Artificial intelligence (AI) is poised to become a major part of the nation’s healthcare system. In just the past five years, the Food and Drug Administration (FDA) has approved more than 500 AI-enabled clinical software applications for use in healthcare settings. Today, AI helps clinicians perform a range of clinical tasks, from analyzing medical imaging data and interpreting the results of MRI scans, to aiding in the diagnosis of diseases.

While the clinical and diagnostic benefits of AI are clear, the impact these new technologies have on healthcare costs is less understood. There is limited evidence suggesting that AI has the potential to reduce healthcare spending in a number of ways, from pre-reading test results and saving clinicians time to replacing more costly and invasive diagnostic tests.

Despite these anecdotal examples, however, the potential cost-saving measures from AI technology have not been fully analyzed, and no empirical evidence exists on the causal relationship between the adoption of AI-enabled clinical software applications and healthcare spending.

With funding from the National Institute for Health Care Management Foundation (NIHCM), researchers from the University of Minnesota School of Public Health (SPH), the University of Chicago, and Harvard University aim to address the impact of AI-enabled SaaS applications on cost and other aspects of health care. Using Medicare-reimbursement data from 2016-2024, the researchers will explore AI-enabled SaaS applications’ impact on spending, testing, diagnosis, and overall health outcomes by data from clinicians that adopted the software versus non-adopting clinicians.

By providing the first causal evidence on the spending and health effects of AI in healthcare the researchers noted that their findings would be of immediate and long-term use to the Centers for Medicare and Medicaid Services (CMS) and other health care payers.

“CMS and other payers are currently wrestling with questions regarding the reimbursement, spending implications, and overall value of AI-enabled clinical software applications, and we aim to shed light on those questions,” says SPH Assistant Professor and study lead Hannah Neprash. “If we document financial savings and health improvements associated with these AI-enabled tools, this evidence will support rapid and broad adoption by healthcare delivery organizations.”

The study is expected to last one year. The results will be of interest to several audiences, including health economists, policymakers, and AI researchers. To reach them, researchers plan to share their results in peer-reviewed papers and conferences related to health economics, health-services research, and machine learning. They are also planning to engage policymakers like CMS and others through public commenting and in-person briefings.