Stayin' Alive -- or not

I've been giving some props lately to JAMA and NEJM for making some of their articles of compelling public interest available to the selfsame interested public, but let's not forget that most of their material is still behind the prescription wall, while PLoS Medicine puts it all out there. This number crunching exercise by Ezzati et al has gotten some media coverage, but it does call for a brief comment by YT.

They used county-level data to do historic mapping of changes in life expectancy. There are various statistical acrobatics they used that I won't go into, but take it from me that this is basically legit. From 1961 to 1983, the disparities in life expectancy among U.S. counties broadly declined. Specifically, life expectancy* increased almost everywhere, but it tended to increase faster in those places where it had been lowest to begin with. From 1983 to 1999, however (well, not exactly 1999 -- sort of till the mushy end of the century. They used a three-year moving average to dampen noise in their data), the disparities started increasing. In fact, life expectancy for women actually declined in many counties, and for men in fewer but more than before. For both sexes, these were mostly the places where life expectancy was already on the low side, hence the growing inequality. Meanwhile, the places where longevity was highest saw continued increases.

I had already guessed the basic reasons before the authors got to the punch line. Cardiovascular disease, diabetes and lung cancer are the major contributors for women; homicide and HIV contribute more for men. We are seeing the delayed effects of the increase in smoking by women during the 1950s and 60s; the obesity epidemic; and HIV and violence. With the advent of antiretroviral therapy, the effect of HIV has largely attenuated since 2000, but we can't see that in this data. However, as people with HIV begin to die from the side effects of HIV medications -- yup, those are diabetes and vascular disease -- we may lose that benefit.

Now, here's the kicker: the declines in life expectancy are heavily concentrated in one region, the historic confederacy, from Texas to West Virginia. Virginia and the Florida penninsula, which are no longer so characteristically southern, are exempt. (For some reason Wyoming, which is very sparsely populated, was also a bad place for women, along with a spot here and there such a southern Oregon and Down East Maine.) This is also the region that is most politically conservative, most Republican, and also includes most of the poorest states.

We keep making similar findings -- that the Old South has the worst of everything, from health to education to income. You'd think they'd be the least conservative region, wouldn't you?

*"Life expectancy" is an artificial construct. Basically, you look at the current death rate within each age cohort, usually five year cohorts, then assume that the same death rates will be experienced by people born today as they pass through the life course, and compute the median age at which they would die in that hypothetical future. Of course the future won't be like that.