Effects of Medicaid coverage on cardiovascular health outcomes

Cardiovascular disease is the leading cause of mortality globally, substantially impacting health and increasing healthcare costs.¹ According to a 2023 World Heart Federation report, deaths from cardiovascular disease increased worldwide from 12.1 million in 1990 to 20.5 million in 2021.² In the United States, nearly half of adults, approximately 127.9 million individuals, are affected by at least one cardiovascular disease condition.³ The prevalence of cardiovascular disease increases with age in both men and women. Despite the covid-19 pandemic, heart disease remained the leading cause of death in 2022.⁴ Hypertension is the primary risk factor contributing to the burden of cardiovascular disease. Other risk factors include tobacco use, unhealthy diet, physical inactivity, air pollution, obesity, diabetes, harmful use of alcohol, and socioeconomic deprivation.⁵

In their study, Inoue and colleagues used machine learning techniques to investigate the effect of Medicaid coverage on cardiovascular health outcomes (doi:10.1136/bmj-2024-inok079377).⁶ Medicaid is a US public health insurance programme for people with a low income. The authors found that Medicaid coverage significantly reduced systolic blood pressure and glycated haemoglobin (HbA_1c) levels for certain groups of adults, with a clinically meaningful reduction in blood pressure by approximately 5 mm Hg (−4.96 mm Hg (95% confidence interval −7.80 to −2.48)) in people with low or no prior healthcare charges.⁶

This study is commendable for several reasons.⁶ Firstly, the use of a machine learning causal forest model provides a nuanced understanding of heterogeneous treatment effects, often overlooked by traditional methods that focus on average treatment effects. Secondly, this approach aligns with the growing emphasis on personalised medicine and targeted health interventions, making the findings relevant to policy makers. Thirdly, using data from the Oregon health insurance experiment, a randomised controlled trial, enhances the findings’ credibility by minimising selection bias.

However, future research should confirm that the benefits shown in the Medicaid group were not confounded by other factors. More detailed baseline characteristics and stratification, including smoking status, alcohol consumption, physical activity, mental health status, and family disease history, should be accounted for to strengthen future analyses.⁷

While the findings from the randomised Oregon health insurance experiment are robust, replication in other states or countries is necessary to ensure generalisability. Future studies should validate these results using diverse populations. For example, previous observational research suggests that Medicaid’s impact varies across US states⁸ and within subgroups such as children and adolescents (≤18 years) with congenital heart disease.⁹ Additionally, older patients with non-ST segment elevation myocardial infarction who have polyvascular disease have a significantly higher long term risk of recurrent events or mortality within three years compared with those with coronary artery disease alone.¹⁰ Inoue and colleagues’ new analysis, based on a 17 month follow-up, might not capture Medicaid’s longer term effects on cardiovascular health.⁶ Further studies are needed to determine whether the observed benefits persist over time.

This study discussed potential mechanisms through which Medicaid coverage may improve cardiovascular outcomes, such as increased access to healthcare and reduced financial stress, but did not provide empirical analysis.⁶ Future research should investigate how changes in healthcare use, medication adherence, and lifestyle adjustments contribute to the observed health benefits. Previous evidence suggests that Medicaid access can improve outcomes by managing cardiovascular risk factors, for example.¹¹

While Inoue and colleagues’ study identified subpopulations with important health improvements, a deeper analysis of these groups’ characteristics and needs would be beneficial.⁶ Understanding the specific barriers this population face, and how Medicaid addresses these barriers, would provide actionable insights for designing targeted interventions. Extending Medicaid coverage post partum could reduce cardiovascular maternal mortality, showing the importance of tailored interventions.¹²

The wider implications of this study include showing the importance of personalised health interventions.⁶ By identifying subpopulations that benefit most from Medicaid coverage, policy makers and healthcare providers can tailor interventions to maximise health benefits, aligning with the broader movement towards precision medicine and personalised healthcare. The Medicaid expansion under the Affordable Care Act was associated with modest improvements in cardiovascular risk factors, supporting the need for personalised approaches.¹³

Findings also highlight the need for equitable health insurance policies that address the diverse needs of different populations. Medicaid coverage offers significant benefits to individuals with low prior healthcare charges, reflecting limited access to care before coverage. Ensuring that these economically disadvantaged populations receive adequate health insurance could reduce health inequities and improve overall public health. A study using data from the 2014 behavioural risk factor surveillance system observed that Medicaid expansion reduced cardiovascular disparities, indicating the policy’s potential to address health inequities.¹⁴

The application of machine learning techniques in health policy research enables the identification of varying treatment effects across different groups, showing insights that traditional methods might miss. While artificial intelligence (AI) encompasses machine learning, machine learning is specifically useful for such targeted analysis. A cross-sectional study across European countries showed that data linkage is commonly used in public health activities, but AI application is less frequent. Barriers such as data regulation laws, resource limitations, and governance issues hinder the broader adoption of AI.¹⁵

Inoue and colleagues’ study makes an important contribution to our understanding of the impacts of health insurance coverage.⁶ While the study has some limitations, its findings have implications for health policy design and implementation. Future research should build on these insights, focusing on external validation, longer follow-up periods, and a deeper understanding of the mechanisms underlying the observed benefits. Ultimately, this study underscores the potential of Medicaid coverage to improve cardiovascular health for specific subpopulations, informing more equitable and effective health policies.