Epistemic status: I have operated on the receiving end of this machine for twenty years, which is both my authority and my bias, and you should weigh it as both. The history and mechanics are confident and sourced; every number traces to a primary source in the references. Section 07 is the strongest case against this piece, written as well as I can write it.
01The Bed
On a Tuesday, a doctor orders a hospital bed for a patient going home.
That sentence should end the story. A physician examined a patient, made a clinical decision, and wrote an order. In most industries, that is the moment the product ships.
In healthcare, it is the moment a second process begins. One the patient cannot see and mostly does not know exists. Before that bed rolls off a truck, someone who has never met the patient has to say yes.
I run home health and hospice companies in Southern California. My teams live downstream of that yes. Hold on to the bed. We are coming back for it.
02What It Actually Is
Prior authorization is a permission system for payment. It is not a medical event. The medical decision already happened, in the exam room, on Tuesday. Prior auth is the insurance company requiring approval, in advance, before it agrees to pay for what the doctor already ordered. No approval, no payment. And for most families, no payment means no bed.
I chose a hospital bed on purpose. Durable medical equipment is the single most prior-authorized category in Medicare Advantage, with more than seven in ten enrollees in plans that require permission for it. The bed is not an edge case. It is the machine’s most common customer.
Here is the request from the inside. The mechanics vary by market, Medicare Advantage, Medicaid, commercial, so I am going to show you the machine as it faces a Medicare-age patient, because that is my ground and the best-documented version of it.
Someone at the doctor’s office or the equipment company, usually an intake coordinator whose entire job is this, assembles a packet. The order, the chart notes, the diagnosis codes, the documentation proving the patient meets this plan’s definition of medically necessary. It goes to the insurer through a web portal or, still, in 2026, by phone and fax. That is not a flourish. Industry’s own measurement found only about a third of medical prior authorizations moved fully electronically in 2023, and even the newest count has it at 40 percent. Most of these requests still travel the way paperwork traveled in 1995. It lands in a queue.
On the other end a reviewer opens it. Often a nurse. Increasingly, software screens it first, and here is a mechanical truth worth holding: under the current rules, the machine is allowed to say yes by itself. Saying no is supposed to require a human. The packet gets compared against criteria, and since 2024, Medicare Advantage plans must follow traditional Medicare’s own coverage rules where they exist, and cannot use proprietary criteria to override them. If the packet matches, approved. If a page is missing, the request gets pended, and a clock starts while the office hunts down one more note.
The official clock, for the record: 72 hours for an urgent request, seven calendar days for a standard one. That seven was fourteen until this January, when new federal rules cut it in half. Which means that for decades, the legal answer to a doctor’s Tuesday order could arrive a full two weeks later and be on time.
Meanwhile the patient is home. The bed is not.
If the answer is heading toward no, the rules require that someone with expertise in the relevant field of medicine review it first, a physician or other qualified clinician. Not necessarily a doctor, and that guardrail was only codified for Medicare Advantage in 2024. In many plans the ordering doctor can then request a peer-to-peer, a phone call where one clinician who knows the patient tries to convince one who does not. If that fails, a letter goes out explaining appeal rights. Most people never use them. Hold that thought too, because it is the strangest fact in this whole story.
That is the machine. The better question is why it exists at all, because nobody designed this on a whiteboard. It accreted. And the accretion has a start date.
03Why It Started
In 1965, Medicare and Medicaid launched, and public money began flowing into a fee-for-service system with no brakes. Every service performed was a service paid. Spending did what spending does under those rules, and within a few years Washington had found real fraud and real overuse in the bills it was paying.
Congress answered in 1972. The Social Security Amendments created Professional Standards Review Organizations, boards of physicians assigned to check whether the hospital care Medicare was buying had actually been necessary.
Look at the design, because it matters. Review after the fact. A doctor could still act first and answer questions later. The checkpoint was an audit, not a gate. Care was never held hostage to permission.
