The Healthcode System

The Healthcode System  ...  Chapter 1

There is a familiar complaint about American medicine, and it is correct as far as it goes. We do not have a healthcare system; we have a sick-care system  ...  one that pays handsomely to treat illness and has little financial reason to prevent it. The complaint is true. It is also unfinished. It names the appetite without naming the organ that feeds it.

Sick-care describes the incentive. It does not describe the machine. To find the machine, move a single letter. Close the gap in the middle of "health care" and you get the word printed on every hospital brochure and every insurer's mission statement. Insert a hyphen instead and you get a different word  ...  health-code  ...  and the second word is the one the system actually runs on.

Consider what has to happen for an act of medicine to count. A physician can take a careful history, reach the right diagnosis, choose the correct treatment, and deliver it with skill and patience. None of that, on its own, moves a dollar. The encounter must first be translated  ...  rendered into standardized strings that name the diagnosis, the procedure, the supply or the drug, and that together establish what the industry calls medical necessity. Submit the right codes and the work becomes reimbursable. Submit the wrong ones, or none at all, and the work  ...  however real, however skilled  ...  did not financially occur.

That is the quiet rule beneath all the noise: care that cannot be coded cannot be paid for, and care that cannot be paid for, in a system organized entirely around payment... struggles to exist. The code is not the record of the care. The code is the condition of the care.

Here is where the usual reform conversation goes wrong. It treats coding as friction  ...  a layer of paperwork bolted onto the real work of healing, regrettable but removable, the sort of thing a better software vendor or a leaner billing office might fix. The central claim of this series is harder than that. The code is not bolted onto the system. The code is the system. Healing is what gets to happen in whatever room the code leaves standing.

The two words sit a hyphen apart and a worldview apart. One promises that the institution exists to make people well. The other admits that the institution exists to process claims, and that wellness is a hoped-for byproduct of correctly processed claims. We say the first word aloud. We have built the second.

This is not an argument that anyone inside the system is acting in bad faith. The coder assigning the digits, the nurse charting at midnight, the physician clicking through templates  ...  they are not the problem; they are among the people the machine costs the most. The argument is structural. A system reveals its true purpose not in its slogans but in what it requires to function, and American medicine requires, above all else, the code.

The chapters ahead trace that requirement to its roots. How a claim actually becomes money. How a profession that once set its own fees handed the dictionary to a committee and the pricing to the government. What the apparatus costs  ...  in dollars, in hours, in trust  ...  and who pays. Why even the most thoughtful attempts to reward "value" keep running aground on the same hidden rock. What it would take, in the end, to build something honest enough to wear the unhyphenated name.

For now, only the reframing. Walk into any hospital in the country and read the sign above the door. It says health care, two clean words, no punctuation. Walk inside, follow the money, and you will find the hyphen everywhere  ...  in every claim, every denial, every necessity that had to be proven in five characters before the care itself could be paid for. The sign is aspirational. The hyphen is operational.

We have a health-code system. The rest of this series is about what that costs us, and what it would take to earn back the word we keep pretending we already mean.

The Healthcode System  ...  Chapter 2

Start with something ordinary. A child falls off a bike and lands wrong, and the wrist swells into a shape it should not hold. A parent drives to urgent care. A physician reads the X-ray, confirms the break, sets the bone, wraps the cast, and sends the child home with a lollipop and a follow-up date. The care is finished. The work  ...  the actual medicine  ...  is done.

Now the second job begins, the one the patient never sees. Before that visit can become money, it has to be translated out of the language of medicine and into the language of billing. That translation is the real subject of this chapter, because the translation is the precise point where healthcare quietly becomes health-code.

The translation runs through three dictionaries, each owned and maintained by a different authority. The first names what is wrong with the patient: the diagnosis codes, known as ICD-10. The second names what the provider did about it: the procedure codes, known as CPT  ...  a dictionary owned outright by a private association of physicians,[1] a fact the next chapter will have reason to revisit. The third names what was used in the doing  ...  the splint, the dose, the durable equipment  ...  under a set called HCPCS. A claim is a sentence assembled from these three vocabularies, and like any sentence, it can be ruled grammatical or thrown out as nonsense by something that never has to understand a word of it.

