Appendix K: The evidence behind the transition design

This appendix supports the section "The transition: what changes, and when" in the plan. Each of the plan's transition rules is written against a specific failure. This appendix describes the failures.

The demand shock: what the Oregon experiment showed

Making care free in year one creates demand in year one, and the buildings and the workforce do not arrive until years six to 10. The plan has to budget for the gap, and it can, because the size of the surge is known from the only randomized experiment ever run on it.

In 2008 Oregon had money to add about 10,000 low-income adults to its Medicaid program and roughly 90,000 people who wanted in, so it held a lottery. The lottery produced something health policy almost never gets: a randomly selected group of people who gained coverage and an otherwise identical group who did not. Researchers followed both for years.

The results are the best evidence on what free coverage does. People who gained coverage used more care: office visits, prescriptions, and hospital admissions all rose, and total use was up by roughly a quarter to a third. Emergency department visits rose by about 40 percent, contrary to the hope that coverage would move people out of the emergency room and into primary care. Medical debt and catastrophic out-of-pocket spending fell sharply, and rates of depression fell. But after two years there was no measurable improvement in blood pressure, cholesterol, or blood sugar control.

The plan reads that two ways. The cost of induced use is real and is budgeted at $100 to $200 billion a year (Appendix J). And coverage alone does not produce health, which is the whole argument for building capacity rather than only paying for care. Money is not the constraint in the early years. Clinicians and exam rooms are.

The national appointment search, which goes live in the first year, will make the queue visible. That is the point of building it and also the risk: it will be the first real-time national measurement of how long Americans wait, and in the early years it will show waits lengthening, accurately, on a system this plan built. The demand arrives before the buildings, and the plan says so.

Deleting the appointments that exist only as friction

The best response to a demand surge in a capacity-constrained system is not to suppress demand but to delete the appointments that were never about care. Every one of these is already in the plan as a consumer convenience, and each is also capacity policy.

• The prescription that renews itself. A large share of primary care visits exist solely to renew a prescription for a condition that has not changed. Drop the renewal requirement for chronic maintenance drugs and the visit disappears. Nobody is denied anything.
• The standing referral. A permanent condition is authorized permanently, so the annual reauthorization visit disappears too.
• Asynchronous primary care. E-visits and secure messaging for the large share of primary care that never needed an appointment slot.
• The automatic search past the guarantee. When a service cannot be delivered locally inside the waiting-time deadline, the system finds a slot elsewhere, books it, and pays the travel. It creates no capacity. It uses the geographic slack that already exists, which is the difference between 11 weeks at the nearest hospital and nine days an hour away.

Rule 1: no eligibility determination. The Medicaid unwinding.

During the pandemic, Congress barred states from removing anyone from Medicaid, and enrollment grew to more than 90 million. When the protection ended in April 2023, states began redetermining everyone's eligibility, a process that came to be called the unwinding. Over the following 18 months roughly 25 million people were removed from Medicaid, and the large majority of them were removed not because they had been found ineligible but for procedural reasons: a renewal form sent to an old address, a deadline missed, a document not received. Many were children. Many were eligible and had to reapply, and some states' systems were so overwhelmed that they were ordered to pause.

The lesson the plan draws is not that redeterminations should be done better. It is that a universal system should not have them. Everyone physically present in the United States is covered. There is no form, no verification, no redetermination, and no renewal. The provider creates coverage by treating you. This is also the largest single administrative saving in the plan, because eligibility determination is the most expensive thing Medicaid does that is not care.

Rule 2: pay first, recover later. The run-out problem.

When a cohort moves to the public payer, there is a window in which claims for care already delivered are still arriving at the old plan. The plan's rule is that the public payer pays every claim with a service date within 180 days either side of the cutover, regardless of what plan the patient held, and recovers from the prior plan afterward. The patient is never caught between two systems.

The reason this rule matters most for the people with the worst coverage is a feature of employer insurance most people never see. About two thirds of workers with employer coverage are in self-funded plans, in which the employer itself bears the financial risk for its employees' claims and hires an insurer only to process the paperwork, under what is called an administrative-services-only arrangement. The insurer's name is on the card, but the money is the employer's. When such an employer's plan ceases to exist at cutover, nobody is left to fund the run-out of claims already incurred, and an employee's only remedy against a marginal or failing employer is a lawsuit under the federal benefits law, ERISA, in the middle of the transition. Pay first, recover later, is the only design that does not turn a cutover into a mass-litigation event.