04The Move
The gate came in the 1980s, and it came from panic over premiums.
Employers were watching health costs eat their payroll, and a new industry of utilization review firms sold them a simple upgrade. Instead of auditing care after it happens, require permission before it happens. Certify in advance. Ask first. Researchers who study this period are blunt about how fast it happened: for practical purposes, this kind of medical management barely existed before 1984, and within a decade it was an industry of massive proportions.
It is the same review, moved one square earlier on the board. That one move changed everything, because now the delay lands on the patient instead of on the paperwork.
05The Migration
By the 1990s managed care had scaled the machinery to everything, and the public revolted. The HMO backlash was a genuine consumer uprising, and insurers loosened the most visible gatekeeping, the referral slips, the permission to see a specialist.
But prior auth did not die. It migrated. Into the pharmacy, through PBM formularies. Into imaging in the mid-2000s, through radiology benefit managers built for exactly that job. And eventually onto ordinary, routine services. Hospital beds. Walkers. Home health visits. By 2024, Medicare Advantage plans collectively required prior authorization for more than 4,000 distinct services, and virtually every enrollee was in a plan that required it for something.
Why did it spread to places the original logic never covered? Because the checkpoint got cheap. Reviewing a request used to take a physician’s time. Then a nurse with criteria software. Now, increasingly, an algorithm and milliseconds. When the cost of asking first falls toward zero, ask first gets applied to everything, whether or not anyone re-runs the math on whether it should.
I spent a year learning to run the five-step algorithm on my own company. Question every requirement, and every requirement has to come with a name attached. Here is a requirement that now sits in front of tens of millions of medical decisions a year, and in fifty years of accretion, no one with the power to do it has stood back and asked the first question. Who asked for this one, here, on this service, and is it still doing what it was built to do?
What the Machine Produces
Now the output, counted the way a manufacturer would count it.
Sit with that combination. A checkpoint that waves nearly everyone through. Whose refusals usually do not survive a second look. And whose refusals mostly go unchallenged anyway, because the family on the other end is exhausted, or confused, or out of time.
Whatever that system is producing, its primary output is not no. Its primary output is later.
Later, multiplied by 52.8 million requests. Plus an administrative economy running on both sides of the wall: practices completing 39 prior authorizations per physician per week, physicians and their staff burning around 13 hours a week on them, four in ten practices employing people who do nothing else. The federal government’s own estimate, when it wrote new rules to speed this process up, was that removing the friction would save the system about 15 billion dollars over ten years. That is a government agency putting a price tag on the delay itself.
Which raises the question a regular person should be asking by now. Why would an insurance company run this machine at all? Follow the money, because it is not hidden. Medicare Advantage plans are paid a fixed amount per member, in advance, whether the member uses a little care or a lot. Under that arrangement, every service the machine delays, shrinks, or quietly talks a family out of pursuing is money the plan keeps. The federal watchdog that audits these plans named that exact incentive, in writing, as the central concern with the payment model before it ever opened a single denial file. The machine does not need to say no to earn its keep. Later is profitable all by itself.
07The Strongest Case Against Me
This piece does not deserve to exist unless it can survive the best argument for the machine, so here it is at full strength.
Low-value care is real. Researchers have documented enormous spending on services that do not help people and some that harm them, and a few service lines carry genuine fraud. The defenders of prior authorization say the 90 percent approval rate proves nothing, because the checkpoint works through deterrence. The truly bad requests were never submitted at all. You are judging a dam by the water that gets through.
And in my own industry, the defenders hold a real receipt. In 2016 Medicare tried a permission requirement on home health in Illinois, and researchers later measured what happened: spending dropped 13 percent almost immediately, even though about 97 percent of requests were ultimately affirmed. Nearly everything approved, spending down anyway. That is the deterrence effect, demonstrated, in my back yard. I am not going to pretend it away.
