The Workforce Equation

The math is not close anymore.

HRSA's National Center for Health Workforce Analysis published the new nurse projections in December 2025. The model projects an eight percent national shortage of registered nurses in 2028, and a twenty-four percent shortage in nonmetropolitan counties in that same year. By 2038, even after fifteen years of pipeline correction, the nonmetro RN gap is still eleven percent. Metropolitan counties accounted for two percent in the same year. The rural-urban ratio runs four to five times across every interval, in every modeled scenario.

The shortage is structural. The pipeline is set well above any individual hospital's pay grade, and the people inside it are leaving faster than they are entering.

Across every role

The 2024 NCSBN National Nursing Workforce Survey surveyed 800,000 nurses and reported that 138,000 nurses left the workforce since 2022. Almost forty percent intend to leave by 2029. The share of registered nurses under age thirty dropped from 11.1 percent in 2022 to 7.9 percent in 2024. The licensed practical nurse share under thirty fell to 7.0 percent, the lowest since 2015. Of the nurses planning an exit, 41.5 percent named stress and burnout as the primary reason.

The downstream view is in the AHA's 2025 Health Care Workforce Scan. National RN turnover was 16.4 percent in 2024, down two points from the prior year, and hospital staff turnover was 18.3 percent. Certified nursing assistants and licensed practical nurses transition out of their roles within two years at roughly an eighty percent rate. Improvement is real. Recovery is not.

Physicians tell the same story with a different timeline. The AAMC's March 2024 report projects a national physician shortage of 13,500 to 86,000 by 2036, including 20,200 to 40,400 in primary care. 11 percent of physicians practice in rural counties, which account for 20 percent of the population. The bipartisan Senate Medicare GME Working Group's December 2024 draft proposes 5,000 new residency slots from FY2027 through FY2031, with 10 percent earmarked for rural programs. The bill has not been formally introduced in the 119th Congress. Even if it passes intact, the first cohort of new residents would not reach independent practice until well after 2030.

The administrative side broke quietly but no less completely. AAPC's 2025 salary and labor market reporting, paired with successive MGMA staffing surveys, identifies medical coding as the hardest revenue-cycle role to fill in a community hospital, with billers, schedulers, and authorization staff close behind. The work has only grown more complex as scrutiny of Medicare Advantage intensified and initial denial rates climbed past 11 percent industry-wide.

What the pipeline cannot do

People who have never run a hospital often assume the pipeline expands when demand rises. It does not. Nursing schools turn qualified applicants away every year for faculty and clinical placement reasons, not for lack of interest. Residency slots are federally rate-limited and have been for decades. Coder credentialing takes years and pays so poorly that strong candidates leave the path before the credential pays back. None of these constraints responds to one hospital raising wages. They are structural ceilings.

Mercer's 2028 update reads, at first glance, like good news. It projects a modest national surplus of registered nurses by 2028. Look at the state-level detail, and the picture changes. The surplus concentrates in a few growth states, while New York, Massachusetts, Kentucky, and Tennessee remain in deficit. HRSA's simulation model, using a different methodology, projects that ten states will have shortages above eight percent by 2038, led by California at twenty-two percent, with North Carolina and Georgia at twenty percent each. The geography is the story. The average tells the wrong story.

Why rural sits at the empty-seat case

The independent rural hospital does not run on slack. A twenty-five-bed Critical Access Hospital typically operates its entire revenue cycle with six to twelve people, most carrying three or four roles. Total IT staffing across many of these facilities runs two to five full-time equivalents. Rural RN vacancy rates exceed 15 percent in CAH and rural community hospital surveys, roughly double the NSI Nursing Solutions national benchmark. Black Book Research's 2024 work, summarized in Stratevora's ICP intelligence brief, found that seventy-five percent of rural hospitals cannot fill the essential revenue-cycle roles they have already funded. Sixty-three percent of providers report ongoing staffing gaps in those same departments.

The numbers compound. When a 25-bed CAH loses one coder and one prior-authorization specialist for six months, the headline staffing number understates the damage. The hospital runs with multiple core revenue-cycle workflows impaired. Initial denial rates rise. AR days extend. Bad debt grows. Cash days on hand drop below 45, the threshold rural finance analysts use to mark a closure-risk inflection point. 41.2 percent of rural hospitals operating at negative margins in the 2026 Chartis State of the State did not get there in one quarter. They got there as the workforce equation slowly stopped balancing.

The financial chain

The chain runs in one direction. An open coder seat or an empty prior-authorization role turns into a backlog within a week. Backlogs surface as initial denials and pended claims. Denials and pends extend AR days. AR drag erodes cash on hand. Cash erosion narrows the operating window. The operating window is the first to close when rural hospitals lose control of AR, cash, and service-line economics. The empty seat is not an HR issue. It is the first link in a financial chain that ends at a service-line decision.

When a seat is already empty, the labor math changes. In a fully staffed system, a tool that saves a clinician an hour a day eventually compresses headcount because the work that hour absorbed was already being done. In a hospital where the seat sits empty, the same tool fills it. It does not displace anyone because no one was in the chair to be displaced. The tool now performs work that was either being done poorly or not at all.

This is the divergence. The same technology lands as compression in well-staffed urban systems and as a multiplier in understaffed rural and community ones. Both effects are real. The political conversation collapses them into one story about AI taking jobs. The operational conversation has to hold both at once, because hospitals making capital decisions today need an honest read on which case applies to them.