The diagnosis dictionary alone is vast. ICD-10 holds close to seventy thousand distinct codes, up from roughly fourteen thousand in the system it replaced[2]  ...  a nearly fivefold expansion in the number of ways a body can be officially described as broken. Each code is a string of three to seven characters, and the later characters carry a startling precision: not merely a fracture, but which bone, which end of it, which side of the body, and whether this is the first visit for the injury or a later one.[3]

Our child's wrist does not enter the system as a broken wrist. It enters as S52.501A  ...  a fracture of the lower end of the right radius, initial encounter, closed.[4] Change the arm and the code changes. The system needs to know which arm before it will agree to pay to fix it. There is a kind of marvel in that specificity, and a kind of madness, and the two are difficult to pull apart... which is the first hint that the dictionary was not built for the patient's benefit.

Specificity is only the entry fee. The codes do not merely describe the visit; they have to agree with one another. The diagnosis must justify the procedure. Payers keep lists  ...  local and national coverage determinations  ...  of exactly which diagnoses are accepted as grounds for which procedures, and those lists are not suggestions. They are the codes and conditions approved for payment, and the documentation has to support both halves or the claim does not stand.[5] The diagnosis, in this system, is not simply a fact about the patient. It is the argument for the bill.

Here is the part that ought to unsettle anyone still picturing a claim being read by a person. Most of the gate is automatic. Before any human reviews the visit, the claim runs through prepay software edits that check the codes against each other and against the approved tables. A procedure paired with a diagnosis the tables do not accept comes back denied  ...  flagged, in the trade, as a simple mismatch  ...  with no clinician ever looking at the chart.[6] Two legitimate services delivered in one honest visit can trip a denial on their own, not because the care was wrong but because the codes collided and no one appended the small modifier that tells the software to stand down.[7]

None of this changes what happened in the exam room. The wrist was broken; the wrist was set; the child went home mended. Whether that mending becomes money depends entirely on whether the translation satisfied the tables  ...  whether the sentence assembled from three dictionaries parsed as necessary in the eyes of a system that cannot see the child, only the string.

The code, again, is not the record of the care. The code is the condition of the care. A claim is medicine that has agreed to speak the only language the money understands... and the chapters ahead are about everything that gets lost in the translation.

Notes

1. The Current Procedural Terminology (CPT) code set was created by the American Medical Association in 1966 and remains its property; CMS uses it to price more than 10,000 services. Mercatus Center, "The Medicare Physician Fee Schedule: Overview, Influence on Healthcare Spending, and Policy Options."

2. ICD-10-CM contains roughly 69,800 diagnosis codes, against about 14,025 in ICD-9-CM. Rhode Island EOHHS, "ICD-10 Frequently Asked Questions"; Outsource Strategies International, "ICD-10-CM  ...  Overview, Significance and Top Diagnostic Codes."

3. ICD-10-CM codes run three to seven alphanumeric characters and encode laterality and encounter type, where ICD-9 codes ran three to five and mostly numeric. Rhode Island EOHHS, "ICD-10 FAQ"; U.S. National Library of Medicine, "The Coming Wave of Change: ICD-10."

4. S52.501A  ...  "Fracture of the lower end of the right radius, initial encounter for closed fracture." Auctus, "ICD-9 vs. ICD-10: Why The Change?"; Coronis Health, "The Difference Between ICD-9-CM and ICD-10-PCS."

5. Local and National Coverage Determinations list the CPT and ICD-10 codes approved for payment within a region; documentation must support both the procedure and the diagnosis reported. AAPC/CMS coverage-determination guidance.

6. NCCI edits are automatic prepay edits applied to claims based on the codes submitted; a triggering code pair is denied automatically, surfacing as a diagnosis–procedure mismatch (denial code CO-11) without clinical review. CMS, "Medicare NCCI FAQ Library"; Noridian, "NCCI Edits."

7. Example: an evaluation visit (CPT 99213) for Type 2 diabetes billed alongside a same-day skin-lesion removal can be auto-denied for want of Modifier 25. TheNexus, "CO 11 Denial Code: Description, Causes & Resolution."

The Healthcode System  ...  Chapter 3

A dictionary can feel like weather. The codes arrive each year, revised and reissued, and to the people who must use them they have the quality of climate: given, impersonal, no more authored than rainfall. That feeling is the most useful illusion the system keeps. Someone wrote the dictionary. Someone decides, every year, what the words are worth. Tracing who did the writing, and who holds the pen now, explains more about American medicine than any chart of premiums.

