Debate Surrounds Medicaid Eligibility Changes

There’s been a lot of discussion regarding the recent reduction in Medicaid eligibility that Congress approved. The implications for existing Medicaid recipients are particularly concerning. A key piece of evidence in this debate stems from the Results of the Oregon Health Insurance Experiment (OHIE), which I led to investigate the effects of providing Medicaid coverage to low-income, uninsured adults over one to two years. It’s gratifying to see this research referenced in policy discussions, yet it’s a bit disheartening when the findings are misinterpreted.

A crucial point of confusion involves what are termed “null results.” These are findings indicating that the impact of Medicaid is statistically insignificant. For instance, the OHIE demonstrated no statistically significant effects of Medicaid coverage on mortality or several health measures like hypertension, high cholesterol, and diabetes.

Sadly, misunderstandings are frequent. We often confuse a lack of evidence for influence with proof of no effect.

Two economists recently wrote in the Wall Street Journal about this:

“The best evidence regarding the health effects of Medicaid expansion comes from the Oregon health insurance experiment. The OHIE is a randomized controlled trial, or RCT. The gold standard for such studies.”

Null results can be very reasonable. They prompt us to reevaluate our assumptions and foster innovation. When making decisions based on evidence, understanding what doesn’t work holds equal importance to knowing what does.

However, it’s vital to interpret null results correctly. The OHIE’s findings indicated NO statistically significant impacts of Medicaid on specific health measures or mortality.

That said, I can’t assert there’s no evidence that Medicaid has any effect whatsoever. The distinction between a lack of evidence for influence and actual evidence of no influence might seem trivial, yet it plays a crucial role, especially when considering that 12 million people are at risk of losing their health insurance. Understanding this difference is essential.

To assess the estimated effect size and uncertainty involved, we should look beyond a straightforward summary of whether evidence of a statistically significant effect of Medicaid exists. Research outcomes come with a pragmatic value range that represents statistical uncertainty about the true effect. If this range includes zero, we can’t exclude the possibility of no effect. But it’s also true that other values within this range remain plausible.

Take some health outcomes from the OHIE where statistical significance wasn’t found regarding Medicaid’s effects. Some of these “null results” could still offer valuable insights.

For example, concerning hypertension, my co-authors and I observed that Medicaid coverage did not significantly reduce hypertension rates. Our study was precise enough to exclude larger estimates found in earlier research indicating Medicaid could effectively reduce hypertension.

In other words, the potential benefit in hypertension reduction was smaller than previously thought, based on this “null outcome.” Again, it doesn’t prove Medicaid has no effect on high blood pressure, which is a helpful addition to the conversation.

Yet, the null results linked to diabetes control (like high glycated hemoglobin percentages) and mortality provided no useful insights. This was likely due to the relatively small sample size of the Oregon study—only around 10,000 people received Medicaid coverage—and the low incidence of diabetes (about 5%) and mortality rates (less than 1%) in the study group, resulting in high uncertainty.

For diabetes, the plausible range of Medicaid effects included zero, but it also allowed for significant improvements based on the anticipated increase in diabetes medication use. Clinical literature suggests such advancements in medication therapy could lead to better glycated hemoglobin levels.

Thus, we cannot entirely dismiss the potential for Medicaid to have no effect on diabetes, nor can we overlook its possible effectiveness, based on its influence on diabetes medications. I categorize this a specific type of null result.

Similarly, the mortality rates in this study were not informative. The data could not exclude the possibility that Medicaid reduces or increases mortality rates. A much larger randomized controlled trial involving around 4 million participants prompted to enroll in health insurance found a statistically significant impact on reducing mortality rates in those aged 45 to 64. The study authors noted that their findings aligned with the range of mortality effects estimated in the OHIE.

The authors of the Wall Street Journal letter emphasize that randomized studies can yield some of the strongest evidence regarding program effectiveness. However, to use this evidence correctly in policy discussions, we must understand that null results don’t automatically determine a program’s efficacy. Both researchers and policymakers need to handle evidence, including null outcomes, carefully and responsibly.