-You may have noticed that there's a bit of a debate going on about health insurance in this country.

For the video version, click here.

Reflecting the great diversity of American political life, we have those calling for a universal, single-payer healthcare system:

And those who feel, well, that the government should take a more limited role:

But in the end, there’s one question that seems to really cut through the political debate: does health insurance save lives? Multiple groups are synthesizing decades worth of data to answer that question. Last week, we saw a report co-authored by Atul Gawande and appearing in the New England Journal of Medicine that concluded:

“Arguing that health insurance doesn’t improve health is simply inconsistent with the evidence”.

His provocative headline on Vox.com read:

But it turns out, this is a trickier quandry than it may appear at first glance. In a manuscript appearing in the Annals of Internal Medicine, researchers Steffie Woolhandler and David Himmelstein take a look at the data.

Before we get to the results, I want to highlight just why this is such a hard thing to study.

First – who do you compare the uninsured to? Like, what's the control group?  If you compare uninsured people to those who get insurance due to disability or through Medicare, you might be stacking the deck against being insured – since that comparison group is a bit sicker than average. In contrast, if you compare uninsured people to people with employer-sponsored insurance, you might make being uninsured look worse than it really is due to the so-called "healthy worker effect".  Accounting for these imperfect comparisons requires statistical adjustment, which can often be quite difficult.

Second –  not all insurance is the same. The main mechanism by which insurance could save lives is through greater access to care (though, yes, perhaps there is some psychological benefit to just knowing you are insured, whether you use it or not). But not all insurance gives equal access to care.

Finally, if you're talking about mortality, the rates in adults are (fortunately) very low. That means to detect significant differences between uninsured and insured individuals you need a very large study.

What would be ideal is a randomized trial that took a bunch of uninsured people, gave some insurance, and left some uninsured. I mean, it would be ideal scientifically, maybe not ethically.

Actually, this happened in Oregon when the state held a lottery for Medicaid slots.

The study looked at roughly 12,000 individuals, half of whom "won" the Medicaid lottery.  The study reports a lot of outcomes between the groups, but let me highlight just a few:

The rate of catastrophic expenditures was higher in those who lost the lottery, as was the death rate. The rate of diabetes treatment was higher in the lottery winners, suggesting that having insurance does increase access to care.  But the rate of treatment was not 0 in the lottery losers.  That's because some who lost the lottery ended up getting insurance through other means, and some, of course, paid their own way.

But this is Oregon – we might not be able to translate the results to the broader US.

The best national-level data comes from the NHANES study, which is considered a representative sample of Americans.

This study examined around 9000 individuals and their death rates over the next 20 years or so. The risk of death, after accounting for demographics, smoking, alcohol use, BMI, physician-rated health and other factors found that the relative risk of death was 40% higher in the uninsured individuals.

But what about the Affordable Care Act itself. No one would argue that more people aren't insured under the ACA. This graph sums it up nicely:

So how are these newly insured folks doing? We don't really know. It's a bit too soon to tell what effect, if any, the ACA has on mortality. There simply hasn't been enough time to link the potential benefits, particularly in terms of preventive medicine to long-term health outcomes. So we can't evaluate the impact of Obamacare on mortality directly.  But we can look at Medicaid coverage expansions that occurred prior to Obamacare.

Actually, we don't have to look at it.  Harvard researchers already did in this study:

New York, Maine, and Arizona all expanded Medicaid prior to the ACA. The authors compared the change in mortality rates in these states before and after expansion with their neighbor states.

Here's what they found:

What you see is roughly parallel death rates before Medicaid expansion, and a gradually widening split after Medicaid expansion. The authors figured you'd have to cover 176 adults with Medicaid to save one life per year.

There are a lot more studies covered in Woolhandler and Himmelstein's paper, but they all suggest roughly the same thing – that insurance has a modest, but real, effect on all-cause mortality. Something to the tune of a 20% relative reduction in death compare to the being uninsured.

With that backdrop, the question really becomes – how can we ensure that people have access to care? And while we can argue about how that access should be guaranteed – through the government, through private enterprise, or through some combination, it is hard to think that forcing people off the insurance roles in this country can occur without some cost in terms of lives lost.