Market analysis · for investors
U.S. healthcare economics, the failure of digitization, and the operational-control opportunity. The system is not under-resourced — it is mis-operated. This is the objective case for the problem, and the market case for the layer no incumbent ships.
Sections I–IV are an objective reading of U.S. healthcare built entirely on public data — CMS, OECD, CDC, the Commonwealth Fund, JAMA/NEJM, ACEP, AHA — cited inline. Section V is the authors' market interpretation. Any figure drawn from the Nightingale OS digital pilot is a modeled projection, not an observed outcome, and is labeled as such.
Executive summary
In 2023 the United States spent $4.9 trillion — 17.6% of GDP — and the federal actuary expects that share to keep climbing toward 20%. For that money the country buys a life expectancy 3.7 years below comparable nations, the highest maternal and infant mortality in the high-income world, and a health system the Commonwealth Fund ranked dead last among ten wealthy peers in 2024.
The thesis
Healthcare does not need a new idea. It needs an old one every other complex, high-stakes domain already adopted: operational and logistical control over the entire system. The military perfected it; its core mathematics — queuing theory — explains precisely why hospitals fail. The category that wins will not be a better dashboard bolted onto the record. It will be the system of action that sits above the records, coordinates the whole operation, and is auditable by design. Nightingale OS is built to be that layer.
I · The macroeconomic case
Health spending was about 5% of GDP in 1960. It has roughly tripled as a share of the economy over six decades, and the federal actuary projects it reaches 19.7% of GDP by 2032. Americans do not consume dramatically more care than their peers — the gap is higher prices and higher administrative and operational overhead per unit of care.
The U.S. spends ~$14,775 per person — about 88% more than the ~$7,860 peer average — yet life expectancy (79.0) sits 3.7 years below the peer average of 82.7. Peterson-KFF; CDC/NCHS
The Medicare Hospital Insurance trust fund is projected to deplete in Q2 2033 — covering only 89% of benefits. That date moved three years closer in a single report. Medicare Trustees, 2025
~38% of Americans skipped or delayed needed care over cost in 2024; ~100 million carry at least $220B in medical debt. Commonwealth Fund; KFF
The landmark estimate put waste at $760–935 billion a year — roughly 25% of all health spending. The largest, most addressable categories are administrative and operational, not clinical overtreatment. Such waste does not shrink when you spend more. It shrinks when you operate better — an engineering and logistics problem, and a market.
II · Where it breaks, at the point of care
Macroeconomic waste is an abstraction until it is traced to the floor of a hospital — where it appears as a collapsing workforce, a documentation burden that drives clinicians from patients, a capacity system that jams, and a steady stream of preventable harm.
RN turnover hit 17.6% in 2025 at $60,090 each — the average hospital loses $4.2–6.2M a year to nurse turnover alone. Understaffing is backfilled with agency labor at ~$91/hr vs $59 for staff, and each extra patient in a nurse's load raises 30-day mortality odds ~16%.
NSI, 2026 · Aiken/Lasater et al.
For every hour with patients, physicians spend ~2 more on the EHR plus 1–2 hours of after-hours "pajama time." The bedside nurse is worse off — the human integration layer between every system in the building, doing in their head the joins a database does in milliseconds.
Sinsky et al., Annals 2016
Occupancy rose from ~64% to 75% as the staffed-bed supply fell ~128,000 (≈160 hospitals). ED boarding over three hours rose from 22% (2017) to 36% (2024); ~a quarter of inpatient days — 10.8M a year — are avoidable.
UCLA/JAMA Network Open, 2025 · EDBA · Advisory Board
23.6% of admissions — nearly one in four — involve an adverse event, 22.7% of them preventable. Since 2008 Medicare won't reimburse the worst "never events," so the hospital pays twice: first to understaff, then for the harm understaffing produces.
Bates et al., NEJM 2023 · CMS HAC policy
| Never event | U.S. incidence / year | Cost to hospital / event | National bill |
|---|---|---|---|
| Inpatient falls (with injury) | 700K–1M falls; 30–35% injure | ≈$62,000 | ≈$20B |
| Pressure injuries (HAPI) | ≈2.5M cases | ≈$21,800 (to $150K severe) | ≈$26.8B |
| CLABSI (central-line) | Tens of thousands | ≈$48,000 | — |
| CAUTI (catheter UTI) | Hundreds of thousands | ≈$13,800 | — |
| All hospital-acquired infections | ≈687,000 infections | — | $28–45B |
These are not random misfortunes but the measured downstream of how the operation is staffed and run — the events real-time monitoring, protocol enforcement, and closed-loop workflow are built to prevent.
III · Why the technology hasn't worked
It digitized records and optimized fragments while the operation as a whole stayed uncoordinated. Understanding how the existing tools fall short is the key to seeing the opening.
HITECH spent more than $35 billion to drive EHR adoption from 3% of hospitals to 96%. On its own terms a spectacular success — and it did not bend the cost curve, improve outcomes, or reduce burnout. It made them worse. The EHR was built as a billing and legal record: a digital filing cabinet, not an instrument for running the operation. Digitizing a filing cabinet yields a faster filing cabinet, not a coordinated hospital.
Command centers, ambient scribes, schedulers — each a local optimum. Optimizing every component in isolation does not optimize the system; it just moves the bottleneck. The hospital needs one layer that coordinates the gardens, not ten more to tend.
The most-deployed predictive model — the Epic Sepsis Model — externally validated at AUC 0.63 vs its advertised 0.76–0.83, missing ~two-thirds of cases. As vendors pour in more black-box models, clinicians trust them less, and regulators have noticed.
Investment chased the physician's diagnostic work — where the human still usually wins — and skipped bed assignment, staffing math, throughput, registry assembly, and compliance evidence, where a machine plainly does better.