Begin in 1965, the year Medicare was signed into law and the federal government became the largest buyer of medical care in the country.[1] At the start the arrangement was almost quaint: physicians were paid according to their "usual, customary and reasonable" charges  ...  which is to say, more or less what they said they charged.[2] The doctor named the fee; Medicare, by and large, paid it. A profession accustomed to setting its own prices simply kept setting them, now with the Treasury as the payer.

Predictably, spending climbed faster than anyone had budgeted. The cure for an open-ended fee was a fixed one, and over two decades the government built the machinery to impose it. Hospitals went first. In 1983 Medicare began paying them by diagnosis-related group  ...  a flat, predetermined sum tied to the patient's diagnosis rather than to whatever the hospital spent, leaving the institution to absorb any costs above the set amount.[3] Physicians followed. Legislation in 1989 laid the groundwork, and in January 1992 the "usual and customary" world ended: each service would now be priced by a Resource-Based Relative Value Scale, the RBRVS, which assigns every procedure a bundle of relative value units and multiplies them by a national conversion factor to produce the fee.[4] The values are meant to capture the resources a service consumes  ...  physician work, practice overhead, malpractice exposure  ...  and the work component alone accounts for roughly half of most payments.[5]

Which raises the question that the whole system depends on not being asked too loudly: who decides what a procedure's work is worth? The answer is a committee most physicians have never heard of. Anticipating the 1992 change, the American Medical Association  ...  already the owner of the CPT procedure dictionary  ...  convened a body called the Specialty Society Relative Value Scale Update Committee, the RUC, in 1991.[6] Its method is straightforward and, from a certain angle, entirely reasonable. Specialty societies survey their own members on the time and difficulty a procedure demands; the specialists who actually perform that procedure present the proposed value to the committee; the committee forwards its recommendation to the government.[6] Who, after all, understands the work of a heart operation better than the cardiologists who do them?

The trouble lives inside that same sentence. The people best positioned to describe the value of a procedure are also the people paid by it. The committee that recommends the prices is convened by the association that owns the words. The government, holding the formal authority to decide, has agreed almost every time: by the AMA's own count, CMS has accepted more than ninety percent of the RUC's recommendations  ...  by some figures, more than ninety-four.[7] One health economist described the agency as accepting them slavishly. The AMA's position is that it merely advises, and that CMS makes every final call. Both things are true at once, which is precisely how the most durable arrangements work.

A committee weighted toward proceduralists tends, over time, to price procedures generously and to price the quieter work  ...  the conversation, the examination, the counsel that keeps a procedure from ever being needed  ...  comparatively thin. That tilt is not the scandal of any one meeting; it is a property of who sits in the room... Hold that thought. It returns, with consequences, several chapters from now.

One more fact closes the circuit. The RBRVS is not Medicare's private instrument. It is the scale used by Medicare and most other payers, the reference against which the wider commercial market sets its own rates.[8] When the committee revalues a code and the government assents, the new price ripples outward through nearly every insurer in the country. A single fee schedule, recommended in large part by the interested, governs what almost all of American medicine is worth.

The dictionary, then, is not weather. It was authored  ...  and the authors did not stop at the definitions. They stayed on to help set the prices, and the word for someone who both writes the language and decides what it is worth is not author but authority. We did not stumble into a health-code system. We appointed its lexicographers, handed them the pen, and agreed not to look too hard at the conflict of interest sitting in plain view... a habit the chapters ahead examine, as the bill for it comes due.

Notes

1. Medicare was signed into law in 1965, making the federal government the country's largest purchaser of medical care. Medical Economics, "Medicare reimbursement rates explained"; Mercatus Center, "The Medicare Physician Fee Schedule."

2. Under the original Medicare framework, physicians were reimbursed on "usual, customary and reasonable" (UCR) charges. Medical Economics, "Medicare reimbursement rates explained"; The Hospitalist, "Where 400 Years of Fee for Service Has Led Us."

3. In 1983 Medicare adopted prospective payment for hospitals via diagnosis-related groups (DRGs), a flat per-diagnosis sum that placed hospitals at financial risk for costs above the set amount. Medical Economics, "Medicare reimbursement rates explained."

4. The Omnibus Budget Reconciliation Act of 1989 laid the groundwork for the Resource-Based Relative Value Scale, effective January 1992, replacing UCR for physician services. Medical Economics, "Medicare reimbursement rates explained"; Mercatus Center, "The Medicare Physician Fee Schedule"; AAPC, "Learn the Three Rs of Physician Payment."

