Medicaid coverage had a significant impact on helping some people lower their blood pressure, a UCLA-led study suggests.
The findings, published in the peer-reviewed journal The BMJ, fill a gap left in the Oregon Health Insurance Experiment finding that Medicaid coverage, which is intended for lower-income populations, leads to improved financial risk protection, better access to care, and lower mental stress, but found no impact on physical health such as blood pressure.
“These findings are important because the Oregon Health Insurance Experiment is one of the few randomized controlled trials that enables us to assess the causal impact of health insurance coverage,” said Dr. Yusuke Tsugawa, associate professor of medicine in the division of general internal medicine and health services research at the David Geffen School of Medicine at UCLA and the study’s senior author. “Our results should be informative to policymakers and health policy researchers, as they provide robust evidence that health insurance not only improves mental health, as the original study has found, but also improves physical health, such as lowering blood pressure.”
The Oregon experiment was launched in 2008 to measure the effects of Medicaid coverage on a range of outcomes such as healthcare use, mental and physical outcomes, and financial strain. It did not, however, capture what is called heterogeneous treatment effects, which is how treatment effects can vary based on people’s individual characteristics.
The researchers used a machine-learning algorithm called causal forest to pick out people whose cardiovascular health could improve with Medicaid. Causal forest can estimate the effects of an intervention on an outcome based on a person’s characteristics, allowing for personalized predictions of how each individual will benefit from a given treatment.
The researchers dug into data for 12,100 Oregon experiment participants. They found Medicaid coverage lowered systolic blood pressure by 4.96 mmHg in people that the algorithm predicted would benefit most.
They also found that improvements in blood pressure were limited to individuals with low utilization of care prior to receiving Medicaid, and that those with higher prior utilization did not see similar improvements. These findings suggest that the null average effects of Medicaid coverage in the original study might be attributed to the benefits gained by individuals who previously lacked access to care being offset by the lack of benefits experienced by those who already had access to care before receiving Medicaid coverage.
“Our findings highlight the importance of looking beyond ‘average effects’ to consider that the impact of interventions often varies heterogeneously based on the characteristics of study participants,” said Tsugawa, who is also an associate professor of health policy and management at the UCLA Fielding School of Public Health. “By focusing too heavily on average effects without comprehensively evaluating the "heterogeneity" in the effectiveness of the intervention, such as health insurance policies, we risk incorrectly concluding that an intervention is ineffective when, in fact, specific identifiable subgroups are benefiting.”
The study has some limitations. Among them, the researchers lacked information on characteristics such as drinking, smoking, obesity status, and family disease history; characteristics were self-reported, making them potentially susceptible to measurement error and misclassification bias; and they examined only limited health outcomes.
The findings, however, highlight how the effect of treatments can vary among population subgroups, Tsugawa said.
“By using advanced analytical tools and considering a wider range of patient characteristics, future studies can help develop more personalized treatment approaches,” he said. “This will ensure that medical interventions are tailored to the unique needs of different patient groups, ultimately improving healthcare outcomes for everyone.”
Study co-authors are Dr. Kosuke Inoue of Kyoto University, Susan Athey of Stanford University, and Katherine Baicker of the University of Chicago.
The study was funded by Gregory Annenberg Weingarten, GRoW @ Annenberg, the National Institutes of Health (P01AG005842 and R01AG034151), the Japan Society for the Promotion of Science (22K17392 and 23KK0240), and the Japan Science and Technology Agency (JST, JPMJPR23R2).
