Send health workers where they help the most.
Health systems never have enough doctors and nurses to go everywhere. ACWIS helps answer one question well: given a fixed budget, where should the next clinician go so the most people gain care? It predicts where shortages will grow, measures where an extra clinician makes the biggest real difference, and turns that into a clear, explainable plan.
Three jobs, in plain terms.
Most tools stop at a dashboard that shows where things look bad. ACWIS goes two steps further: it works out where help actually changes outcomes, and then where to put a limited budget.
Predict
It forecasts where shortages of doctors and nurses will deepen next, place by place.
Prove
It measures where adding a clinician truly improves access, not just where numbers happen to move together.
Place
It spreads a fixed budget across places to help the most people, and explains every choice.
Sources: HRSA Designated HPSA Quarterly Summary, data as of 31 Dec 2025; AAMC, The Complexities of Physician Supply and Demand, 2024; WHO, Health Workforce.
Care looks different across the world, and so do the gaps.
ACWIS is built to work in very different health systems, from well-resourced hospitals to clinics that run on a single nurse. The goal is the same everywhere: get skilled hands to the people who need them.
Frontline clinician
In the operating room
The care team
Decisions with dataPhotographs illustrate roles and settings, not specific countries. Free images via Unsplash; studio portraits separately licensed.
The next clinician, and the next unit of budget, should go where the evidence shows it does the most good, not where a ranking or a negotiation happens to point.
Scarce clinicians, placed by ranking instead of evidence.
Under a fixed budget, every system faces the same question: where should the next clinician go, or which places should get an incentive, so the money helps most? Today that call is usually made by ranking places on a shortage score, by negotiation, or by habit, rather than by the measured effect of the help itself.
Data-rich systems
e.g. United StatesRecords are plentiful, yet clinicians are still placed largely by ranking a shortage score, and the real effect of a placement on local access is rarely measured well.
Data-scarce systems
e.g. NigeriaStaffing and access must be pieced together from a registry, surveys, and maps. Migration and insecurity cloud what is seen, and incentives are spent with little evidence of impact.
The gap is real, and it is measurable.
How many doctors and nurses each country has, per 1,000 people. Figures load live from the World Bank, so they stay current.
Source: World Bank, World Development Indicators (SH.MED.PHYS.ZS and SH.MED.NUMW.P3), latest available year per country. Dashed lines mark the world and OECD averages. Health-workforce data is reported yearly, not moment to moment; this shows the most recent figure the World Bank publishes.
Projected total U.S. physician shortfall. Source: HRSA, National Center for Health Workforce Analysis. Hover a bar for the figure.
U.S. counties grouped into ten equal bands by primary-care supply. Real data (County Health Rankings 2026). This is a correlation (r = −0.24), not proof of cause; isolating the causal effect on a specific program is what ACWIS does on partner data.
Down to the county.
Shortages are local. This map shows primary-care physicians per 100,000 people for every U.S. county with data. Counties in red have the fewest doctors per person and sit at or below the federal shortage line; green counties are the best served. Hover any county for its figure.
Source: County Health Rankings & Roadmaps 2026 release, from the Area Health Resources File (AHRF), U.S. Health Resources and Services Administration. Rate is primary-care physicians per 100,000 residents. The shortage line marks a 3,500:1 population-to-physician ratio, HRSA's primary-care threshold.
Doctors gravitate to cities. The gap does not close on its own.
The map is not random. Physicians concentrate where the amenities, incomes, professional networks, and training hospitals are, which means large cities and affluent suburbs. Rural and less-populous places are left short, decade after decade, because the choice of where to practise follows careers and lifestyles rather than need. This is why a market left to itself does not fix the shortage, and why deliberate, evidence-based placement has value.
Cities draw the doctorsPhoto: Joel Filipe / Unsplash
Rural areas are left shortPhoto: Acton Crawford / UnsplashIncentive programs exist to counteract this, from loan repayment to shortage-area designations, but they are typically handed out by ranking a shortage score, not by where an extra clinician would help the most people. That is the gap ACWIS is built to close.