5. RBRVS prices each service from work, practice-expense, and malpractice relative value units times a conversion factor; the physician-work component averages slightly more than half of the total. American Medical Association, "RBRVS overview"; AAPC, "The Three Rs."

6. The AMA, which owns the CPT code set, convened the Specialty Society RVS Update Committee (RUC) in 1991; specialty societies survey their members and the specialists who perform a procedure present its proposed value to the committee, which forwards recommendations to CMS. AMA, "An Overview of the RUC Process"; KFF Health News, "Little-Known AMA Group Has Big Influence On Medicare Payments."

7. Historically CMS has accepted more than 90 percent of RUC recommendations  ...  more than 94 percent by AMA figures; economist Uwe Reinhardt characterized CMS as accepting them "slavishly," while the AMA holds that CMS makes all final decisions. KFF Health News, "Little-Known AMA Group"; Specialty Society Relative Value Scale Update Committee (encyclopedic overview citing Reinhardt); AMA, "RVS Update Committee (RUC)."

8. The RBRVS is the physician-payment scale used by CMS "and most other payers." American Medical Association, "RBRVS overview."

The Healthcode System  ...  Chapter 4

Every business carries overhead. It is the cost of keeping the lights on and the doors open: the rent, the utilities, the back office, the necessary expense that sits at the edge of an enterprise, distinct from the thing the business actually makes and sells. A reasonable question to ask of any industry is how large that edge has grown. Ask it of American medicine and the edge stops being an edge.

In 2017, by the most careful accounting available, insurers and providers in the United States spent eight hundred twelve billion dollars on administration  ...  roughly twenty-five hundred dollars for every person in the country, and a little more than a third of all national health spending.[1] A third of every health dollar. None of it sets a bone, reads a scan, or sits with a frightened patient. It pays for billing departments and claims processing, for prior authorizations and appeals, for the vast clerical machinery that exists to turn care into the codes of the previous chapters and to argue, endlessly, over whether the translation was done right.

This is not a law of nature, and the cleanest proof is to look north. Canada spends about seventeen percent of its health dollar on administration, half the American share, and reaches that figure through a payment system simple enough not to require the apparatus.[2] The pattern holds across borders. In a comparison of eight wealthy nations, American hospitals carried the heaviest administrative burden, a quarter of their total spending, while hospitals in Scotland and Canada ran at less than half that, and the lightest burdens of all fell where hospitals were funded by a single lump-sum budget rather than paid code by code.[3] The lesson is uncomfortable and unambiguous. The more a system pays for individual coded services, the more it must spend proving the codes. The overhead is not incidental to the health-code system. The overhead is what the health-code system costs to run.

The dollars are only half the bill. The other half is paid in the one currency medicine cannot print: the clinician's attention. Time-and-motion research finds that for every hour a physician spends with patients, nearly two more go to the electronic record and the desk work around it, with another one to two hours of documentation waiting at home after the children are asleep.[4] The exam room shrinks; the screen swells. A profession trained for years to read a body now spends most of its day feeding a database. That this ranks among the leading drivers of physician burnout is no mystery to anyone living it.[5]

Here is where the accounting word becomes a lie. We call it overhead  ...  the term itself insists the work sits overhead, above and apart from the real labor of healing, a manageable margin at the top of the ledger. It is not at the top of the ledger. A third of the money and the larger part of the clinician's day is not a margin around the system; it is the system... asserting itself in every coded line. The paperwork is not a tax on the economy of care. The paperwork is an economy  ...  with its own millions of workers, its own software vendors, its own quarterly growth and its own career ladders  ...  and the product it manufactures is not health. The product is the successful conversion of a human being's care into a string a payer will honor.

Recall the rule from the first chapter: the code is not the record of the care, the code is the condition of the care. Four chapters in, that condition has a price tag and a payroll. It employs more people than many of the country's largest industries, consumes a third of the most expensive health system on earth, and quietly takes the scarcest thing a sick person needs, which is a clinician's undivided minute.

The overhead, then, is not overhead. It is the building. The patient pays for it twice  ...  once in the bill, once in the rushed and screen-lit visit. The public pays for it in a third of every health dollar that buys no health at all. All of this, the staggering sum and the borrowed hours and the burnout, describes only the cost of claims the system means to pay. The next chapter takes up the other kind: the claims engineered, from the start, to be denied.