Here is why I think the argument still loses. The Illinois result is an argument about targeting, not a license for everything the machine now does. Medicare aimed that requirement at a specific program, in a specific place, with documented improper payments, and even its defenders describe it as a fraud tool. Nothing in that logic explains a permission requirement on a hospital bed for a dying patient. And when a checkpoint’s refusals are overturned four times out of five whenever anyone pushes back, the deterrence story has to explain why the machine is so wrong precisely when it says no.
You do not have to take my word that the targeting failed. Take the industry’s. Roughly fifty insurers covering 257 million Americans pledged last June to shrink prior authorization, and by this spring reported eliminating 11 percent of their requirements, six and a half million fewer requests. UnitedHealthcare, the largest of them, went further: in April 2025 it eliminated prior authorization for home health services in its Medicare Advantage plans across more than thirty states. You do not delete a third of a safeguard. You delete a third of a tax.
08My Industry’s Version
Here is what the machine does at my station on the line.
Ninety percent of Medicare Advantage enrollees are in plans that require prior authorization for home health care. Medicare’s own advisory commission compared those patients to patients in traditional Medicare, where no such requirement exists, and found what you would now predict. MA patients who use home health receive about 11 percent fewer visits. Agencies report referral denials several times higher from some MA plans. Hospitals hold patients for days waiting on the yes. A federal watchdog found that 13 percent of the denials it examined were for care that met Medicare’s own coverage rules. And a study published this year of more than 400,000 people in their last year of life found the requirement does not land evenly: prior authorization widened the gap in home health use for Hispanic and Asian patients at the end of life. The checkpoint is not just slow. It is slower for some families than others.
There is one part of my world the machine never conquered. Hospice. When a person elects the Medicare hospice benefit, the permission structure falls away, and care for the terminal illness proceeds without an insurer’s advance yes. The industry tried an experiment that would have brought hospice under managed care’s roof, and it was shut down at the end of 2024. I have watched what the absence of the machine looks like, at the hardest moment families ever face. It looks like care that starts when the doctor says so.
09The Commission on No
Which brings the history to this year, and to my industry’s door.
In January, Medicare itself, the traditional program that processed only about 625,000 prior authorization requests in all of 2024, switched on an experiment called WISeR in six states. Artificial intelligence now reviews permission requests for a list of services that includes wound care products. The reviewing is done by private technology vendors, and here is the design detail that tells you everything: the vendors are paid a share of the savings generated when care is not delivered, reported at ten to twenty percent. A commission on no.
The early returns leaked out of Texas. The AI approved 62 percent of requests on first pass. Human review pulled it up to 84. The Medicare Advantage norm is 92. A digital rights group is suing the government just to see how the algorithm works. A House committee voted unanimously to block it. It is still running.
Step back and look at what the renovation actually is. Faster clocks, cleaner pipes, real-time answers by 2027, fewer requirements on paper, and a brand new AI checkpoint growing in the program that had mostly escaped one. Everyone in the room is asking how to run the machine faster. The 1972 question, whether this checkpoint still earns its place on this service, sits unasked. A faster version of the wrong checkpoint is not reform. It is a shorter line at the wrong gate.
10Take This With You
If you take one practical thing from these two thousand words, take this. The no is usually not final. Four out of five denials that get appealed are overturned, and the machine is counting on you not knowing that. The denial letter in your hand lists the appeal steps because the law makes it. Your doctor’s office fights these every day and will help. The families who push back usually win. The machine’s greatest ally is the moment a tired person decides it is not worth the fight, and that moment is a choice.
11The Ledger
Every field note that makes a prediction gets a dated line on the public ledger. Here’s this one.
Ways I’m Wrong
The deterrence effect is the honest wound in this piece. The Illinois evidence shows a permission requirement changes behavior upstream even when it approves almost everything, which means the machine’s defenders can always claim its value is invisible, and I cannot fully falsify that. I have argued targeting, not existence, and a smarter defender than me might close that gap. My data leans on Medicare Advantage because that is what gets published; commercial insurance may look better or worse, and I do not know which. My seat means I feel the machine’s costs every day and its benefits never, which is exactly the vantage most likely to underweight them. And the insurers’ 11 percent reduction figure is their own self-reported scorecard, unaudited; I have used it as an admission, not as proof of relief, but you should know its source.