What the dividend actually is

The clinical evidence has matured enough to talk about with numbers. A 2025 quality improvement study of forty-six clinicians using Nuance DAX Copilot ambient documentation reported a 20.4 percent decrease in documentation time per visit, from 10.3 minutes to 8.2 minutes. A separate multi-clinician evaluation of Abridge AI measured a drop from 6.2 to 5.3 minutes per encounter. A 2025 rapid review in JMIR AI synthesized real-world evidence across multiple ambient tools, settings, and countries, reporting documentation time savings that aggregate to two to three hours per day per clinician when ambient scribing is deployed at scale. None of these studies is vendor marketing. All are peer-reviewed evaluations published by independent academic groups.

McKinsey's Q4 2025 healthcare survey reported that every responding leader has some generative AI work in motion. Fifty percent of organizations have implemented at least one use case. Eighty-two percent expect or have already measured positive ROI on those investments. Separately, McKinsey's earlier opportunity modeling puts broader AI deployment at five to ten percent of U.S. healthcare spending, between $200 billion and $360 billion annually, with hospital-specific savings of $60 billion to $120 billion.

The dividend exists. The interesting question is where it goes.

In a hospital operating below its safe staffing line, the dividend funds the work that must remain human. The nurse who got two hours back at the end of a twelve-hour shift can use it to round, to coach a new graduate, to sit with a family that just heard hard news, or to leave on time and come back tomorrow. The coder who has algorithmic assistance for routine cases can spend the recovered hours on the complex cases that require a person. The dividend, in the empty-seat case, is the only realistic path to keeping the unit open, the OB ward staffed, and the emergency department covered after midnight.

In a hospital running above its safe staffing line, the dividend looks different, and the leadership conversation gets harder. Some seats compress. Some roles change. Some functions consolidate. That conversation deserves honesty rather than soft language. Executives in those settings owe their workforce a clear plan for reskilling and for what the AI dividend buys for the patients and the staff who remain.

Same technology. Different math. Different moral weight. The senior operator's job is to know which case applies in their building.

The senior operator's frame

Plan as if hiring will not keep up, because it will not. HRSA's projections already assume historical patterns of attrition, graduation, and labor force participation. Both attrition and graduation are trending in the wrong direction, per NCSBN. The bipartisan rural GME proposal would add residency capacity that lands after 2030 at the earliest, if it lands at all. Wage growth helps recruitment in any individual market, but it does not expand the national supply, which is capped by the pay scale of any one facility.

Design work around what each side does best. AI handles pattern matching at scale: denial prevention, coding suggestions, prior authorization preparation, eligibility verification, and real-time coverage status checking under the accelerated Medicaid redetermination cadence that followed the unwinding, ambient documentation, and clinical summarization. The volume and complexity of those tasks already exceed what a rural staffing model can absorb manually. Humans handle the work that needs judgment when the data is incomplete: bedside presence, hard family conversations, the difficult payer call, the leadership moment that decides a hospital's reputation in its community.

Sequence the deployment. Hospitals that have done this well did not start with the most visible clinical use case. They started where the work was most invisible and most expensive: the prior authorization queue, the denial worklist, the eligibility verification step that fails quietly and shows up later as bad debt. Those workflows produce measurable revenue within 90 days and free the human team to take on the next set of unfilled work. Ambient clinical documentation tends to come later, after the operational case for the dividend has been earned in dollars rather than borrowed on faith.

Measure what the dividend funds do, not just what they save. A pure cost-reduction view will miss the point. The dividend's job in an empty-seat hospital is to fill seats. Track the work that the open seats were supposed to do and report on whether that work is now happening. Denial rates, AR days, time to authorization, cash days on hand, and charity care screening completeness. Those are the executive scorecard. The hours-saved figure is the input, not the outcome.

Protect the dignity of the worker who stays. The worker who stays needs room to do the work that matters most, and the AI dividend should fund exactly that. Pretending the pipeline will refill, or deferring the AI question for another year, accomplishes neither.

For community and rural hospitals, the AI dividend is the only realistic response to a broken pipeline, a tightening payer mix, and the January 2027 CMS interoperability and prior authorization API compliance date that will not slow down. Forty-one percent of rural hospitals operated at negative margins in the most recent Chartis State of the State, and 52.2 percent in non-expansion states. The hospitals that close in the next twenty-four months will not lose because they failed to recruit harder. They will lose because they ran the old equation against the new variables, and it no longer balances.

The workforce equation has four variables: demand growth, pipeline yield, attrition, and the AI dividend. Three of those variables are moving against rural and community hospitals at the same time. The fourth is the only one inside a hospital CEO's control.

Start where the seats are already empty. Deploy AI first against the RCM workflows that bleed cash: denial prevention, prior authorization, and eligibility verification. Measure denial rate, AR days, and cash days on hand within ninety days. Move to ambient clinical documentation once the operational case has paid for itself.

Solve for it.

References

HRSA, National Center for Health Workforce Analysis. Nurse Workforce Projections, 2023-2038. December 2025.

National Council of State Boards of Nursing (NCSBN). The 2024 National Nursing Workforce Survey. Journal of Nursing Regulation, 2025.

American Hospital Association. 2025 Health Care Workforce Scan. November 2024.

Association of American Medical Colleges (AAMC). The Complexities of Physician Supply and Demand: Projections from 2021 to 2036. March 2024.

Mercer. Future of the U.S. Healthcare Industry: Labor Market Projections by 2028.

McKinsey and Company. Generative AI in healthcare: Adoption trends and what's next. 2024-2025.

AAPC. 2025 Medical Coding and Billing Salary Report. February 2025.

Chartis Center for Rural Health. 2026 Rural Health State of the State. February 2026.

Quality improvement evaluation of Nuance DAX Copilot ambient documentation in outpatient care, 46 clinicians, 2024-2025.