Notes

1. In 2017 U.S. insurers and providers spent $812 billion on administration  ...  $2,497 per capita, or 34.2% of national health expenditures. David U. Himmelstein, Terry Campbell, and Steffie Woolhandler, "Health Care Administrative Costs in the United States and Canada, 2017," Annals of Internal Medicine (2020).

2. Canada spent roughly 17.0% of its health expenditures on administration, about $551 per capita  ...  half the U.S. share. Himmelstein, Campbell, and Woolhandler, Annals of Internal Medicine (2020); coverage in Time and Healio.

3. In an eight-nation comparison, U.S. hospital administrative costs were the highest at 25.3% of total hospital spending (2011), against 11.6% in Scotland and 12.4% in Canada; the lowest burdens occurred where hospitals received single lump-sum (global) budgets. David U. Himmelstein et al., "A Comparison of Hospital Administrative Costs in Eight Nations," Health Affairs (2014); Commonwealth Fund summary.

4. Time-and-motion research finds physicians spend nearly two hours on electronic-record and desk work for every hour of direct patient care, plus one to two additional hours of documentation after hours. Altarum Healthcare Value Hub, "Excess Administrative Spending in Healthcare," citing time-and-motion study data.

5. Administrative burden is identified as a key driver of physician burnout. Altarum Healthcare Value Hub, "Excess Administrative Spending in Healthcare."

The Healthcode System  ...  Chapter 5

The last chapter described the cost of the claims the system means to pay. This one is about the claims it means to refuse, and refusal turns out to have an architecture of its own.

In the systems that run the apparatus there is a security posture called default deny. The phrase describes a gate built to refuse everything by default and to admit only what clears an explicit exception. It is a sensible setting for a firewall. Applied to a sick person's claim, it becomes the quiet governing logic of American coverage: the default answer is no, and yes is the thing that has to be won.

The refusals are not holding steady; they are climbing. Between 2022 and 2023, one industry analysis found denials rising by roughly a fifth on commercial claims and by more than half on Medicare Advantage.[1] Whatever else is true of the gate, it is closing faster.

The gate has also moved. It used to sit behind the care, deciding after the fact whether to pay. Now, through prior authorization, it sits in front of the care, deciding whether the treatment may happen at all. Ask the physicians who spend their days at that gate what it does. In the American Medical Association's annual survey, more than nine in ten report that prior authorization delays necessary care, and roughly four in five say patients, worn down by the process, abandon treatment that was clinically indicated.[2] The figure that should stop a reader cold: about one in four physicians report that prior authorization has led to a serious adverse event for a patient in their care  ...  hospitalization, permanent harm, or death.[3] This is not overhead at the edge of the system. This is the apparatus reaching past the bill and onto the gurney. The doctors pay their share too, in the thirty-nine authorization requests and thirteen hours the average physician now loses to the process every week.[4]

Here is the detail that turns an inefficiency into an indictment. When denials are challenged, they collapse. The same analysis found a health system reporting that more than half of its denied Medicare Advantage claims were overturned on appeal.[1] Sit with the arithmetic. A denial reversed on appeal was, by the payer's own eventual admission, a claim that should have been paid the first time. A handful of such reversals is error. A majority reversed is not error... it is method. A system that refuses valid claims and pays them only when pressed is not malfunctioning. It is working exactly as designed, with the burden of proof inverted and handed to the patient.

In recent years the denial machine has stopped being a metaphor. Investigative reporting and a wave of lawsuits describe claim refusal carried out at the speed and scale of software. ProPublica reported that one large insurer used an automated system to deny more than three hundred thousand claims in two months, with physicians spending an average of 1.2 seconds of review on each, signing off on batches without opening individual files.[5] The insurer responded that the tool is a simple code-matching process, that most claims run through it are paid, and that it operates only after care has been delivered.[5] A separate class action alleges that another insurer leaned on an algorithm to cut off post-acute care for elderly patients, pressuring staff to keep nursing stays within one percent of the model's prediction, and that roughly nine in ten of the resulting denials were reversed when appealed; the company maintains the tool only guides clinicians and does not itself decide coverage.[6] Courts have allowed the central claims to proceed and ordered the algorithm disclosed.[6] The facts remain contested. What is not contested is the architecture they reveal: adjudication at machine scale, with human review compressed toward zero.

