Twice The Money, Years Less Life: What Is Actually Broken In American Healthcare

A physician’s diagnosis of three health systems, and why venture capital keeps funding the wrong machine.

You have read a version of this article before. It usually ends with one side shouting “single payer” and the other shouting “socialised medicine,” and nobody learns anything. I understand the fatigue, because I shared it for years. I spent two decades running clinical protocols for Novartis, Roche, AstraZeneca and Otsuka. I watched excellent American clinicians deliver excellent care inside a machine that then sent the patient a bill nobody in the room could explain. For a long time I assumed the explanation was greed. It mostly is not. It is architecture. If you invest in health technology, that single distinction decides whether your portfolio is buying health or buying friction.

Here is the number that should stop you. In 2024 the United States spent $14,775 per person on health consumption. The average across comparable wealthy countries was $7,860. American health spending was 17.2% of GDP against 11.2% for its peers. That same year, US life expectancy reached an all-time high of 79 years and still trailed those peers by several years.

Twice the price. Worse outcome. A clinical trial with that risk-benefit profile would be stopped at interim analysis.

First principles: A health system is three machines

Machine one decides who is allowed in. That is financing and coverage. Machine two delivers the care. That is clinicians, nurses, hospitals. Machine three decides what a thing costs and who pays for it. Almost every argument about American healthcare confuses machine two with machine three.

The evidence separates them cleanly. The Commonwealth Fund compared ten wealthy countries across seventy performance measures. The United States ranked last overall. It ranked tenth of ten on access, tenth on equity, tenth on health outcomes, and ninth on administrative efficiency. But on care process, the domain covering prevention, safety, coordination and patient engagement, the United States ranked second, behind only New Zealand.

Read that again, because it answers the question most people get wrong. The doctors are not the defect. What clinicians actually do at the bedside is world-class. The payment machine is what fails, and it fails hard enough to drag the outcome numbers down with it.

What the payment machine actually does

It runs an argument. In the American system, every clinical decision is a transaction that a second party may refuse, and that right of refusal has to be staffed on both sides of the table.

The American Medical Association surveyed 1,000 practising physicians in December 2025. Each physician completes about 40 prior authorisation requests a week. Physicians and their staff spend an average of 13 hours a week processing them. Ninety-four per cent say the process contributes to burnout. Twenty-six per cent report that it led to a serious adverse event for a patient in their care, including hospitalisation, permanent impairment or death.

Now multiply. Thirteen hours per physician per week, plus the mirror-image workforce inside every insurer whose job is to evaluate and often decline. You have built a second health system that produces no health. Estimates of the administrative share of US spending run from roughly a quarter to nearly a third depending on what you count. The classic cross-national study put it at 31% of total expenditure, or $1,059 per person against $307 in Canada (a 2003 figure; no update of comparable granularity has been published, which is itself telling).

Richard Feynman’s test was that if you cannot explain a mechanism in plain words, you do not understand it. So, plainly: America does not overpay because its doctors are lavish. It overpays because it employs an enormous number of people to argue about the bill, and because prices are set privately rather than negotiated nationally. That is a design choice, not a moral failing. Design choices can be changed.

India: the cash constraint

India runs the opposite failure. The most recent National Health Accounts, released in May 2026, put government health expenditure at 1.43% of GDP in 2022–23 (1.48% on the revised GDP series) and out-of-pocket expenditure at 43.4% of total health spending, down from 64.2% a decade earlier. That decline is real and it is not linear: out-of-pocket spending had fallen to 39.4% during pandemic-era public spending, then rose again when that spending stopped.

What that means at the bedside is severe. Out-of-pocket payments are estimated to push around 55 million Indians into poverty every year, and roughly one household in six incurs catastrophic health expenditure. The counter-intuitive part, and the part most health-tech founders miss, is where the damage happens. One national analysis found catastrophic spending was more common for outpatient care than for hospitalisation, at 47.8% of affected households against 43.1%. Medicines alone account for approximately two-thirds of out-of-pocket health spending.

India’s strength is unit cost and throughput. Its weakness is that the flagship insurance scheme covers hospitalisation, while the money that actually ruins families flows through clinics, chemists and diagnostic labs.

Europe: the capacity constraint

Europe largely solved financing and then ran into physics. You cannot manufacture a consultant surgeon overnight. In England, the referral-to-treatment waiting list stood at 7.28 million pathways in May 2026, around 6.16 million individual patients, with roughly 105,000 waiting more than a year. Only 65.6% were treated within 18 weeks against a 92% standard. Nobody goes bankrupt. Some people wait a year for a hip.

