09 May 2014 New Study Shows Link Between Health Insurance And Mortality

A new study in the Annals of Internal Medicine researches the expansion of insurance under RomneyCare in Massachusetts and finds a link between insurance status and mortality.  Specifically, the study (henceforth the Sommers-2014 study) found that about every “830 adults gaining health insurance [prevented] 1 death per year.”

the Breaking Chain on gray

Thus, we have another entry into the debate over whether there is a link between having insurance and mortality.  And unlike the silly Wilper-2009 study, this one is a pretty solid study, using a quasi-experimental design.  Basically, the researchers compared a set of counties in Massachusetts with similar counties in other states both before and after RomneyCare came into effect in 2006.  The mortality rate in the Massachusetts counties dropped 2.9% post-RomneyCare relative to the other counties.

That said, there are only twostudies in this debate that use the best type of study design, the randomized controlled trial, and both of those studies found no link between insurance status and mortality.  (If you want to know more about difference in study design, this power-point presentation is pretty good.) Both of those studies have their shortcomings, though, including small sample size and abbreviated period of study.  The Sommers-2014 study had relatively large sample size and was able to study affects over four years.

However, the Sommers-2014 study has its shortcomings too.  Phil Klein at the Washington Examiner notes that Massachusetts has the most per capita physicians of any state in the nation and has some very high quality medical institutions.  Klein writes:

So it’s possible that what the Massachusetts study picked up is not necessarily that increasing health insurance coverage decreases mortality in general, but doing so in an area with a vast pre-existing medical infrastructure with a high concentration of skilled doctors and world-class medical facilities is what makes the difference.

Another shortcoming in the Sommers-2014 study is that its unit of analysis is counties, not individuals.  Because of this, the study can’t measure the impact of health status, smoking status and body mass index.  A 2009 study by Dr. Richard Kronik found that excluding those factors increases the strength of the relationship between insurance status and mortality.  Include them and the relationship all but evaporates.

In fairness, the Sommers-2014 study did compare counties that were very similar in their demographic characteristics such as age, median income, employment and poverty rates, and so forth.  Thus, it’s possible that health status, smoking status and body mass index are also relatively similar in all counties in the study.  But since they aren’t measured, we ultimately don’t know, and it leaves open the question whether the study’s findings on insurance status and mortality would weaken or disappear if those factors were included.