Default deny works for one reason above all others, and it is the coldest fact in this chapter. Contesting a refusal is costly, confusing, and rare. By the allegations in that same case, only about two patients in a thousand ever appeal.[6] Run that figure against a ninety percent reversal rate and the business logic resolves into something stark: the system can be wrong nearly every time it is questioned and still come out ahead, because it is questioned almost never. The denial does not need to be correct. It only needs to go uncontested. The house edge is not accuracy. It is exhaustion.

Chapter two established that under this system care must prove its necessity in code before the money will move. This chapter is what proving it has become: a gate whose default is refusal, widened by software, and kept profitable by the near-certainty that the patient lacks the time, the health, or the stamina to fight back. The cruelty is not in any single no. It is in a design that earns its margin from the friction it manufactures.

The worst of it is still not the care wrongly refused. It is the care that was never assigned a code worth offering  ...  the prevention the system cannot see, and will not pay for, no matter how cleanly the claim is filed. That is the hinge of this entire argument, and the subject of the next chapter.

Notes

1. Between 2022 and 2023, care denials rose an average of 20.2% on commercial claims and 55.7% on Medicare Advantage claims; one health system reported MA denial rates of 10.5%–15.5%, with as much as 56% of denied claims overturned on appeal. American Hospital Association, "Skyrocketing Hospital Administrative Costs, Burdensome Commercial Insurer Policies Are Impacting Patient Care" (September 2024), citing Strata Decision Technology and system-reported data.

2. In the AMA's 2024 prior-authorization survey, roughly 94% of physicians reported that prior authorization delays necessary care and about four in five reported that patients abandon clinically indicated treatment because of it. American Medical Association, "2024 Prior Authorization Physician Survey."

3. Roughly one in four physicians (24%–29% across recent survey years) reported that prior authorization had led to a serious adverse event for a patient in their care, including hospitalization, permanent impairment, or death. AMA, "2024 Prior Authorization Physician Survey"; AMA press release, "AMA survey indicates prior authorization wreaks havoc on patient care" (June 2024).

4. Physicians complete an average of 39 prior authorizations per week, spending about 13 hours on the process. AMA, "2024 Prior Authorization Physician Survey."

5. ProPublica reported that Cigna's PxDx system was used to deny more than 300,000 claims over two months in 2022, with physician reviewers spending an average of 1.2 seconds per claim and signing batches without opening individual files; Cigna characterized PxDx as a code-matching/sorting process that mostly results in payment and operates after care is delivered. ProPublica (March 2023); Healthcare Dive; company statements.

6. A class action concerning UnitedHealth/Optum's use of the nH Predict algorithm (developed by naviHealth) alleges it was used to deny post-acute care for Medicare Advantage members, that staff were pressured to keep skilled-nursing stays within 1% of the model's projection, that roughly 90% of resulting denials were reversed on appeal, and that only about 0.2% of patients appeal; UnitedHealth states the tool guides clinicians and does not make coverage decisions. Federal courts allowed breach-of-contract and good-faith claims to proceed (2025) and later compelled disclosure of the algorithm. STAT investigation; class-action filings; Healthcare Finance News; Bloomberg Law.

The Healthcode System  ...  Chapter 6

Every system built on codes obeys one quiet law, and it is the law on which this entire argument turns. What gets coded gets done. What cannot be coded slips, gradually and without anyone deciding it should, out of existence  ...  not because the system rules against it, but because the system has no way to see it, pay it, or count it. A code is a kind of attention. Whatever falls outside the codes falls outside the system's attention, and in medicine, attention is the difference between a thing that happens and a thing that does not.

The law operates first inside the codeable, where the apparatus plays favorites, and its favorites are procedures. Recall the committee from the third chapter, weighted toward the specialists who perform the interventions, and the seed planted there: a body of proceduralists tends to price procedures generously and to price the quieter work thin. To this day that committee reserves only about a fifth of its seats for primary care, though those physicians handle better than a third of the nation's visits.[1] The tilt it built into the fee schedule three decades ago has compounded ever since, until what began as a tendency became a chasm. In 2022 a primary-care office visit drew an average of two hundred fifty-nine dollars, while a single gastroenterology procedure drew one thousand ninety-two  ...  the conversation worth roughly a fifth of the scope.[2] The physicians who perform the procedures out-earn the physicians who do the thinking by well over a hundred thousand dollars a year.[3] Primary care, the place where most prevention actually lives, receives less than five cents of every health dollar, a share that has been falling, not rising.[4] The market reads these prices and responds with perfect rationality. Students choose the lucrative specialties; practices invest in the equipment, not the hour of counsel. What gets coded richly gets done abundantly. What gets coded thinly gets done rarely. The fee schedule is not describing medicine's values. It is dictating them.

