Inside the Beltway, it’s easy to forget that the fate of the Affordable Care Act is a question of life and death.
If President Trump and the Republican-majority Congress repeal or destroy the law without replacing it, people will die. But getting a firm number on exactly how many deaths an ACA repeal would lead to is surprisingly contentious.
Right now we don’t have any direct evidence of Obamacare’s impact on mortality, since it takes a long time for changes in things like health insurance to show up in death statistics and the law was only enacted in 2009. (Studies looking at whether the ACA improved people’s health status and access to health care have had positive results, which suggests it should increase longevity.)
More from Vox:
Marine Le Pen, the far-right politician topping the French polls, is thirsting for a Frexit
Trump’s new sanctions targeted Iran. So why is China angry about them?
The Philippine president says he’s going to hell for his vicious drug war. And he’s proud of it.
We do have something else though: many different studies on the impact of health insurance on mortality before Obamacare was in the picture. Since the US is unique among wealthy nations in that it doesn’t provide universal health coverage for its citizens, estimating the tollof this failing is an obsession among health economists. But all these studies come to different conclusions, which opens the door for wild speculation about how much death an Obamacare repeal could lead to.
There is a better way. By focusing on the strongest of these health insurance and mortality studies, we can come to some pretty clear estimates — and throw out some of the claims people make that are based on shaky science. The best research out there suggests the very high and very low numbers you’re reading are probably wrong.
Here’s a quick tally of the most often cited research in this debate, which studies to trust, and what they reveal about life, death, and health insurance.
There are two main types of studies that can tell us something about the effect repealing the ACA might have on mortality in America: First, there are simple non-randomized observational studies that use large health data sets to compare mortality rates between people who have had health insurance (government or employer-sponsored) and people who have not. Second, there are quasi-experimental studies, a stronger type of observational study design, which compare states that have expanded health insurance with those states that haven’t in a natural experiment.
The simple observational studies have come to pretty mixed conclusions.
- This 2009 paper in Health Services Research by Richard Kronick of UC San Diego followed some 640,000 respondents of a large health survey — the National Health Interview Survey — from 1986 to 2002, looking for differences in the mortality rates of people who had insurance coverage compared with those who didn’t. (The study was a reaction to an oft-cited 2004 report from the Institute of Medicine, which determined that the lack of universal health coverage in the United States led to an extra 18,000 deaths per year.) Kronick found something quite different: Lacking health insurance in his study was not associated with a greater risk of mortality. So if you use this study as the basis for estimating the impact repealing Obamacare could have on death in America, you might think it would lead to no extra deaths.
- Another observational paper, published in the American Journal of Public Health that very same year (2009), used a similar method on a different data set (the National Health and Nutrition Examination Survey), again looking at how the health of people with insurance differed from those without. This time, its lead author, the University of Washington’s Andrew Wilper, found being uninsured seemed to come with a significant increased mortality risk. “Lack of health insurance is associated with as many as 44,789 deaths per year in the United States, more than those caused by kidney disease.”
There’s a catch to these papers that belies one of the reasons they came to such contradictory results: The study design was weak.
The researchers compared people who had insurance with those who didn’t, but there are a lot of things that differ about these two groups: The uninsured tend to be poorer, sicker, and have worse health behaviors, for example. It’s difficult to fully control for these factors — and that’s why the results of these studies are somewhat squishy.
Kronick thought the mortality gap between the insured and uninsured could be explained by these other factors — income, health behaviors, etc. — and in his study, the difference in life expectancy disappears when you control for these variables. Meanwhile, Wilper found that after accounting for a similar set of factors, people without insurance still die younger. So the two study authors used similarly weak methods on two different data sets, which explains their divergent findings.
“The primary conclusion I would draw is that it is not possible to have much confidence in the results of the sort of observational analysis used in my study or [in Wilper’s],” said Kronick over email. More recent quasi-experimental studies “are more fruitful, and do support the common sense conclusion that expanding insurance coverage will lead to reduced morbidity and mortality.”
So let’s turn to those.
Traditionally, Medicaid covered only low-income children, parents, and pregnant or disabled people. Over the past 10 years, several states expanded Medicaid coverage to include non-disabled adults who don’t have children — a group that is similar to the population that gained eligibility under Obamacare. But Medicaid expansion has not been spread evenly throughout the US, and the Supreme Court ruled it was an optional feature of the ACA in 2012. This led to a series of natural experiments across the country, in which researchers have compared the mortality rates in places that expanded coverage with places that didn’t.
- The researchers behind this 2012 New England Journal of Medicine study took advantage of that variation: They compared what happened to health in three states (New York, Maine, and Arizona) that expanded Medicaid eligibility since 2000 with neighboring states without expansions, covering a period of five years before and five years after each state’s expansion. They found mortality declined in places that expanded Medicaid by 20 deaths per 100,000, unlike neighboring states that didn’t expand Medicaid. Extrapolating that to the estimated 20 million who could lose health insurance with an ACA repeal, other researchers have suggested this would translate to 43,956 deaths in the US per year.
