The US health workforce gap is at least 6× larger than the headline.

The problem

HRSA's National Center for Health Workforce Analysis publishes the official US health workforce projection through 2038. Its 2025 release pegs the 2035 physician shortfall in metro areas at ~52,000 FTEsunder the published “Status Quo” scenario. That number is the one reporters and policymakers cite.

But HRSA's own technical documentation concedes the Status Quo is a lower bound. Three critiques are baked into the method itself:

1. Demand is defined as utilization, not need.HWSM Ch. I: “Workforce demand is defined as the number of health care workers required to provide a level of services that will be utilized given patient health-seeking behavior, and ability and willingness to pay.” And explicitly: “demand is different from need. Demand reflects the level of care that people are likely to use, while need is usually a clinical definition.”
2. Historical baseline assumes equilibrium.HWSM Ch. XIII: “Historically, HWSM operated under the assumption that national demand equals national supply in the starting year for most health professions.”
3. Zero scenarios for AI, scope, staffing ratios, or financial impact.HRSA ships 13 what-if scenarios. None model AI productivity, NP scope-of-practice expansion, mandatory nurse staffing ratios, team-based care leverage, telehealth capacity, induced demand, insurance shifts, or GDP impact. Not because the model couldn't — because the microsim doesn't have the policy surface to express them.

Our finding

We ingested HRSA's FY2025 data file verbatim (21/21 spot-checks pass — see Compare to HRSA) and built a three-layer extension platform that lets anyone dial in literature-backed parameters and watch the gap — and its dollar cost — recompute live.

For All Physicians / US Metro / 2035, stacking just our three headline extensions:

• E1 Need-Based Baseline — demand is scaled by a per-profession multiplier derived from unmet-need literature (Borsky et al., Health Affairs 2018 — only 8% of US adults receive all recommended preventive services; CDC NHANES 2024 — 20.7% of hypertensive adults have BP controlled; SAMHSA NSDUH 2023 — 22.8% of adults have AMI but only 23% received treatment).
• S1 AI Productivity — +10% effective clinician supply from ambient AI scribes (Kaiser Permanente 7,260-physician JAMIA 2024 deployment saved ~16,000 hours; NEJM AI 2024 longitudinal study found no net productivity gain; the default sits at the midpoint).
• S3 Mandatory Staffing Ratios — +17% RN demand by 2030 (California AB 394; Oregon HB 2697; Spetz et al., Health Affairs 2013).

…produces a 2035 physician gap of −352,720 FTEs — 6.7× HRSA's headline.

What this costs

Layer 3 translates the gap into dollars using health-system P&L benchmarks (Kaufman Hall, NSI, Moody's), preventable hospitalization rates (AHRQ PQI), and GDP multipliers (McKinsey Global Institute, Deloitte):

• 2035 annual health-system impact: −$117.8 billion (revenue loss + premium labor + turnover cost)
• 2035 annual preventable hospitalizations: 14.1 million
• 2035 annual GDP drag: −$805 billion
• Cumulative 2023–2038 health-system impact: −$1.69 trillion
• Cumulative 2023–2038 GDP drag: −$11.5 trillion

For context: McKinsey Global Institute (2020) estimated that poor health reduces global GDP by ~15% per year and that equity improvements could add $12 trillion by 2040. Deloitte (2024) projected US health inequity costs could exceed $1 trillion by 2040.

The 14 extension knobs

The Scenario Builder exposes every knob as a slider with its top citation. The six demand-side levers (E1–E6) let you replace utilization-anchored demand with need, override disease trajectories, model induced demand, add AI-driven visit friction reduction, select Census aging trajectories, and model insurance coverage shifts. The eight supply-side levers (S1–S8) let you dial AI productivity, scope of practice, staffing ratios, team-based care leverage, IMG immigration, training pipeline growth, burnout early-exit waves, and telehealth capacity gains.

Every knob is a pure function over the HRSA baseline. When all knobs are off, the extended output matches the HRSA baseline exactly (enforced by the composer's invariant and verified on the Compare to HRSA page).

The counter-argument

We've tried to be fair to HRSA. Two notes matter:

• HRSA is partially right that AI may not expand supply.The NEJM AI 2024 longitudinal study of DAX at a major academic medical center found no net EHR or financial productivity gain, even as physicians reported reduced cognitive burden. We default S1 to +10% (the Kaiser midpoint), and the slider goes down to 0 to reflect the null finding.
• AI may also drive demand up via Jevons paradox.Our E4 knob lets you add a demand uplift from AI-enabled triage and async care. If you believe AI productivity gains will be fully offset by induced demand, the extended gap doesn't shrink — it may even widen. This is easy to test: turn on S1 + E4 together and watch the extended demand curve rise faster than extended supply.

The platform doesn't decide which scenario is right. It makes every scenario legible, citable, and one URL away.

Explore live: Scenario Builder · Financial Impact · US State Map · Methodology & Sources.

Want the longer story behind the numbers? Read the evidence walk →

Underlying data: HRSA NCHWA Health Workforce Projections, FY2025 release, data currency 2025-12-18. All coefficients are in 2023 USD unless otherwise noted. Every extension default has a citation in data/docs/external_research.md (48 primary sources across alternative projections, need-vs-utilization literature, AI impact studies, staffing-ratio research, and financial-impact literature).