need to start watching more Glenn Beck to make yourself stupider, and that you need to take up a tobacco habit. You’re looking at the tree again and ignoring the forest. What these correlations suggest is that we should be looking into some property of the population that unites them — that each one in itself is not necessarily causal, but that they are common symptoms of the true link. We need to see the big picture to puzzle out the answer.
And sometimes interpreting the phenomenology of a single parameter analysis would lead to a bad result: I can pretty much guarantee you that being a heavy smoker during pregnancy is much worse for the fetus than non-smoking.
Willingham does see the bigger picture.
This study didn’t show that induction or augmentation during childbirth substantially increases the risk for autism, although it hints at a greater influence of socioeconomic status and by implication, healthcare access. If anything, based on earlier literature, it adds a slight if only mathematical confirmation of the perception that births involving autistic children can be associated with more complications, such as the presence of meconium, gestational diabetes, and fetal distress, than births involving non-autistic children. And that points to induction and augmentation as useful in these situations, not as problematic, and certainly does not affirm them as a risk.
Oh, look, it’s practically a jungle!