Skeptics Stack Exchange is a question and answer site for scientific skepticism. It only takes a minute to sign up.
Sign up to join this community
U.S. Senator Bernie Sanders recently tweeted:
As Republicans try to repeal the Affordable Care Act, they should be reminded every day that 36,000 people will die yearly as a result.
(emphasis mine). How accurate is this statement? Where is the source of his information?
My TL;DR summary is: there is too much uncertainty at this stage to know what the specific effect of an ACA repeal would be.
The rest of this answer explains in more detail that:
On the effect of health care reform and expanded coverage on mortality rates,
Multiple sources claim that upwards of 20 million people could lose insurance if the ACA is repealed, with the specific number depending on the mechanism of the repeal and the reactions of the insurance market (this assuming no replacement - it is impossible to say what the effect would be of a replacement whose details have not been revealed, as of now). These estimates are based on
Items (3) and (4) are predictions of the effect of repeal on a very complex and sensitive system, and so have quite a bit of uncertainty. (See [4], and [5], [7].)
Sanders shared this ThinkProgress piece on his Facebook page on January 5, 2017; it appears to be the source of his claim.
That article claims:
Nearly 36,000 people could die every year, year after year, if the incoming president signs legislation repealing the Affordable Care Act.
which is already less certain than the Sanders tweet (which says 36,000 will die).
They compute the 36k number based on:
A study [1] that looked at changes in mortality in Massachusetts after health care reform and concluded that
The number needed to treat was approximately 830 adults gaining health insurance to prevent 1 death per year.
A brief [4] by some group called the Urban Institute that estimates
The number of uninsured people would rise from 28.9 million to 58.7 million in 2019, an increase of 29.8 million people (103 percent).
They then compute the 36,000 number as: 29.8 million no longer insured/830 insured corresponds to one death prevented.
This is a problematic calculation, because the 29.8 million number includes children, while the 830 number is for adults insured. But that aside, how credible are those numbers?
First, let's look at the effect of health care reform on mortality.
The only nationwide study that I have seen is a 2009 study [6] that looked at the mortality rates among the insured and uninsured. At the individual level, it tells us the increased risk of mortality that is attributed to lack of insurance:
After [age and gender adjustments and] additional adjustment for race/ethnicity, income, education, self- and physician-rated health status, body mass index, leisure exercise, smoking, and regular alcohol use, the uninsured were more likely to die (hazard ratio = 1.40; 95% CI = 1.06, 1.84) than those with insurance.
And at the population level, it estimates:
approximately 44,789 deaths among Americans aged 18 to 64 years in 2005 associated with lack of health insurance.
This is not to suggest that repealing the ACA would bring about that many deaths; the reduction in number of uninsured US adults due to the ACA is not equal to the total number of uninsured adults in 2005. But, at the population level this suggests a ballpark number of approximately 1.2 deaths per thousand uninsured adults per year attributed to uninsurance. (44789 deaths in 2005 among 18-64 year olds attributed by the study to lack of insurance, divided by the 36.864 million adults aged 18-64 who were uninsured in 2005.)
For comparison, the article Sanders was referring to (which uses risk estimated by a different study, not the study we have just discussed) assumes there will be about 1.25 (36000/28900) deaths per thousand uninsured per year.
Other studies on the effect of increasing health insurance coverage and other health care reforms suggest this varies tremendously by state:
830 adults gaining health insurance prevents 1 death per year, according to the study in Massachusetts [1], as quoted above. This is the estimated risk the ThinkProgress article uses in its analysis.
According to another study that compared states expanding Medicaid coverage (NY, AZ, and ME) and states that didn't [2],
176 additional adults would need to be covered by Medicaid in order to prevent 1 death per year
This study found a much greater benefit of expanded Medicaid in reducing mortality, compared to the one the ThinkProgress piece cites (i.e. using this number, they could have arrived at a much higher estimate of deaths).
A study on expanded coverage in Oregon [3] did not look at mortality, but did look at other measures of health, and found no significant benefit to increasing Medicaid coverage. This is obviously a much smaller (i.e. no) benefit of expanded Medicaid in reducing mortality, compared to the study the ThinkProgress piece uses.
Not only is there wide variation across these in terms of benefits of expanded insurance coverage, the authors of all these studies caution that "results may not generalize to other states" [2] (and identify specific variables in the states under consideration that could cause the benefits to be greater than or less than other states). For example, expanding health insurance coverage in a state where most of the previously uninsured do not have doctors or hospitals nearby, will not have the same effect as expanding health insurance coverage in a state with a much higher density of healthcare providers. A state with a higher baseline rate of insurance coverage before enacting reform will not see as great an effect on mortality rate after enacting reform [1]. Similarly, the Oregon paper [3] notes that a small-scale healthcare reform effort is substantially different from a large-scale healthcare reform effort, for better or worse:
the newly insured participants in our study constituted a small share of all uninsured Oregon residents, limiting the system-level effects that insuring them might generate, such as strains on provider capacity or investment in infrastructure.
