Health OS City

Health OS

A city for testing the future of care

2036Next
HealthyAt risk Managed conditionUnmanaged condition On the way to hospitalCare at home

How this model works

Health OS is a teaching sandbox for keynotes, workshops and early design conversations. It is a synthetic city with illustrative numbers. It is not a forecast, a validated digital twin, or a basis for patient, purchasing or reimbursement decisions.

What you are looking at

  • Six districts. Each roof stands for a share of that district's residents: healthy, at risk, living with a managed condition, or living with an unmanaged one.
  • Everything drawn comes from the model's output for that month. The line outside the hospital follows the wait for care. Red hospital windows follow team strain. Trees follow prevention reach. Neighborhood hubs rise as access improves. Roof antennas follow care moved home. Links from the data hall follow how integrated the system is.
  • The three bands in the ground are what is needed first: trust and governance, workflow and adoption, equitable access.
  • Drag the divider to see the same city under sick care as usual.

The engine

A deterministic cohort model: six districts, five states (healthy, at risk, managed chronic disease, unmanaged chronic disease, death) and 120 monthly steps from 2026 to 2036. The population is closed, with no births or migration. Every plan is compared with a matched reference that keeps today's fragmented system. Every coefficient is a hypothesis chosen for teaching, not an estimate from the literature.

What changed from v0.1

  • Now, Near, Next scheduling. Each capability, technology and building starts in 2026, 2029 or 2032. Spending starts when the work starts and benefits ramp up after. Near and Next technologies cannot start before their act.
  • Order has consequences. Changes under way slow each other down and load care teams. Strained teams lose capacity. Trust is earned slowly and lost quickly.
  • Results per living resident. The health index and cost are measured across living residents, and care capacity follows the population. Healthy years gained and lives saved are running totals against the reference. In v0.1 the index fell in every scenario and cost drifted down by itself, because both were divided by the starting population.
  • Costs anchored. All dollar figures are scaled by 1.7 so the reference city starts near $15,400 per resident, close to US health spending of $15,474 per person in 2024 (AMA summary of the CMS national health expenditure accounts). Dollars are constant and there is no medical inflation.
  • Mechanisms v0.1 left out. Human care capacity now improves chronic-disease control and trust as well as throughput. Prevention reduces relapse and admissions more directly, and is priced at $520 instead of $740 per 100 index points before scaling. Public health has a small everyday benefit. On its own, prevention now costs about $127,000 per healthy year at ten years, in the same range as the full plans. That matches the broad finding that cost-effectiveness is distributed similarly for prevention and treatment (Cohen, Neumann and Weinstein, N Engl J Med 2008;358:661-663).
  • Not carried over yet. The v0.1 optimizer, sensitivity sweep and scenario export. The optimizer's default weights valued a healthy year at about $13,000, which is why it recommended cutting prevention and the workforce.

What the model tends to say

  • Full redesigns cost roughly $80,000 to $105,000 per healthy year gained at ten years, and about half that at twenty. Better is not automatically cheaper.
  • Trust, workflow and aligned incentives are the best buys, because they make every other investment work.
  • AI layered on weak foundations costs almost as much as a balanced transition and buys no health.
  • The same investments in a different order give different results.

Limits

Nothing here is calibrated to a real health system. There are no specific diseases, treatments, payers or prices. The equity measure is only the gap in access between districts. Keeping sicker people alive can lower the average index among the living, so read it alongside lives saved. Shocks are stylized.

Keys

Space plays and pauses. Left and right arrows move a year, or a step when the story is running. C compares with sick care, S starts the story, F toggles stage view, Escape closes.

Concept: Daniel Kraft, Reimagining the Health OS. Engine v0.2.0, synthetic prototype.