The bed, by the way, eventually comes. It usually does. A truck arrives some number of days after Tuesday, a driver carries the frame up the walk, and nobody in that living room ever finds out why it took so long, or that a machine older than most of them had to say yes first. Now you know.
No comment box here on purpose. The discussion for every field note lives on LinkedIn, in public, with names attached.
The next one ships when it's true. Leave an email and it finds you.
Unsubscribe anytimeReferences · 21 sources
1. MA prior authorization volume, approvals, denials, appeals, overturns: 52.8 million determinations, 7.7% denied, 11.5% of denials appealed, 80.7% of appeals overturned, 2024. KFF, January 2026. kff.org
2. Professional Standards Review Organizations. Social Security Amendments of 1972, P.L. 92-603.
3. Growth of private utilization review after 1984. Health services research literature on utilization management, including the GAO characterization of the industry’s scale.
4. More than 4,000 services requiring prior authorization in Medicare Advantage, 2024. KFF.
5. Durable medical equipment as the most prior-authorized MA category, 70%+ of enrollees. KFF data notes; Am J Manag Care 2024;30(3).
6. Electronic versus manual prior authorization share: 31% fully electronic in 2023, 40% in the latest index. CAQH Index. caqh.org
7. Expert-reviewer requirement for MA medical-necessity denials; MA plans bound to traditional Medicare coverage criteria. CMS Contract Year 2024 MA Final Rule, CMS-4201-F.
8. Decision clocks, 72 hours expedited and 7 calendar days standard, effective January 2026, previously 14 days. CMS Interoperability and Prior Authorization Final Rule, CMS-0057-F.
9. Algorithms may not override individualized medical-necessity determinations in MA. CMS FAQ guidance, February 2024; insurer pledge on licensed-clinician review, June 2025.
10. Physician burden: 39 prior authorizations per physician per week, about 13 hours weekly, 40% with dedicated staff. AMA Prior Authorization Physician Survey, 2024. Self-reported survey. ama-assn.org
11. $15 billion ten-year savings estimate from removing prior authorization friction. CMS-0057-F regulatory impact analysis.
12. Capitated payment incentive named as the central concern with the MA model. HHS Office of Inspector General, April 2022, OEI-09-18-00260. oig.hhs.gov
13. 13% of examined MA prior authorization denials met Medicare coverage rules. Same OIG report.
14. Illinois pre-claim review: roughly 13% spending decline with about 97% of requests affirmed. Health Affairs, 2025, 10.1377/hlthaff.2025.01356; CMS Review Choice Demonstration FY2024 statistics. healthaffairs.org
15. Insurer pledge and scorecard: roughly 50 insurers covering 257 million Americans; 11% of requirements eliminated, 6.5 million fewer requests. AHIP and BCBSA, June 2025 and April 2026. Self-reported.
16. UnitedHealthcare elimination of home health prior authorization, effective April 1, 2025. UHC provider notice, March 2025.
17. 90% of MA enrollees in plans requiring prior authorization for home health. Thomas et al., Health Affairs Scholar, 2025.
18. MA home health users receive about 11% fewer visits, 18.7 versus 20.7. MedPAC, March 2025 analysis and June 2025 Report to Congress. medpac.gov
19. Prior authorization linked to racial and ethnic differences in end-of-life home health use. Jones et al., Journal of the American Geriatrics Society, 2026, 10.1111/jgs.70556.
20. Hospice carve-in termination, effective December 31, 2024. CMS VBID model announcements.
21. WISeR model design, vendor compensation reported at 10 to 20 percent of savings, Texas approval data of 62% first pass versus 84% after human review, EFF v. CMS lawsuit, congressional actions. CMS Innovation Center; Washington Post analysis; EFF filings, 2026.