Europe’s underrated asset for anyone building health technology is that a payer will actually buy evidence. Germany’s DiGA pathway lets a physician prescribe an approved app, reimbursed by statutory insurance covering roughly 73 million people. As of mid-2025 the directory listed 44 permanently and 14 provisionally approved products, with over a million prescriptions issued by the end of 2024 and about €234 million in cumulative statutory spending. Eleven products had been delisted, six of them because they could not demonstrate a positive healthcare effect.

Be honest about the other half of that story. A 2025 systematic review of all 23 published DiGA approval studies found every one of them carried an overall high risk of bias, mostly from missing outcome data and unblinded patient-reported measures. Europe has built the best reimbursement door in the world and has not yet built an equally strong evidence lock behind it.

The counter-case, taken seriously

Three objections deserve real weight. First, the comparison is not fair to the clinic. American life expectancy is dragged down by obesity, firearm deaths, road deaths and overdose, none of which a hospital causes and few of which it can fix. Second, the excess spend buys optionality: faster specialist access and a disproportionate share of global drug and device innovation that other systems partly free-ride on. Third, universal systems ration too, just in a different currency, and India’s low spending is not thrift but an unpriced transfer of risk onto households.

So the tidy sentence “America is broken and Europe is fine” is wrong. Every system rations. The United States rations by price, Britain rations by time, India rations by wealth. Choosing which failure you can live with is the actual policy question, and pretending otherwise is how bad investment theses get written.

Where the money actually goes

Now the venture question. US digital health startups raised $14.2 billion in 2025, up 35% on 2024 and the highest since 2022. AI-enabled companies took 54% of those dollars, up from 37% the year before. Deal count fell to 482 from 509, and raises over $100 million made up 42% of all funding. Strip out the top nine companies by dollars raised and the year’s total falls below 2024.

Follow the value propositions rather than the headline. In the first half of 2025 the three best-funded categories were non-clinical workflow ($1.9B), clinical workflow ($1.9B) and data infrastructure ($893M), together 55% of all digital health funding, a first in Rock Health’s dataset.

Read that plainly. Venture capital is not principally funding cures. It is funding the argument machine: coding, documentation, prior authorisation, denials, revenue cycle. That is a rational trade. Administrative waste is the largest, most liquid and most obviously wasteful budget line in the richest health market on earth, and the buyer has cash and a CFO who can sign. But name it correctly. It is a margin trade, not a health trade. It makes the existing architecture cheaper to operate rather than replacing it.

There is an uncomfortable corollary. The more efficiently AI resolves prior authorisation, the more prior authorisation the system can afford to run. And the operational reality is unforgiving: only about a third of health AI pilots scale to full system-wide deployment.

What I would underwrite

• Name the payer and the budget line before the demo. Which code, which capitation pool, whose medical-loss ratio. “Hospitals will save money” is not a payer.
• In the United States, prefer companies whose value survives payment reform. If prior authorisation vanished tomorrow, what is left in the box?
• In Europe, underwrite evidence capability, not app polish. A DiGA-style pathway delists products that cannot prove effect. That is a clinical-trial competency inside a software company’s cost structure, and most founders have not budgeted for it.
• In India, underwrite against household cash, not insurance. The real market is outpatient care, medicines and diagnostics, priced in rupees per episode.
• Treat generalisability as a diligence question, not a technical footnote. A model validated at one health system is a hypothesis everywhere else.

What would change my mind

If a portfolio of administrative-AI companies could show, in audited data across more than one health system, that total cost of care fell rather than merely the cost of billing, I would drop the “margin trade, not health trade” framing. I would also revise if the current US prior authorisation reform commitments, with deadlines running through 2027, actually land. In that world the administrative market shrinks and the thesis inverts: today’s workflow winners become tomorrow’s stranded assets. I am watching that deadline more closely than any funding round.

The question I would rather have argued than agreed with

If you invest: name one healthcare AI company in your portfolio whose value does not depend on administrative complexity persisting. If you are building: tell me which of the three failures you are actually selling against, price, time or wealth, because the same product cannot fix all three. If you practise medicine in any of these systems: tell me where this diagnosis is wrong. I would rather correct it now than underwrite it later.