Next, let's consider the number of people who will lose insurance. This, too, is a subtle and complicated problem involving many variables. The Urban Institute brief claims
a higher rate of uninsurance than before the ACA because of the disruption to the nongroup insurance market.
i.e. it estimates more people will lose insurance than just the individual exchange enrollees or people who gained access to Medicaid as part of the ACA's expanded Medicaid.
Specifically, they estimate that more people will have employer-covered health insurance, fewer will have Medicaid, and fewer will have nongroup coverage:
This is based on the following prediction:
The near “death spiral” in the private nongroup market described earlier is likely to occur immediately after the reconciliation bill’s provisions take effect. Insurers would recognize the unsustainable financial dynamics of broad-based pooling policies (e.g., guaranteed issue, no preexisting condition exclusions, essential health benefits, modified community rating) combined with no individual mandate and no financial assistance to spur enrollment. Similar near market collapse has occurred in the past under similar conditions. When New York’s and New Jersey’s state governments implemented community rating and guaranteed issue in their private nongroup markets without also providing for an individual requirement to obtain coverage or financial assistance to make coverage affordable for people with modest incomes, the nongroup markets unwound (Monheit et al. 2004).
In general, it is plausible - expected, even - that repealing the ACA would not suddenly revert health care back to exactly what it was before it was passed. There have been massive changes throughout the industry as a result of ACA (as alleged both by people who think they have been net positive and people who think they have been net negative, such as this one). However, it's not clear at this point where the dust will settle, so it's hard to say exactly what the effects on insurance coverage and access to healthcare will be.
The U.S. Congressional Budget Office in 2015 [5] arrived at a lower estimate:
CBO and JCT estimate that the number of nonelderly people who are uninsured would increase by about 19 million in 2016; by 22 million or 23 million in 2017, 2018, and 2019; and by about 24 million in all subsequent years through 2025, compared with the number who are projected to be uninsured under the ACA. In most of those years, the number of people with employment-based coverage would increase by about 8 million, and the number with coverage purchased individually or obtained through Medicaid would decrease by between 30 million and 32 million.
However, this number should be revised downward because it is based on what was an overestimate by about 8 million of the number of enrollees predicted in individual exchange plans for 2016:
But then potentially revised upward in the case of a repeal through a reconciliation bill that leaves in place market reforms, because in a letter to the Senate Committee on the Budget, they add:
CBO and JCT have not estimated the changes in coverage from leaving in place the ACA’s insurance market reforms while repealing the subsidies and mandate penalties. However, the agencies expect that, relative to the numbers provided above, leaving the market reforms in place would lead to a further reduction in the number of people covered in the nongroup market and an additional increase in the number of uninsured and people with employment-based insurance.
More recently, in January 2017 [7] the CBO estimated that first 18 million, then 27 million, and finally up to 32 million by 2026 would lose insurance (compared to the situation under the current law) if the ACA was repealead via an H.R. 3762-like reconciliation mechanism, leaving the market reforms in place:
The number of people who are uninsured would increase by 18 million in the first new plan year following enactment of the bill. Later, after the elimination of the ACA’s expansion of Medicaid eligibility and of subsidies for insurance purchased through the ACA marketplaces, that number would increase to 27 million, and then to 32 million in 2026.
based on the following "accounting":
The estimated increase of 32 million people without coverage in 2026 is the net result of roughly 23 million fewer with coverage in the nongroup market and 19 million fewer with coverage under Medicaid, partially offset by an increase of about 11 million people covered by employment-based insurance.
because:
eliminating the mandate penalties and the subsidies while retaining the market reforms would destabilize the nongroup market, and the effect would worsen over time. The ACA’s changes to the rules governing the nongroup health insurance market work in conjunction with the mandates and the subsidies to increase participation in the market and encourage enrollment among people of different ages and health statuses. But eliminating the penalty for not having health insurance would reduce enrollment and raise premiums in the nongroup market. Eliminating subsidies for insurance purchased through the marketplaces would have the same effects because it would result in a large price increase for many people. Not only would enrollment decline, but the people who would be most likely to remain enrolled would tend to be less healthy (and therefore more willing to pay higher premiums). Thus, average health care costs among the people retaining coverage would be higher, and insurers would have to raise premiums in the nongroup market to cover those higher costs. CBO and JCT expect that enrollment would continue to drop and premiums would continue to increase in each subsequent year.
If the market reforms were also repealed, they estimate:
(The number of people without health insurance would be smaller if, in addition to the changes in H.R. 3762, the insurance market reforms mentioned above were also repealed. In that case, the increase in the number of uninsured people would be about 21 million in the year following the elimination of the Medicaid expansion and marketplace subsidies; that figure would rise to about 23 million in 2026.)
