If you're a regular reader of this blog, you probably realize this essay may not tell you the answer to that question. You know that I often examine questions to which we do not really know the answer, and I am often more interested in how we try to answer the question than in the answer itself.
The pediatricians say no. They've been saying that for years. The August issue of Pediatrics, the official journal of the American Academy of Pediatrics, includes an article (just published online) telling us of an association between physical punishment (such as spanking) and subsequent diagnosis of mental illness. The story has already hit the popular press.
The authors used a database called the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) to discern an association between physical punishment during childhood and later being diagnosed with any of various mental illnesses, including depression and bipolar disorder, anxiety and panic disorder, post-traumatic stress disorder, substance (alcohol and other drug) abuse and dependence, and an array of personality disorders. They separated "harsh" physical punishment from more severe maltreatment. The former included being "pushed, grabbed, shoved, slapped, or hit" by parents (or any adult living in the household), but not being hit so hard it left a bruise or caused an injury.
Pediatricians have long believed (and assert that there is abundant scientific evidence) that more severe physical abuse is associated with subsequent development of mental illness. In this study they find an association between milder (yet still "harsh") physical punishment and mental illness.
So the first question is, do we care? And the answer depends on whether the association reflects a cause-and-effect relationship. Let us examine that question. We could start by looking at the journal article itself, which says the study's "design precludes determining any causal ... relationship between harsh physical punishment and mental disorders." And we could stop right there.
But the authors didn't stop right there. They are clearly inclined to believe there is some causality. So let us consider that question more carefully. Why might there be an association without cause-and-effect relationship?
The most common situation in which A is associated with B, but A does not cause B, is that A and B have contributing factors in common. Examples from everyday life are abundant. "People who get speeding tickets face higher costs to operate their cars." When we look at this, we find that they are not getting into more crashes, so that's not the source of the expense. Maybe if we stopped giving them speeding tickets, and they saved the money on fines and higher insurance premiums, they wouldn't face higher costs, and they'd be better off. But when we look more closely, we find that people who get speeding tickets are people who tend to drive faster. (In other words, they aren't just failing to keep an eye out for the highway patrol.) And people who tend to drive faster get worse fuel economy and burn through tires faster. And that's what causes the higher operating expense. So the speeding tickets don't cause the higher expenses (when you separate out fines and insurance premium hikes). The speeding tickets and the higher operating expenses are both caused by faster driving.
Now back to beatings and mental illness. Maybe the beatings don't cause the mental illness. Maybe they are associated because they have common contributing factors. The authors tried to adjust for some possibilities by looking at whether parents were dysfunctional (meaning parents were alcoholics, or drug addicts, or had mental illnesses, or had been in prison, or had attempted suicide). They also tried to take into account marital status, race/ethnicity, level of education, and household income. From the paper it appears that they looked at these variables for the children rather than the parents, which seems odd (why not both?), but perhaps they didn't have adequate data for the parents. The questions about parents being dysfunctional, like the questions about being beaten, were based on the recollections of the (now grown) children.
There could be any number of things (other than being "dysfunctional" as defined in this study) about the parents that made them inclined to beat their children and made them "bad" parents in other ways that put their kids at risk of later being diagnosed as mentally ill. Maybe it was those other things (and not the beatings themselves) that were the important contributing factors to later diagnosis of mental illness.
We could also look at factors inherent in the children. Maybe there were innate traits in the children that caused them to behave so badly that their parents found it appropriate to beat them, and those same traits resulted in their later being diagnosed with mental illnesses.
Why should we try to figure out whether there is a cause-and-effect relationship?
It is only if there is a cause-and-effect relationship that the finding of an association has any implications for medical practice or social policy. The pediatricians believe parents should be counseled not to beat their children but to use other disciplinary measures instead. And the authors of this paper point out that 32 nations (the U.S. and Canada not among them) prohibit such punishment, the clear implication being that the U.S. and Canada should climb onto this bandwagon.
This study has found an association, and the authors themselves admit the design of the study "precludes" a determination of causality. So if we want to know the answer to the question of causality, how might we find it?
We must first recognize the existence of a potentially large number of what scientists call "confounders" - additional variables that could be influencing the relationship between the two we are trying to study. We must also recognize that we may not be able to identify all possible confounders. And so we must design a study in such a way as to eliminate (or at least minimize to the greatest extent possible) the influence of confounders, both known and unknown. The way medical scientists do this is by conducting a large, randomized study. Randomization and sample size are the keys to minimizing the potential influence of confounders.
So here is the design. We take a very large number of children (say ten thousand) whom we are going to follow from birth. We assign them at random to either of two groups. The first group has their parents counseled (early and often) by pediatricians (or family practitioners) that beating children is not a good idea (and why) and is given advice about alternative methods of discipline. The second group receives no specific counseling of parents about disciplinary measures. We then follow the children into adulthood and see whether they develop mental illness (or anything else we think might be influenced by whether they are beaten as children).
(We could examine lots of other questions and assumptions. For example, we think we know that boys who are beaten as children are more likely to beat their wives, and girls who are beaten as children are more likely to accept such mistreatment as adults. I am as sure as I can be that this is true - it makes so much sense that it has to be true - without compelling scientific evidence. Are they also more likely to beat their own children? Many of us think so. The kind of study I'm describing could answer these and many other questions.)
Of course we realize that in each group there will be children who are beaten and children who are not, so we will also be sorting the subjects into those two groupings. We will compare subjects who were beaten with those who were not, and we will compare those whose parents were counseled with those whose parents were not.
We want to answer two fundamental questions. First, does counseling seem to reduce the likelihood of beating children - in other words, do parents follow the advice they are given? Second, does not being beaten seem to produce emotionally healthier adolescents and adults? If the answers - from a study like this one - to both questions are "yes," then we know this is the advice doctors should be giving parents. And then we can talk about whether this should go beyond advice from doctors to parents and become a matter of social policy (and law).
This study will not be done. Pediatricians will not randomize children to having their parents counseled not to beat them or not counseled about that. And that is because they believe they already know the answer.
We design studies to answer questions to which we believe we do not know the answers. This is called "equipoise" - the answer could be A or B, and we really don't know.
The premature dismissal of equipoise is one of the major challenges in science. You cannot answer a question if you do not study it correctly, and you will not study it correctly if you believe you already know the answer.
No comments:
Post a Comment