A rebalanced fee schedule could narrow that chasm, and the chapters ahead argue it should. Beneath it, though, lies a problem no adjustment of prices can reach, because it is not a problem of price. It is a problem of grammar. A code names an event. Prevention's masterpiece is a non-event. Consider the difference the ledger sees. A coronary stent is a richly coded procedure, a clean and billable event with a date and a price. The years of steady blood-pressure control that meant the stent was never needed are scattered across a hundred unremarkable visits and resolve, on the ledger, into almost nothing. A cancer removed at stage three is a long cascade of well-paid events. The same cancer caught early at a screening and never permitted to become a crisis is one cheap line item followed by silence, and the system reads silence as the absence of value, because silence is precisely what it cannot bill. You cannot code an absence. The disaster has a price. The disaster averted leaves no trace the apparatus knows how to read.

Painters have a name for this. Negative space is the shape of everything surrounding the subject  ...  the void that gives the figure its edges. A skilled hand can draw a vase by drawing only the space around it, letting the absence describe the form. Medicine's billing apparatus can do the opposite and nothing more: it renders the figure and never the ground. It captures every line of the disease  ...  each procedure, each crisis, each coded intervention  ...  and none of the quiet that good health actually consists of. A system that pays only for the figure will, over the decades, produce exactly what it pays for: a medicine crowded with events and starved of the calm that prevents them. We built an apparatus that can see sickness in exquisite detail and cannot see health at all, because health, rendered honestly, is mostly negative space.

Two different invisibilities hide in that negative space, and the rest of this series turns on telling them apart, because they call for two different cures.

The first is the undervaluing of clinical prevention  ...  the screening, the counsel, the patient management of a chronic disease that the code can see but consistently underpays. That invisibility a better payment model can largely correct, by rewarding the outcome rather than the event, the health of the patient rather than the volume of the procedures. Whether such models actually work, and what an honest one would demand, is the subject of the next two chapters.

The second invisibility runs deeper and lies outside medicine altogether. The stable housing, the reliable food, the years of schooling, the breathable air, the paycheck that does not vanish: these shape a human being's health long before any clinician enters the story, and they answer to no code, arrive on no claim, and fit inside no fee schedule, however enlightened. No reform of medical payment reaches them, because they were never medicine's to bill. Reaching them requires something the health-code system was not built to contain and cannot be retrofitted to hold. Naming that thing, and what it would take to build it, is the work of the final chapters.

The first chapter offered a rule: the code is the condition of the care. This chapter supplies the corollary that turns it from observation into indictment. If the code is the condition of the care, then care the code cannot name is care that, under this system, will not happen  ...  and the most valuable medicine we possess, the prevention of suffering before it begins, is precisely the medicine the code was never able to name. We did not sit down and choose to neglect prevention. We built a machine that cannot see it... and then mistook everything the machine could see for the whole of health.

Notes

1. Primary care physicians hold roughly 19% of the seats on the AMA/Specialty Society RVS Update Committee (RUC) while accounting for nearly a quarter of the physician workforce and about 35% of all patient visits. Commonwealth Fund, "Improving Payments for Primary Care Physicians" (2025).

2. In 2022, primary care physicians' reimbursement averaged about $259 per visit, compared with roughly $1,092 for gastroenterology; primary care's per-visit revenue is approximately one-fifth that of procedure-heavy specialties. Milbank Memorial Fund / Robert Graham Center, "The Health of US Primary Care: 2025 Scorecard Report."

3. In 2023 the average specialist salary was about $394,000, against roughly $277,000 for a primary care physician. Health Affairs Forefront, "Valuing Cognitive Effort in Primary Care: Rebalancing Medicare Physician Payment."

4. Primary care accounted for under 5% of total U.S. health spending in 2022  ...  about 3.4% within Medicare and 4.3% within Medicaid  ...  and the share has been declining. Milbank Memorial Fund and The Physicians Foundation, "The Health of US Primary Care: 2025 Scorecard Report."