- Massachusetts has also offered a natural experiment for researchers who want to understand the impact of expanding health insurance on mortality rates. The state underwent a health reform in 2006 with the goal of providing insurance to almost all of its residents — and it became the model for the ACA. The best paper on this, published in 2014 in the Annals of Internal Medicine, compared the mortality rates in Massachusetts counties from 2001 to 2005 (before health reform expanded insurance) and 2007 to 2010 (after health reforms) with changes in control counties with similar demographic and economic conditions. Here, they found that the health insurance expansion prevented 320 deaths per year since it began in 2006. If that pattern holds for the ACA, the White House Council of Economic Advisers has estimated that it means 24,000 deaths per year nationwide are averted because of the ACA. (Others, including Harold Pollack, have made the same calculation.)
Unlike the Wilper and Kronick studies, which simply compare the uninsured with those who have insurance, these more recent quasi-experiments tell us what happens when uninsured people gain coverage. This is a much more reliable way to find out how many deaths insurance may have averted. And even though they come to different conclusions, they both suggest we’d see a lot of people die if Obamacare were repealed.
*There are high quality randomized trials in this space, like the Oregon and RAND experiments, but they can’t tell us much about mortality. For more on why, see the footnote.
So if we trust these “quasi-experiments” more than the simple observational studies, we’re looking at somewhere between 24,000 and nearly 44,000 extra deaths per year if 20 million people lose health insurance with an Obamacare repeal.
That’s still a sizable range, so I reached out to the author on these last two studies, Benjamin Sommers, to ask him about which he thinks is the better model for estimating the impact of repealing the ACA.
Sommers, a health economist and physician based at Harvard University who has extensively researched the effects of health insurance, argued for the Massachusetts figure as the “more cautious and reasonable estimate.” The 2006 health care reform in the state became the model for the ACA, so he thinks the experience there — and the population it affected — is a better match for what’s happened nationally than the Medicaid expansion study in New York, Maine, and Arizona.
Still, this figure doesn’t come without some uncertainty. “On the one hand,” Sommers said, “Massachusetts has the most doctors per capita in the country and a strong health care infrastructure.” These factors may have enhanced the benefits of gaining coverage, and they wouldn’t translate to places that have a patchier health infrastructure. “On the other hand, Massachusetts already had a low uninsured rate before 2006 from prior coverage expansions and because it is a relatively high-income state.” Most other states didn’t have these features before the ACA, which means that the benefits of expanding health insurance under the law might be even greater outside Massachusetts.
Again, the Massachusetts study suggested 24,000 deaths were averted each year as a result of expanding health coverage to 20 million people. “Is it the case that it’s exactly the opposite if we repeal the law?” Sommers asked. “It’s a fair question.” For now, though, it’s the best estimate we have. And if the ACA is repealed, “repaired,” or replaced with something else in a patchwork across the US, we may never have a clearer result.
It would be an oversight not to mention that some of the best data out there on the impact of health insurance comes from Oregon. There, researchers tracked what happened to the winners and losers of a state lottery that offered Medicaid to 10,000 randomly selected Oregonians in 2008. This is the closest we have to a randomized trial on health insurance, which was certainly a scientific breakthrough: There had never been a randomized trial — the gold standard study methodology — that gave some people insurance and withheld it from others, since that would be unethical to do. So this study was a first.
But unfortunately, the Oregon research can’t tell us all that much about mortality. The population the researchers looked at was too small (10,000 people compared with the 270,000 in the Massachusetts study), and it lasted for only two years, which isn’t enough time to evaluate the mortality impact of Medicaid versus no insurance, Harvard’s Katherine Baicker, a lead author on the study, said over email. “While the Oregon study gives what I think are the most scientifically robust findings on the outcomes it examines, it does not have sufficient sample size to detect the changes in mortality that one might expect to find from Medicaid expansion.”
Still, added CUNY’s David Himmelstein, “[i]ts findings of small — although not statistically significant — improvements in blood pressure and diabetes control, Framingham Risk score, and substantial statistically significant improvements in depression, are consistent with estimates from other studies of decreased mortality from gaining health insurance.”
The RAND Health Insurance Experiment, which ran from 1974 to 1982, is another important randomized trial on health insurance. But there was no uninsured group in the study (it only assigned people different types of health insurance and followed them up). The research is also pretty old, predating many current lifesaving medical interventions like statins to lower cholesterol, or stents for people suffering heart attacks — or all the things people who gain health insurance could access today. So it also doesn’t reveal much about insurance and death.