On this blog in the past, we have looked at some intriguing ways in which social issues such as violence may be considered as epidemics. We have also looked at some of the problems in public health with confusing correlation with causation; a classic statement of the fallacy is often given as follows: in summer ice cream sales go up, and murder rates go up. Therefore, eating ice cream causes murder.
The example of ice cream and murder is absurd, but it points out just how difficult it can be to ascribe causation definitively in matters of public health. Clearly, both ice cream sales and murder rates are independently affected by the same actual cause (heat waves), but one could easily imagine compelling data showing that ice cream sales go up just before each wave of violence. And in fact, a fascinating new piece in Mother Jones has been getting a lot of attention in public health circles this week because it shows exactly that kind of compelling relationship between violence and a different factor: leaded gasoline.
Through a pretty careful analysis of past publication, the article makes an extraordinary claim: “Gasoline lead may explain as much as 90 percent of the rise and fall of violent crime over the past half century”. But it has the data to back it up, and what’s really intriguing is that these correlations hold from the macro- all the way down to the neighborhood level. In neighborhoods where lead is removed, crime rates drop a predictable number of years afterward. If there really is a causal relationship between lead exposure and violent crime, we should be making the removal of lead from the environment a top priority- and maybe we should also be reconsidering the effectiveness of the police campaigns that are claiming the credit for the tremendous decline in violent crime America has been experiencing in recent decades.
But is this really ice cream and murder all over again? Scott Firestone has an excellent blog post about the MJ piece that does a nice job discussing why we might temper our enthusiasm about these findings somewhat (although he also finds the data very compelling), and it’s worth reading just to think more about how hard it can be to prove anything with certainty, even when the evidence is extraordinary (think of how successful tobacco lobbyists were for so long in creatively interpreting the data on the health effects of smoking). There’s a brief and well-executed discussion this week in Scientific American about just how hard it can be to establish causation in health on another issue: whether even very moderate amounts of drinking during pregnancy has any negative effects on babies. This should be easy to establish, but it isn’t: in part because of ethical considerations (you can’t set up a control for potentially harmful behavior), in part because of the reliability of self-reports, and in part because of confounding variables like “lifestyle” associations (the same arguments tobacco lobbies make).