Nearly $60 million flowed from artificial intelligence device companies to American physicians between 2017 and 2023, according to the first comprehensive analysis of industry payments tied to AI-powered medical devices. The findings reveal a troubling pattern: payments are concentrating at well-funded academic hospitals while rural and under-resourced facilities are being left behind, potentially widening healthcare disparities as AI becomes increasingly central to modern medicine. Researchers at the University of Pennsylvania's Leonard Davis Institute of Health Economics tracked these payments and discovered that cardiologists, brain surgeons, and radiologists were the biggest recipients. The analysis, published in Health Affairs Scholar, raises urgent questions about how artificial intelligence will be distributed across American healthcare—and whether it will help close gaps in care or deepen existing inequalities. Who's Getting the Money, and Why It Matters? The $60 million in payments went to approximately 46,000 doctors, but the distribution was far from even. A handful of companies dominated the spending landscape. Johnson & Johnson's cardiac device unit, Biosense Webster, accounted for 82.5% of all payments in cardiac electrophysiology, the medical specialty focused on the heart's electrical systems. Medtronic, another device giant, spread its payments across multiple specialties including neurosurgery (23.1% of their AI device spending) and orthopedic surgery (18.5%). The concentration of payments matters because research shows doctors are more likely to use devices from manufacturers who pay them. When a handful of companies control most of the payment flow, they gain outsized influence over which AI tools hospitals adopt and which physicians learn to use them. The Geography of AI Inequality Perhaps the most concerning finding is where these payments went. The vast majority flowed to physicians working at larger, urban teaching hospitals—institutions that already have robust research programs and the financial resources to invest in cutting-edge technology. Rural hospitals and under-resourced urban facilities received minimal payments, suggesting that artificial intelligence adoption will likely follow existing healthcare inequalities rather than help remedy them. "We're laying the groundwork for how AI will function in medicine," explained Alon Bergman, one of the study's lead researchers. "How it's built will determine whether AI closes gaps in care or widens them." This geographic divide has real consequences. Health systems lacking their own robust evaluation capabilities may become even more dependent on vendors when making purchasing decisions about which AI tools to buy. Without targeted funding and infrastructure support, smaller hospitals could fall further behind in adopting AI technology, creating a two-tiered system where wealthy institutions get the latest diagnostic tools while others make do with older technology. The FDA Approval Problem Nobody's Talking About Behind the payment story lies an even deeper concern: the regulatory framework governing AI medical devices is full of gaps. By mid-2025, over 1,200 AI-enabled devices had received Food and Drug Administration (FDA) approval, representing a 350% increase in clearances over five years. But most of these devices were approved through the FDA's 510(k) pathway, which requires manufacturers to show only that a new device is substantially equivalent to a previously approved product. The problem is that those predicate devices often lack artificial intelligence, were trained on limited patient data, and have minimal trial evidence proving their effectiveness. Additionally, many AI tools now bypass FDA review entirely—such as AI scribes that record medical encounters—because they aren't classified as medical devices at all. - Regulatory Gap: Most AI devices approved through the 510(k) pathway, which only requires showing similarity to older devices that often lacked AI components. - Unregulated Tools: AI scribes and other ambient technologies widely used in clinical practice receive no FDA review because they fall outside the medical device classification. - Evolving Technology: AI-enabled devices continue to learn and adapt after approval, but the FDA approval process treats them like static products that don't change over time. - Limited Validation: Many AI tools lack robust testing in racially and socially diverse patient populations, raising concerns about bias and effectiveness across different communities. Steps Policymakers Should Take to Ensure Fair AI Adoption The researchers behind this analysis offer concrete recommendations for how the healthcare system can ensure AI benefits everyone, not just wealthy institutions: - Transparent Reporting: Require device makers and clinicians to publicly disclose industry relationships so hospitals and patients understand potential conflicts of interest. - Reimagine FDA Oversight: Develop new approval processes that recognize AI devices continue to learn and evolve, requiring different evaluation standards than traditional medical devices. - Fund Rural and Urban Hospitals: Direct targeted funding and infrastructure support to smaller hospitals so they can adopt and implement AI medical devices without becoming overly dependent on vendors. - Strengthen Transparency Standards: Establish clear requirements for explainable AI, transparent datasets, and multidisciplinary oversight to ensure AI systems remain trustworthy and unbiased. The stakes are high. As artificial intelligence becomes increasingly central to diagnosis, treatment planning, and patient care, the decisions made today about how AI is distributed and regulated will shape healthcare inequality for decades to come. Without intervention, the current pattern of payments concentrated at wealthy institutions suggests that AI will amplify existing disparities rather than help bridge them. "We need to send funds and infrastructure to help weaker rural and urban hospitals adopt AI medical devices," the researchers emphasized. The alternative is a healthcare system where cutting-edge AI-powered diagnosis and treatment become privileges of the wealthy while under-resourced communities continue relying on older technology.