Rule 3: mid-course of treatment is untouchable

No patient in an active course of treatment changes payer or authorization at cutover. Every prior authorization in force on cutover day is honored by the public payer for 12 months without review. Every course of chemotherapy, every transplant workup, and every pregnancy runs to completion under the arrangements it started under. This costs almost nothing, because the care was going to be paid for anyway, and it removes the single most frightening scenario a transition can produce.

Rule 4: fill every prescription. The Part D launch.

Medicare's drug benefit went live on 1 January 2006, and on that day roughly six million people eligible for both Medicare and Medicaid were moved from state drug coverage to new private plans whose enrollment data had not fully arrived. Pharmacies could not confirm coverage, or found patients in the wrong plan, or were told to charge copays of hundreds of dollars to people living on a few hundred a month. Patients left without medication, pharmacists filled prescriptions on trust and ate the cost, and more than 20 governors declared emergencies to pay for drugs with state money. Appendix F tells the story in full.

The plan's rule is the 180-day transition fill: any prescription presented in the first 180 days after cutover is dispensed as a 30-day supply and paid, with clinical checks intact and only the coverage question deferred. Part D's launch failed for six million people. This transition is for 330 million. The cost is small relative to the alternative, and the rule stays permanently available as the fallback whenever the system is down.

Rule 5: never replace the system and cut the people in the same year. Phoenix.

In 2016 the Canadian federal government replaced its payroll system for some 300,000 public servants with a new system called Phoenix. In the same window it consolidated payroll operations into a single center and laid off most of the experienced payroll staff across departments, on the theory that the new software made them unnecessary. The software did not work. For years afterward tens of thousands of civil servants were underpaid, overpaid, or not paid at all, the backlog of cases reached the hundreds of thousands, and the cost of fixing it ran to billions of dollars, many times what the system was supposed to save. The auditor general's conclusion was that the failure was not the software alone but the decision to remove the people who understood the old system before the new one had proven itself.

The plan's rule: you may replace the system, or cut the people who ran it, but never in the same fiscal year. Nothing is switched off until its replacement has run in parallel for a year. The pharmacy switching infrastructure, the Medicare Administrative Contractors that process traditional Medicare's claims, and the state eligibility systems are all contracted forward as regulated utilities rather than terminated. This is also why the insurance-industry headcount reduction belongs in years six to 10 and not year one: the people who process claims are the people who know how the claims work.

The financing timing, and why out-of-pocket share misleads

Delivering relief first costs money before it saves any. It moves roughly $250 to $400 billion a year into years one through three, when the buildout, the training, the transition guarantee, and the new benefits are all running and no savings line has begun to ramp. Two things make the money work in the years the plan has to survive.

The revenue clock runs with the relief clock. The employer contribution and the household contribution begin in year one, when relief begins, not at enrollment in year three to six. The public purse starts buying down everyone's cost-sharing in year one, so it may tax for it in year one.

Insurers cut premiums by what the government takes off their books. From enactment, the government pays for chronic-disease drugs, primary care, glasses and hearing aids, and caps out-of-pocket costs. Those are liabilities that private insurers no longer carry, and the law requires them to cut premiums by that value, and to price no higher than the prior year. That offsets a large share of the cost in the same year. It arrives as lower premiums rather than as revenue, so it does nothing for the federal deficit. What it cuts is the amount each household pays in a year, which is the number that decides whether a family is better off.

On that point, one international statistic needs explaining because it will be used against the plan. The standard measure of "out-of-pocket share" counts only what patients pay at the point of service, copays and the like, as a share of total health spending. By that measure America's out-of-pocket share is low, lower than Denmark's. The measure excludes premiums and deductibles from its numerator, and in America those are enormous. American out-of-pocket cost is distinguished not by its average but by its variance: it is concentrated on the sick, and it is the only rich country where it routinely produces bankruptcy. A family with a $6,000 deductible and a $1,400 monthly premium is, by the international statistic, barely paying out of pocket at all.

The long-term care sequence: Japan

Long-term care is the benefit the plan most wants to deliver and the one it cannot deliver fast, because the workforce that assesses need and manages care does not exist. Japan is the model. When it built its long-term care insurance system, it spent the years from 1997 to 2000 training a corps of care managers and building a national assessment instrument, and only then, in April 2000, switched the benefit on. The plan copies that exactly: an assessor corps in years one to three, staffed largely with the utilization-review nurses displaced from the insurance industry, who are already trained in structured assessment; the benefit live in year four to five. Meanwhile the existing waiting lists for home and community services are cleared, because those people are already assessed, and the caregiver allowance is paid from the day of application whether or not a place has come up, because it has no workforce constraint at all.