The same budget, placed two ways.
A limited number of new primary-care physicians can be spread across the country's shortage counties in different ways. Set how many, then compare ACWIS, which places them to reach the most people, against the common approach of filling the worst-ranked counties first. This runs on real county supply and population data.
Illustrative allocation on real county supply and population data (County Health Rankings 2026), optimising a transparent access target: people in counties lifted above HRSA's 1:3,500 primary-care line. It demonstrates the allocation logic, not a causal health outcome; on a partner's program data ACWIS optimises measured causal benefit with the same engine. Clearing every county-level shortage this way takes roughly 1,500 physicians; the space between the two lines is the cost of allocating by rank instead of by benefit.
From raw signals to a plan you can question.
The same six steps run in any country. Only the data it reads and the local rules change.
Gather
Pull records, or registry, survey and map signals, into one place, and note what is measured and what is missing.
Forecast
Predict where shortages will grow, by role and place, and say plainly that it is a prediction.
Measure the effect
Estimate what adding a clinician really does at the eligibility line, with checks, not a guess.
Allocate
Spread a fixed budget to help the most people, learning a rule rather than ranking raw numbers.
Explain
Attach a plain-language report, a confidence score, and the assumptions behind every recommendation.
Govern
Label every output with how it was produced, and refuse to run on stand-in data pretending to be real.
The gain is in where the workers go, not how many.
Health care is nearly a fifth of U.S. GDP and the largest single source of new jobs, so the sector’s scale is not in question. The economic case for ACWIS is about allocation, not size: the same budget and the same new clinicians produce very different amounts of health depending on where they land. A clinician placed where the marginal return is highest reaches more people and prevents more avoidable, costly illness; the same clinician placed by ranking or habit leaves that value on the table. Reducing that misallocation, the deadweight loss of putting scarce skill in the wrong place, is the return ACWIS targets.
A growing workforce; the question is where it landsPhoto: National Cancer Institute / UnsplashProjected employment change, 2023–2033. Source: U.S. Bureau of Labor Statistics, Employment Projections. "All occupations" is the economy-wide average.
The chart shows where the new workers are; it does not decide where they go. That decision, made clinician by clinician and dollar by dollar, is where a limited budget either reaches the most people or does not, and where a healthier, working population is either supported or left to chance.
Built on real data.
Every figure and relationship on this page is computed from real, public sources, cited where it appears: the World Bank, HRSA, County Health Rankings, the AAMC, WHO, CMS, USDA Economic Research Service, and the U.S. Bureau of Labor Statistics. What ACWIS adds for a partner is the step from these public patterns to a program-specific causal estimate and an allocation plan, produced on the partner’s own data. That pilot, on real program data, is the collaboration we are looking for. You can inspect the method itself: ACWIS forecasts shortages with gradient-boosted models, estimates causal effects with a regression-discontinuity design at a program’s eligibility cutoff (cross-checked against double machine-learning), and allocates with budget-constrained policy learning.
The commitments built in
- Every result carries a label for how it was produced. A prediction is never dressed up as proof, and a model estimate is never shown as certainty.
- A safeguard stops the system from running on substituted data and reports what is missing instead.
- The values in any plan, how much weight goes to access, fairness, and cost, are visible choices you set, not hidden in the maths.
- Every recommendation supports a human decision. It does not replace one.
Who is behind this.
I build ACWIS, an AI system at the meeting point of economics, causal inference, and health systems. The aim is simple: scarce clinicians are too often placed by ranking or habit, when the data and methods now exist to place them by measured benefit. This page explains the system in plain terms and shows it running on real, public data, and invites honest, expert feedback from the people who know these systems best. If you see something wrong, I want to hear it.
Share your feedbackTell me what you think.
ACWIS is in active development, and it gets better with critical, expert input. If you work in health economics, workforce planning, clinical care, or health policy, or you might use a system like this, your view is valuable. Rate what you can, say what is weak as freely as what is strong, and add your name so I can follow up.
Thank you, your feedback has been received.
If you left an email and opted in, I may follow up with a question or two. You can close this page.