How much can we trust artificial intelligence (AI)? How much could AI transform an industry as stodgy as healthcare, where other technologies have failed time and time again? These questions were far from mainstream thought until just a few years ago, when the current wave of AI innovation captured the attention of the public, industry, investors, regulators, and everyone in between.
Now, it's all but a foregone conclusion that AI will fundamentally alter the economy and world we live in, with many industries already undergoing dramatic transformation.
Healthcare is one particularly important industry undergoing transformation with the advance of AI. Healthcare is generally expensive, especially in the United States. The Centers for Medicare & Medicaid Services (CMS) report a 4.1% increase in US healthcare spending in 2022, reaching a total of $4.5 trillion. Although that may seem like a modest amount, the US economy over the past 10 years has generally grown around 2% annually, indicating that we spend relatively more on healthcare every year. Projections from CMS suggest that this rate of growth in healthcare spending will continue, with healthcare spending reaching $6.8 trillion by 2030.
Given this ever-inflating price tag, there is a huge appetite for innovation to make high-quality care more affordable and accessible. The US healthcare system as a whole has largely functioned through a fee-for-service system, paying for services rendered using formulas that factor in time and resources. The transition to value-based care, or fee-for-value, has had a long and difficult history spanning many decades. Although this transition to value-based care has seen varying degrees of success, there have been many more failures related to a variety of factors, including discordance among stakeholders about how to define "value," disparity between payment and costs of delivering high-quality care, and difficulty measuring and assessing quality and outcomes.
AI has emerged as a tool poised to accelerate the transition to value-based care. At their core, all the various types of AI are designed to reduce the burdens of labor, to automate what can be automated to allow for more efficient allocation of workers and their time. As these technologies begin to permeate healthcare, they will lower unit costs for care delivery by decoupling the labor-intensive input required to perform key healthcare services from the health outcomes that those services are designed to deliver. This reduction in the cost of care delivery should in theory act as a pressure release valve to temper historical management mandates from various actors in the healthcare industry — mandates meant to extract every nickel and dime of revenue possible to maximize profits, because costs were previously generally high and fixed. This inflection point will thus provide an opportunity to use the greater amount of headroom to better align incentives to prioritize care that results in improved health outcomes and a better patient experience rather than just rewarding churning out as many procedures as possible.
Although this transformation isn't likely to replace doctors, it will allow physicians to focus their time and attention in ways that prioritize high-quality care. This would manifest as AI-driven technologies supporting physicians and other providers to practice at the top of their license and training. Examples include AI technologies that empower radiologists to read more diagnostic imaging studies more efficiently; primary care providers to manage larger patient panels without having to pore over every detail of every patient chart and recommended preventive service manually; anesthesiologists to more accurately predict or proactively respond to intraoperative risks; and surgeons to perform more complex surgeries through the use of computer vision guidance or robotic support.
Another area of dire need within the healthcare space is curtailing of administrative bloat. Administrative spending is estimated to range from 15%-30% of total healthcare spending annually, with recent estimates reaching $1 trillion. Unfortunately, up to half of this spending has been characterized as wasteful, which of course does nothing but contribute to rising healthcare spending and cost. As in other industries, AI has the potential to reduce administrative spending through automation, with some estimating healthcare savings as high as $200-$360 billion, using existing technologies realized in the next 5 years. This is particularly important in federal and state government programs such as Medicare and Medicaid, which rely on precious tax dollars for funding that should be directed with the utmost care.
We are reaching new milestones every day, with AI technology advancing at breakneck speeds and already producing applications throughout healthcare, including drug discovery, diagnostics, and ambient scribing for providers, among many others. At this early point in the healthcare AI revolution, many advances have been through the application of general technologies to healthcare, but we are seeing a growing need to build healthcare-specific tools because building for healthcare poses unique and complex challenges. Examples include large language models focused on safety and built around patient-provider relationships and AI models built by health systems to lighten burdens specific to patient populations and clinical operations.
With computing power doubling roughly every 6 months for large AI models and the market for AI projected to expand at a staggering clip, we can only expect more on the horizon. As we learn to harness these paradigm-shifting technologies to unlock new potential for our healthcare system, we have every reason to be hopeful.
Robert D. Glatter, MD, is an assistant professor of emergency medicine at Zucker School of Medicine at Hofstra/Northwell in Hempstead, New York. He is a medical advisor for Medscape and hosts the Hot Topics in EM series.
Brian J. Miller, MD, MBA, MPH, is a hospitalist and an assistant professor of medicine at the Johns Hopkins University School of Medicine. He is also a nonresident fellow at the American Enterprise Institute. From 2014 to 2017, Dr Miller worked at four federal regulatory agencies: the Federal Trade Commission, Federal Communications Commission, Centers for Medicare & Medicaid Services, and Food and Drug Administration.
Ted Cho, MD, MBA, is a pediatric resident physician at the University of California, San Francisco. He previously completed his dual MD/MBA degree at Georgetown University.