The category error. The industry spent two decades optimizing the productivity of individual humans, when the science of operations says the lever is system flow. You cannot fix a queuing system by making each server work harder — past a point, that makes it worse. The resulting burnout is not a personal-resilience problem. It is a design problem.
IV · The idea that isn't new
It is the discipline every other complex, high-consequence system — military, aviation, logistics, manufacturing — adopted decades ago and that healthcare, almost uniquely, never did. Operations research was born in WWII to coordinate enormous, variable, life-or-death systems under uncertainty. The military's genuine expertise is not weaponry; it is logistics. That is the capability a hospital lacks.
Administrators, taught to treat empty beds and idle staff as waste, push utilization onto the steep part of the curve — where a single surge or a few call-outs tips the unit into gridlock, diversion, and harm. The slack queuing theory says a variable system requires is treated as the first thing to cut. "It works until it doesn't" is a theorem, not an opinion.
Automate the analog · unlock the human
You cannot fix a queue by working humans harder. You manage the system for flow and remove every task that does not require a human. A nurse relieved of the integration-layer burden is additional capacity created without additional hiring — the only kind a workforce-constrained system can actually add.
Reference points: ambient AI cut clinician burnout 40% at Mass General Brigham and documentation latency 81% at Cedars-Sinai; the company's own pilot observed a 35% documentation-time cut (modeled).
V · Market positioning & opportunity
Hospital software today is a stack of systems of record — destinations a clinician navigates to enter data. The unoccupied category is the system of action that sits on top: a governed layer that consumes the silos, holds a live model of the operation, predicts the predictable, and surfaces the next action — already populated — to the accountable person, on the right device, at the right moment. No incumbent ships this.
If the staff are the integration layer, the product is the layer that replaces them: a governed adapter reading the EHR, lab, pharmacy, bed board, staffing, RTLS, and telemetry through open standards (FHIR R4, HL7 v2, DICOM) across ~two dozen integration categories. An Epic-class install runs $10–300M+ over years; a coordination layer deploys in weeks, improves continuously, and the only hardware a hospital buys is the smartboards.
Above the fabric sits Charlie. Charlie Assist surfaces the next task by clinical urgency behind a k≥5 privacy floor — the pilot modeled ~2,457 nursing hours returned a year. Charlie Companion reaches the one lever hospital software never touches: patient demand. A platform clinicians refuse to give up and patients seek out pulls from both sides at once — almost unheard of in enterprise hospital software.
| Category | Typical annual cost | Structural limitation |
|---|---|---|
| EHR / system of record | $10M–$300M+ install | A billing & legal record, not an orchestrator. Can't be meaningfully updated without buying a new product. |
| Capacity / command center | $150K–$500K+ / facility | Watches one domain (beds, OR, transfers). Another dashboard, another silo. |
| Ambient documentation | $100–$300 / provider / mo | Transcribes the visit. Cloud-only; does not touch the operation. |
| Staffing / scheduling | $50K–$200K / facility | Optimizes labor cost. No flow model, no governance. |
| Quality / safety dashboards | $30K–$150K / facility | Reports harm after the fact. No closed loop. |
| System of action | — unoccupied — | The vendor-agnostic, real-time, governed layer over everything. The gap and the moat. |
Unit economics — anchored by the digital pilot (modeled, not observed)
A modeled 274-bed hospital projected roughly $2.0M in net annual value against a $59,000 platform cost — about $7,500 of modeled value per bed, a 34.9× return. Read against the status quo, the case is concrete: the average hospital loses $4.2–6.2M a year to nurse turnover, one hour of ED-boarding reduction can recapture five figures per day, and Medicare's penalties convert preventable harm directly into lost payment.
The closure crisis & the social-impact market
46% of rural hospitals run negative margins and 432 are vulnerable to closure; 293 dropped obstetrics between 2011–2023. A money-losing facility can't cut its way to survival — it needs to recover margin without adding headcount, which is what coordinating the operation does. These buyers are also structurally excluded from cloud-only vendors. Chartis, 2025
Why now
Three forces converged in 2026: regulation turned transparency into a procurement requirement (Joint Commission's Responsible-Use-of-AI cert launched June 2026; ONC transparency rules); clinician trust in black-box AI collapsed, rotating the market toward governable systems; and the staffing crisis is now understood as structural, so buyers want tools that retain staff by removing burden.
VI · Risks & honest disclosures
As of June 2026 the ADT interface is built and tested, but no live hospital feed has flowed. The first design partner is the make-or-break validation.
Pilot value figures are projections on synthetic data and literature-based rates, not observed outcomes; clinical composites are labeled unvalidated pending advisor review.
"We already have Epic" is a real objection. The answer: Epic is a record, not an orchestrator — but a system-of-record vendor's distribution advantage is real.
Hospital enterprise software averages ~12-month, committee-driven cycles; adoption is slow even when ROI is clear.
Healthcare outcomes have many causes; any value-based contract must define metrics, baselines, and measurement windows precisely.
A non-traditional solo founder at pre-seed; the domain-specific work is done, but the team and capital to scale must still be built.
VII · Conclusion
Apply the operational and logistical control the military perfected, automate the analog work, manage the system for flow, and return the human to care — and the largest, most addressable inefficiency in the U.S. economy becomes a market.
The thesis in motion
The same argument as a short film — the data, why two decades of technology failed, and Nightingale OS as the system of action.
Selected references
Forward-looking and pilot figures are modeled projections, not observed outcomes. A selection of the sources cited inline is below — the complete 45-source citation list is in the full document:
Prepared by Regulation Loop, Inc. (Nightingale OS), June 2026. The complete 45-source citation list and all figures appear in the full document — download the PDF ↓. This analysis does not constitute an offer to sell securities.