HealthcareEconomics #DigitalHealth #VentureCapital #HealthPolicy #HealthTech #AIinHealthcare #GlobalHealth #HealthcareInnovation

SOURCES (APA 7TH EDITION)

Ahmad, F., & Mohanty, P. C. (2024). Incidence and intensity of catastrophic health expenditure and impoverishment among the elderly: An empirical evidence from India. Scientific Reports, 14, 15908. https://doi.org/10.1038/s41598-024-55142-1

American Medical Association. (2026, May 13). AMA survey: Prior authorization reform pledge falls short with physicians. https://www.ama-assn.org/press-center/ama-press-releases/ama-survey-prior-authorization-reform-pledge-falls-short-physicians

Blumenthal, D., Gumas, E. D., Shah, A., Gunja, M. Z., & Williams, R. D., II. (2024). Mirror, mirror 2024: A portrait of the failing U.S. health system. The Commonwealth Fund. https://doi.org/10.26099/ta0g-zp66

British Medical Association. (2026). NHS backlog data analysis. https://www.bma.org.uk/advice-and-support/nhs-delivery-and-workforce/pressures/nhs-backlog-data-analysis

Himmelstein, D. U., Campbell, T., & Woolhandler, S. (2003). Costs of health care administration in the United States and Canada. New England Journal of Medicine, 349(8), 768–775. https://doi.org/10.1056/NEJMsa022033

KFF. (2026). Americans’ challenges with health care costs. https://www.kff.org/health-costs/americans-challenges-with-health-care-costs/

Ministry of Health and Family Welfare, Government of India. (2026, May 27). Union Health Ministry releases the National Health Accounts estimates for India 2022–23 [Press release]. Press Information Bureau. https://www.pib.gov.in/PressReleasePage.aspx?PRID=2265816

Rock Health. (2026, January 12). 2025 year-end digital health funding overview: A tale of two markets. https://rockhealth.com/insights/2025-year-end-digital-health-funding-overview-a-tale-of-two-markets/

Rock Health. (2025, July 7). H1 2025 market overview: Proof in the pudding. https://rockhealth.com/insights/h1-2025-market-overview-proof-in-the-pudding/

Sippli, K., Deckert, S., Schmitt, J., & Scheibe, M. (2025). Healthcare effects and evidence robustness of reimbursable digital health applications in Germany: A systematic review. npj Digital Medicine, 8, 495. https://doi.org/10.1038/s41746-025-01879-6

Sriram, S. (2026). Can a bigger budget shrink the bill at the bedside? Public health financing as a route to lower out-of-pocket spending in India. Frontiers in Public Health, 14, 1903360. https://doi.org/10.3389/fpubh.2026.1903360

Telesford, I., Cotter, L., Wager, E., & Cox, C. (2026, March 11). How does health spending in the U.S. compare to other countries? Peterson-KFF Health System Tracker. https://www.healthsystemtracker.org/chart-collection/health-spending-u-s-compare-countries/

GLOSSARY OF TERMS

Billing and insurance-related (BIR) cost — The staff time and systems cost of generating, submitting, contesting and settling a medical bill, on both the provider and the payer side.

Care process — A Commonwealth Fund performance domain covering prevention, patient safety, care coordination, patient engagement and responsiveness to patient preferences. It measures what clinicians deliver, separately from what the system charges.

Catastrophic health expenditure (CHE) — Household health spending that exceeds a defined share of total household consumption, commonly 10% or 25%. It measures financial shock rather than poverty status.

DiGA (Digitale Gesundheitsanwendungen) — Germany’s category of prescribable, statutory-insurance-reimbursed digital health applications, assessed by BfArM under a fast-track pathway that requires evidence of a positive healthcare effect.

Health consumption expenditure — Spending on health goods and services, excluding capital investment in buildings, equipment and research. The basis for OECD cross-country comparison.

Impoverishment (health-related) — The count of households pushed below a national poverty line specifically by medical payments. Distinct from catastrophic expenditure.

Mega deal — A single venture financing round of $100 million or more.

Out-of-pocket expenditure (OOPE) — Direct payments made by households at the point of care, net of any reimbursement. Expressed here as a share of total health expenditure.

Prior authorisation — A payer requirement that a clinician obtain approval before a service, drug or procedure will be covered. The principal source of administrative friction quantified in this article.

Referral to treatment (RTT) — The NHS England measure of the interval between a GP referral and the start of consultant-led treatment. The constitutional standard is 92% within 18 weeks.

Risk of bias (RoB 2) — The revised Cochrane instrument for judging how far a randomised trial’s design, conduct or reporting could distort its result.

Value proposition (venture usage) — Rock Health’s classification of what a digital health company sells, for example clinical workflow, non-clinical workflow or data infrastructure.

Prepared by Kaveri Rangappa, MBBS, MPH. All figures verified against primary sources on 13 August 2026. Figures older than 24 months are flagged in-line in the article text.