The variation in estimates of the potential number that will lose insurance also highlights the danger of drawing firm conclusions. It is extremely difficult to predict the effect of such a complex and far-reaching change. Also, the number of people losing insurance will depend in large part on the mechanism of a repeal, which in turn would affect what individual parts of the ACA (the Medicaid expansion, the market reforms, etc.) stay or go.
The ThinkProgress piece itself notes some uncertainty:
In fairness, 36,000 is a high estimate of the number of deaths that will result if Obamacare is repealed, as there is some uncertainty about how congressional Republicans will repeal the law. Even in the best case scenario, however, a wholesale repeal of Obamacare may cause about 27,000 people to die every year who otherwise would have lived.
but given everything described above, I would similarly challenge their claim that 27,000 deaths is the "best case" scenario.
[1] Sommers, B.D., Long, S.K. and Baicker, K., 2014. Changes in mortality after Massachusetts health care reform: a quasi-experimental study. Annals of internal medicine, 160(9), pp.585-593.
[2] Sommers, B.D., Baicker, K. and Epstein, A.M., 2012. Mortality and access to care among adults after state Medicaid expansions. New England Journal of Medicine, 367(11), pp.1025-1034.
[3] Baicker, K., Taubman, S.L., Allen, H.L., Bernstein, M., Gruber, J.H., Newhouse, J.P., Schneider, E.C., Wright, B.J., Zaslavsky, A.M. and Finkelstein, A.N., 2013. The Oregon experiment—effects of Medicaid on clinical outcomes. New England Journal of Medicine, 368(18), pp.1713-1722.
[4] Blumberg, L.J., Buettgens, M. and Holahan, J., 2016. Implications of Partial Repeal of the ACA through Reconciliation. Washington, DC: Urban Institute.
[5] United States Congressional Budget Office, 2015. Budgetary and Economic Effects of Repealing the Affordable Care Act. https://www.cbo.gov/publication/50252
[6] Wilper, A.P., Woolhandler, S., Lasser, K.E., McCormick, D., Bor, D.H. and Himmelstein, D.U., 2009. Health insurance and mortality in US adults. American journal of public health, 99(12), pp.2289-2295. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2775760/
[7] United States Congressional Budget Office, 2017. How Repealing Portions of the Affordable Care Act Would Affect Health Insurance Coverage and Premiums. https://www.cbo.gov/publication/52371
[8] United States Congressional Budget Office, 2016. Federal Subsidies for Health Insurance Coverage for People Under Age 65: 2016 to 2026. https://www.cbo.gov/publication/51385
TL;DR: There is no evidence that Obamacare reduces mortality; if anything, it increases it.
If extending taxpayer-sponsored insurance to 15 million people since 2013 has resulted in 21,000 fewer annual deaths, then the mean death rate should decrease from 310.4 to approximately 300.
Returning to the WONDER database for 2014-15 numbers, one finds the mean death rate is … 320.4.
Source: The Federalist, which gets its data from the Center for Disease Control's WONDER database.
To summarize, the mean death rate had been 310.4 and should drop to around 300 if Obamacare had prevented 21,000 annual deaths (much less 36,000). Actual result was an increase to 320.4. So the Patient Protection & Affordable Care Act (PPACA; colloquially known as Obamacare) cannot be shown to have reduced mortality. In raw numbers, it has increased.
The same source then speculates that this might be caused by an external cause. So it runs the numbers again, excluding external causes.
For the decade 2004-2013, the death rate is 247.4 people per 100,000 population. It is more stable than the all-cause death rate, with a low of 244.7, a high of 249.9, and a standard deviation of 1.7.
With Obamacare extending insurance to 15 million more people, this death rate should fall to 238 per 100,000. The 2014-15 data show the actual reported death rate among U.S. adults, excluding external causes, is … 252.9.
To summarize, the average rate during the 2004-2013 time period, the death rate varied between 244.7 and 249.9. So they calculated that if increased coverage was causing the claimed improvement in mortality, that number should have dropped to around 238 in 2015. Instead, it increased to 252.9. On the bright side, that's a smaller increase than with external causes.
It has been suggested that this is caused by an aging population. So what happens if we look at the Age Adjusted Rate in CDC's WONDER?
Centers for Disease Control and Prevention, National Center for Health Statistics. Underlying Cause of Death 1999-2015 on CDC WONDER Online Database, released December, 2016. Data are from the Multiple Cause of Death Files, 1999-2015, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. Accessed at http://wonder.cdc.gov/ucd-icd10.html on Oct 7, 2017 7:41:37 PM
This shows smoother, less variable results than the previous attempts. From 1999 to 2010, the numbers go down rather steadily from 249.6 to 212.3. From 2010 to 2015, the numbers stabilize to 212.0 in 2015. If they had gone down at the same rate as from 1999-2015, that would have been about 195 in 2015.
