Yale preprint projects Medicare for All would cut US health spending $1.04T and avert 114,000 deaths annually

The Yale School of Public Health team, led by Alison Galvani, modeled what would happen if the United States adopted a national public insurance program along the lines of the Medicare for All Act. Against 2024 spending, coverage, and mortality data, they project $1.04 trillion in net annual savings, roughly 20% of current expenditure, and 114,174 deaths averted each year. The savings figure already nets out an estimated $304 billion in additional spending to cover unmet medical needs, reimburse uncompensated care, and provide universal dental coverage. The deaths-averted total combines roughly 62,863 from expanded coverage with about 51,311 from reversing post-2025 coverage rollbacks. Nearly half of the coverage-driven deaths averted, approximately 29,631, would be among working-age adults who already hold insurance but cannot afford the care their plans require.

The model decomposes savings into five buckets: lower pharmaceutical prices, Medicare-level payments to providers, reduced administrative overhead, less fraudulent billing, and fewer avoidable emergency department visits and hospitalizations. A conservative scenario that assumes less aggressive drug-price negotiation and smaller fraud reduction still yields $663 billion in annual savings, which the authors treat as a floor rather than a central case. The study extends earlier work the same group published in The Lancet in 2020, which projected $450 billion and 68,000 lives saved per year. The larger current figures reflect higher baseline health expenditures, a wider gap between commercial and Medicare payment rates, and updated estimates of the underinsured population, which the authors put at over 45 million working-age adults.

The authors are explicit about the model's limits. Direct estimates of excess mortality among underinsured adults do not exist, so that number is itself modeled rather than measured from data. The spending analysis excludes transition costs, administrative job losses, and the behavioral response of hospitals and physicians to Medicare payment rates, all real-world effects a static model cannot capture. The paper is a preprint and has not been peer reviewed. The authors' own framing, that the United States already spends enough to provide universal coverage and the problem is allocation rather than invention, is a policy argument built on top of the model rather than a finding of the model itself.

Despite the 'AI' topic tag attached to this event, the underlying paper is a health economics simulation, not a machine learning or software engineering result. There is no model release, no benchmark, and no code described in the material provided. The single feed carrying the story surfaced it on Hacker News as a comments thread rather than a primary technical discussion, consistent with it being a policy paper rather than an engineering one. The only angle of interest for someone who builds software is methodological: it is an example of how a public dataset plus a transparent set of assumptions can produce a widely cited, politically consequential projection, and how the assumptions one chooses to include or omit, such as transition costs and provider response, shape the headline number that propagates into news coverage.