Data Science & Psychology Data Science applied to Values, Morals, Politics, & things that matter.

Recently, an unprecedented study was done in Oregon where (due to budgetary, not research reasons) a lottery was held to randomly decide which applicants for Medicaid would actually receive the opportunity to receive Medicaid.  There has never been an opportunity to randomly assign people to have access to a program like Medicaid, and so this represents a unique opportunity to learn something about the effects of Medicaid, especially considering the large sample size.  The results were recently published and while there are multiple news reports about the article, none had the depth (or graphs) to satisfy me, hence this post as an excuse to dig deeper.

There is a psychological point to be made here too, about partisan motivated reasoning.  The results of the study are intuitive to most people who are not so ideological.  Of course increased access to health care leads to greater psychological and physical well-being for the insured as well as greater cost to society.  I have to admit that as a liberal, I'm tempted by arguments that the benefits of preventative medicine and reduced use of expensive emergency rooms means that increased health care for the poor will be essentially cost neutral.  While Obama touts improving health care access as a deficit reducing measure and liberal pundits minimize costs, the reality is likely that insuring everyone will cost significant money.  At the same time, arguments that state that medicaid doesn't improve poor people's health or that minimize the effect, seem psychologically motivated as well.  Indeed, the fairest summary of the results of the study would be this fairly obvious sounding ABC News headline which reads, "Medicaid makes poor healthier and states poorer". To add something to the numerous articles out there, I made the following graphs to summarize results:

People who received medicaid (in red) felt happier ("very happy" or "pretty happy" as opposed to "not too happy"), healthier, and less depressed (using this measure) than the control group (in blue).

People who received medicaid (in red) used more preventative services than control group (in blue) (yearly results for just women).

People who received medicaid (in red) used more medical services overall than control group (in blue), costing taxpayers more money, without any decrease in ER visits (yearly numbers, extrapolated from 6 month numbers in article - costs used per event in article, based on previous studies, are in parentheses).

Partisans will surely see it through partisan eyes, as one man's enormous gain in outcomes is another man's modest increase.  The National Review had a fairly detailed critique, but I can't help but feel that statements like "supporters must show not only that expanding coverage improves health but also that it does so at a lower cost to taxpayers than alternative policies" ring hollow unless advocates are forcefully pushing for those policies on the grounds of improving the health of the poor.  It has the same feel as liberal arguments that taxing the wealthy will actually stimulate the economy.  Both groups don't like to make tradeoffs, even obvious ones, but the reality is that expanding health coverage will both cost money and improve health.

Is it an unaffordable amount of money or a trivial amount?  The other neat thing about the study is that it actually translated health usage into actual dollars spent per year.  The control group still spent money on health care, which was presumably taken care of through existing services, charities, or emergency rooms.  The marginal cost of insuring the poor could be seen to be the difference between the experimental and control groups or the total cost of the experimental group.  Under medicaid, the government would pay all those costs, but there may be savings on what government is already spending on emergency room visits to public hospitals and other like societal costs.  In comparison, I found these links for the yearly cost of educating a child or incarcerating a prisoner in Oregon.

Yearly Cost

Of course, the above graph is perhaps misleading as there are far more school children than prisoners, so perhaps multiplying the total cost of care by the 213,000 medicaid eligible uninsured individuals or by the almost 650,000 total uninsured (numbers from statehealthfacts.org), and comparing it to the overall Oregon budget might put the cost of expanding coverage dramatically in context.  Below are yearly Oregon state budget items compared to extrapolated medicaid costs.  Note that the cost of insuring all uninsured is likely lower due to many uninsured being young working adults.  However, there is likely overhead and administrative costs to the program that are not taken into account as well, so perhaps this balances out.

Oregon Yearly Budget Items in Comparison to Potential Expanded Health Care Cost

I learned something from this exercise.  My liberal intuition was that expanding coverage to all the uninsured would not be that large a cost for a state.  In reality, it looks like expanding Medicaid in Oregon would be roughly equivalent to the entire budget of the Oregon University system or at least the community college system, depending on whether you count the entire cost of medicaid health care or just the marginal cost of increased usage.  Either way, it is a significant cost.  At the same time, providing health coverage to all the uninsured is not fiscally impossible.  It costs a fraction of the overall state budget and would cost a fraction of the Oregon health and human services budget.  Behind all the reactions to such studies is the attempt by both liberals and conservatives use motivated reasoning  to avoid a hard choice between a costly government program and failing to provide health care for our nation's poor.  There is a cost, in terms of money or well-being, to either position.

- Ravi Iyer