Is the Age Adjusted Rate the correct one to use? If we look at how it is calculated, we find that it works by aggregating the data by ten-year age brackets and then generating a weighted sum as if the age brackets were as common as they were in 2000. This may be the best that CDC's WONDER makes available, but it's not what we'd use if we were calculating these numbers ourselves.
I would prefer to bucket by year rather than decade. And I would like to use age distributions from the 2010-2015 period, as those are the ones about which we are talking. The 2000 number may give an incorrect estimate of the impact in years like 2013 and 2015.
I also find it worrisome that if we look at just the 25-34 bracket, we see increasing mortality in the 2010-2015 period. At that age, there shouldn't be a significant impact from age. If anything, the younger members are more prone to risky behaviors.
Data from the Centers for Disease Control and Prevention show that overdose deaths per capita rose twice as much on average between 2013 and 2015 in states that expanded Medicaid than those that didn’t—for example, 205% in North Dakota, which expanded Medicaid, vs. 18% in South Dakota, which didn’t. . . . Between 2010 and 2013, overdose deaths rose by 28% in Ohio and 36% in Wisconsin. Between 2013 and 2015, they climbed 39% in Ohio, which expanded Medicaid, but only 2% in Wisconsin, which did not.
From the Wall Street Journal.
The original assumption was that we shouldn't count external causes because the opioid epidemic was independent of PPACA. However, if PPACA makes the opioid epidemic worse, that reason goes out the window.
Centers for Disease Control and Prevention, National Center for Health Statistics. Underlying Cause of Death 1999-2015 on CDC WONDER Online Database, released December, 2016. Data are from the Multiple Cause of Death Files, 1999-2015, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. Accessed at http://wonder.cdc.gov/ucd-icd10.html on Oct 7, 2017 9:29:18 PM
We're back to the original conclusion. Even adjusting for age, mortality rates first stabilized and then increased after PPACA. Prior, mortality rates had been falling.
Prior to PPACA, children were covered under their parents' coverage, Medicaid, or by programs like the Children's Health Insurance Program (CHIP). We can expect that PPACA would have less effect on them.
Prior to PPACA, seniors aged 65 and up were and are covered by Medicare. The only effect that PPACA should have on them would come from the cuts that were used to fund PPACA.
The age adjusted rates require use of the age brackets. So we could either start at age 15 or at 25. We started at 15. This includes those who stayed on their parents insurance from 18 to 25. We could and did end at 64, exactly where the non-Medicare group ends.
When looking at those 15-64, mortality rates have increased or at least stayed the same since PPACA passed. This is true regardless of whether we included external causes or used the age adjusted rates. While mortality rates fell from 1999-2010, rates either held steady after 2010 or increased.
As such, we can't make any claims about what would happen if PPACA were repealed in its entirety. We might hope that mortality rates might fall again, but we don't know that.
It can be argued that this analysis ignores confounding factors and doesn't establish causality. That's true. However, the 36,000 claim ignores almost all data. The only data it tries to incorporate is increased coverage data. It then assumes that increased coverage will have a certain benefit in terms of mortality. Where does it get that benefit? It uses a value that someone made up for use in a previous model. That's not science. It's barely math.
All that ignores the possibility that the increased coverage is coming at the expense of quality of coverage. And there is some evidence of that. For example, Time magazine reported that health insurance satisfaction reached a ten year low in 2015. Gallup reports that continued in 2016. Time reported a satisfaction rate of 69% in 2015 and Gallup reported 66% in 2015 and 64% in 2016.
Columnist George Will said:
When we started this health care debate a year ago, 85 percent of the American people had health insurance, and 95 percent of the 85 percent were happy with it.
PolitiFact ruled that mostly true. They found that Will cherry picked that number, but there was in fact a survey showing an approval of 95% (in 2008). And the other surveys they found were consistently at least 81% in 2009. So we can see that satisfaction dropped for employer plans dropped from at least 81% in 2009 to 66% in 2016 (Gallup).
It is absolutely true that we don't know what would have happened in 2010-2015 without PPACA. But we do know what did not happen with PPACA. And what did not happen is a decrease in mortality rates, even though rates fell consistently from 1999-2010. So at best we can say that PPACA was ineffective at overcoming any confounding effect. At worst, it may have been actively harmful.
I agree that we can't make either statement at this time. Which takes me back to the 36,000 claim being hooey. If we can't even explain the actual behavior of mortality rates since 2010, then there is no way that we can project out behavior if PPACA were repealed. There is at this time, literally zero evidence that PPACA has prevented mortality on an average basis. And any claim that repeal would increase mortality is without a scientific foundation.
Perhaps the evidence is being collected and will be published. But it doesn't exist at this moment in time. What little evidence we have goes the other direction.
