Does providing universal access to preventative care reduce emergency department utilization?: Professor Amanda Kowalski uses multiple methods from economics to explore this health policy question.
Does providing increased health insurance coverage reduce emergency department utilization?
Visiting associate professor Amanda Kowalski set out to understand this question in her first few years as an assistant professor. She researched the Massachusetts health reform of 2006 to understand what lessons it might hold for the implementation of the Affordable Care Act (ACA) nationally.
One of the broad findings of her work was that when people gained access to health insurance coverage through the Massachusetts reform, they used the emergency room less often relative to before the reform and relative to other states. “That was good news from a policy perspective,” remarked Kowalski, “because care provided in emergency rooms is expensive and usually of lower quality due to less continuity of care.”
However, the scalability of these policy findings was called into question when, in 2008, Oregon expanded Medicaid health insurance coverage. Oregon used a lottery to select new recipients which yielded a randomized experiment—the so-called “gold standard” in research. (The Massachusetts analysis used difference-in-differences methods). Surprisingly, people who gained health insurance coverage through the new Oregon policy went to the emergency room more—the opposite of the findings in Massachusetts study and also bad news from a public policy perspective as lawmakers became nervous that further health insurance expansion through the ACA may increase utilization of strained emergency rooms and be prohibitively expensive.
“People were quick to say, ‘well, the Oregon health insurance experiment was the gold standard because it was a randomized experiment. Whereas the work that other people had done—including me—on Massachusetts health reform was based on difference-in-difference methods—or a natural experiment.’ I share the view that randomized experiments are the gold standard, but I saw value in the natural experiment in Massachusetts.”
Kowalski’s next question was, how can you have two experiments that would both be right but yield different answers?
First, Kowalski looked through the data and took a traditional approach that compared the differences between Massachusetts and Oregon, however, she did not find any strong explanation for the difference in emergency room utilization.
Next, she asked, “what if the people who got health insurance coverage in Oregon have a different propensity to go to the emergency room than people induced to get health insurance coverage in Massachusetts?”
In Massachusetts—which was the model for the ACA—there was Medicaid expansion but there was also a health insurance mandate, for which those who opted-out of health insurance paid a penalty. Kowalski thought that perhaps “the people in Massachusetts who didn’t have health insurance coverage and got it through the Massachusetts reform are arguably different people than the people who signed up for a lottery for health insurance in Oregon.
By looking at differences within groups, Kowalski found a possible explanation. “I looked just within the Oregon health insurance experiment to see if I could see some groups of people who—if they gained access to health insurance coverage—would go to the emergency room less. And to the extent that I could find that information within the randomized experiments, that information could explain why I would get different results from the randomized Oregon experiments than I get from the mandated Massachusetts policy.”
Kowalski separated people into three groups: always takers (people who would get health insurance whether or not they won the lottery), compliers (people who we can identify as affected by the policy change), and never takers (people who would not sign up for health insurance whether or not they won the lottery).
What Kowalski found was that around 15% of people ended up getting health insurance whether or not they won the lottery (always takers) and that around 60% of lottery winners did not get health insurance—either because they didn’t submit the eligibility information required or because they weren’t eligible (never takers). So effectively, there are some people that got health insurance even if they lost the lottery (always takers), there are other people who got health insurance if they won the lottery but did not get health insurance if they lost (compliers), and there was a final group of people who did not get health insurance either way (never takers). Her thesis based on this disaggregation is that the compliers in the Massachusetts reform are a lot more like the never takers in Oregon than the compliers because of the individual mandate for insurance.
Kowalski explains, “pulling it all together, if I look at the observed characteristics of these groups, which is normally how we think to look for different subgroups in an experiment, I see that these groups differ a lot in their previous emergency room utilization, and this behavior explains some of their difference in emergency room utilization after the health insurance lottery.” Therefore, from a public policy perspective, health coverage expansion can be but is not necessarily correlated with increased emergency room utilization or increased health system costs. The policy details of who the expansion enrolls matter.
Based on this research, Kowalski became more interested in bringing cutting-edge econometric analysis to healthcare, and her current research uses some of the same methods she used to investigate the Oregon health insurance experiment to examine clinical trial data. Personalized medicine is a new buzzword, but clinical trials mostly just compare people based on whether they are randomized in or randomized out. Kowalski believes that building some within-group assumptions can provide some insight into which people will benefit the most from a particular treatment. Currently, her team is analyzing data from a large clinical trial on mammograms and another large clinical trial on menopausal hormone therapy, estrogen, and progestin. Kowalski says, “I get excited about my research day-to-day, I wake up in the middle of the night wanting to come to work on it.”