The United States spends roughly $16 billion annually through Medicare to fund graduate medical education—the residency training that turns medical school graduates into practicing physicians. That funding system was largely set in place by the Balanced Budget Act of 1997, which capped the number of residency slots Medicare would support at each hospital.

Nearly three decades later, those caps remain mostly intact. The geographic distribution of funded slots still reflects where teaching hospitals happened to be training residents in the mid-1990s, not where physicians are needed today. The result is a workforce pipeline shaped more by historical accident than by deliberate planning.

This creates a fundamental tension in American health policy. We know we face physician shortages in primary care and in rural and underserved communities. Yet the single largest public investment in physician training operates on autopilot, channeling billions into patterns that predate the smartphone era. Understanding how this system works—and why it resists change—is essential for anyone concerned about healthcare access.

Medicare funds graduate medical education through two distinct payment streams. Direct Graduate Medical Education (DGME) payments reimburse teaching hospitals for the direct costs of running residency programs—resident salaries, benefits, faculty supervision, and administrative overhead. Indirect Medical Education (IME) payments compensate hospitals for the higher patient care costs associated with training, such as the additional tests and longer stays that occur in teaching environments.

The formulas behind these payments are where things get complicated. DGME payments are calculated using each hospital's historical cost-per-resident amount, updated for inflation but anchored to what hospitals spent decades ago. IME payments are added as a percentage adjustment to a hospital's regular Medicare reimbursement, tied to the ratio of residents to beds. Hospitals with more residents relative to their size receive proportionally larger adjustments.

Crucially, both streams are subject to the resident caps established in 1997. A hospital that was training 200 residents when the caps took effect can receive Medicare funding for 200 residents—regardless of whether community needs have changed or the hospital has capacity for more. New teaching hospitals or programs that have expanded since then must fund additional slots from other sources, a significant financial barrier.

The IME adjustment also deserves scrutiny. Research has consistently shown that the IME payment substantially exceeds the actual indirect costs of training. The Government Accountability Office and the Medicare Payment Advisory Commission (MedPAC) have both documented this overpayment, estimating that a significant portion of IME funds effectively functions as a general subsidy to teaching hospitals rather than a training-specific investment. This means billions flow to institutions without clear accountability for workforce outcomes.

Takeaway

When you anchor a funding system to a historical snapshot, you don't just preserve the past—you actively reproduce it. The allocation mechanism itself becomes the policy, regardless of what policymakers intended.

The geographic consequences are stark. The 1997 caps froze funding in states that already had dense concentrations of teaching hospitals—primarily in the Northeast and along parts of the West Coast. States like New York and Massachusetts receive GME funding far out of proportion to their populations, while fast-growing states in the South and West remain chronically under-resourced for training. Since physicians tend to practice near where they trained, this funding geography directly shapes the physician workforce map.

Specialty distribution is equally distorted. Large academic medical centers, which receive the lion's share of GME funding, have strong institutional incentives to train subspecialists. These physicians generate higher revenue for the hospital, their programs attract research funding, and they enhance institutional prestige. The funding formula does nothing to counteract these incentives—a hospital receives the same per-resident support whether it trains a family medicine physician destined for rural practice or a cardiac electrophysiologist bound for a metropolitan center.

The result is a persistent mismatch. The Association of American Medical Colleges projects significant shortfalls in primary care physicians over the next decade, and the Health Resources and Services Administration has designated thousands of communities as primary care shortage areas. Yet the financial architecture of GME continues to favor the production of specialists in well-resourced urban settings.

There's also a less visible institutional distortion. Because the caps lock in historical slot counts, they create a kind of property right for incumbent teaching hospitals. These slots have real financial value—each funded resident position can represent hundreds of thousands of dollars in annual Medicare payments. This makes hospitals fiercely protective of their existing allocations and resistant to any redistribution, even when neighboring communities face acute physician shortages.

Takeaway

A system that rewards what already exists will always overproduce what we already have. Without explicit incentives tied to workforce needs, the default output of GME funding is more of the same—in the same places, in the same specialties.

Proposals for GME reform are not in short supply. MedPAC has recommended reducing the IME adjustment to better reflect actual indirect costs and redirecting the savings toward a performance-based funding pool. Under this model, hospitals would receive bonus payments for training residents in shortage specialties, placing trainees in underserved communities, or demonstrating that graduates practice in high-need areas. The idea is straightforward: tie at least some funding to the outcomes we actually want.

Other proposals go further. Some advocates call for shifting GME funding away from Medicare entirely, consolidating it into a dedicated workforce development fund administered by a body with an explicit workforce planning mandate. This would sever the historical link between hospital Medicare volume and training support, enabling funding to flow to community health centers, rural hospitals, and ambulatory settings where an increasing share of medicine is actually practiced.

Congress has also periodically considered simply lifting the caps. The Resident Physician Shortage Reduction Act, introduced in various forms over recent years, would add thousands of new Medicare-funded positions with requirements that a portion go to primary care and underserved regions. While politically popular—it gives something to everyone—critics argue that adding slots without reforming allocation simply amplifies existing distortions at higher cost.

The deeper problem is political economy. Teaching hospitals are major employers and economic anchors in their communities, giving them enormous lobbying influence. Any redistribution of GME funding creates identifiable losers—specific hospitals that would see payments decrease—while the beneficiaries are diffuse and often hypothetical. This asymmetry of political pressure explains why GME policy has remained largely frozen despite decades of expert consensus that reform is overdue. The policy window, as John Kingdon might observe, has yet to open wide enough for the problem, policy, and politics streams to converge.

Takeaway

Knowing the right policy is not the same as achieving it. When the beneficiaries of the status quo are concentrated and organized while the beneficiaries of reform are dispersed and future-facing, inertia has a structural advantage.

The federal GME funding system is a case study in how policy infrastructure outlasts the assumptions that created it. A framework designed for the healthcare landscape of 1997 continues to shape physician training in ways that compound workforce shortages rather than address them.

Reform doesn't require dismantling the system—it requires reconnecting it to purpose. Tying funding to workforce outcomes, expanding where and how training happens, and redistributing resources toward documented needs are all technically feasible. The barrier is political will.

For anyone who cares about healthcare access, GME funding is not an obscure technical issue. It is the upstream decision that determines which communities will have physicians a decade from now—and which will not.