The Healthcode System  ...  Chapter 7

The obvious answer to everything in the last three chapters is to stop paying for procedures and start paying for health. Stop rewarding the volume of things done to a patient; reward instead whether the patient got well, and stayed well. The instinct is correct. It has a name  ...  value-based care  ...  and, after more than a decade of trial, a record long enough to judge honestly. This chapter judges it before the next one proposes, because a proposal that has not reckoned with the ledger is not a plan. It is a wish.

The reform has been pursued in earnest. The Affordable Care Act built a federal laboratory, the Center for Medicare and Medicaid Innovation, to design and test payment models that reward outcomes over volume: accountable care organizations, bundled payments, and dozens of variations besides. Over its first decade the center launched roughly fifty models touching more than forty million patients.[1] No one can accuse the system of failing to try the idea.

The honest ledger is sobering. The innovation center was projected to reduce net federal spending by some billions in its first decade; instead, by the Congressional Budget Office's accounting, it increased net spending by 5.4 billion dollars, with further increases projected for the decade after.[2] Of forty-nine models with published evaluations, only six produced statistically significant savings  ...  all of them modest, none exceeding a few hundred million dollars a year  ...  and only four were ever certified to expand.[3] An independent review of the more mature models found the results split in rough thirds: a third saved real money, a third lost it, and a third did little either way, with quality effects that were similarly mixed.[4] This is not the case for the prosecution. It is the record, kept by the people who believe in the reform.

The record is not uniformly gray, and the bright spots matter more than the disappointments, because they show where the instinct holds. Medicare's largest accountable-care effort, the Shared Savings Program, now saves on the order of two billion dollars a year; its strongest performers, year after year, are the organizations led by primary care.[5] The lesson rhymes with the previous chapter: put the money behind the thinking, the coordination, the prevention, and the savings follow. Paying for value is not wrong. Where it has been pursued seriously and given time, it works. The trouble is what "value" has been forced to become inside a system made of codes.

You cannot pay for value without measuring it, and to measure value in this system is to generate more codes. Quality is assessed by extracting coded metrics from the medical record. Accountable care organizations identify their patients and tally their savings through claims, which are codes. The main program meant to reward physician quality grew into a reporting burden of its own, onerous and bound to the same software, until clinicians spent yet more of the day documenting that they had been good rather than being good.[6] Value-based care did not leave the health-code system behind. It bolted a second story of codes  ...  quality codes  ...  on top of the first.

What follows is the oldest hazard in measurement: once a metric becomes a target, the work bends toward the metric rather than the goal it was meant to capture. The reform changed the question from how much did you do to how well did you score, and the answer is still written in code, still gameable, still adjudicated by the apparatus that produced the problem. A system cannot measure its way out of measurement. It can only choose its measures more wisely... which is worth doing, and is not the same as being cured.

These are half measures in two senses, and both matter. In the plain sense, the results so far are fractional  ...  a third of the models, modest sums, set against a system counted in trillions. In the deeper sense, even the genuine successes only reweight the codeable. They shift payment from procedures toward outcomes, which is right and necessary and overdue. They do not, and cannot, reach past the code to the negative space the last chapter described: the prevention that throws off no codeable metric, and the housing and food and air that generate no claim at all. Paying for value is half the cure. It is the necessary half. It is not the whole, and pretending otherwise is how good reforms get oversold, then abandoned when they fail to do the thing they were never built to do.

The next chapter takes the instinct seriously rather than incrementally. It asks what a payment model would look like if it staggered the majority of a provider's income against the totality of patient and population outcomes  ...  not a bonus at the margin, but the main event  ...  and it asks, just as seriously, what such a model would demand, what it would risk, and where, even at its best, it would still run out of reach.

Notes

1. The Center for Medicare and Medicaid Innovation (CMMI), created by the Affordable Care Act in 2010, launched roughly fifty payment and delivery models touching more than forty million beneficiaries in its first decade. Testimony before the House Energy and Commerce Committee, via Healthcare Dive; CMS.

2. CMMI was initially expected to reduce net federal spending (by roughly $2.8 billion across 2011–2020) but instead increased net direct spending by $5.4 billion over that period, with an additional $1.3 billion projected for 2021–2030. Congressional Budget Office (2023), via Healthcare Dive and the Paragon Health Institute.

3. Of forty-nine CMMI models with published evaluations, only six generated statistically significant savings  ...  all modest, none exceeding roughly $220 million annually  ...  and only four were certified for expansion. Congressional Budget Office (2023); Niskanen Center.

4. An Avalere Health review (December 2024) of eighteen mature CMMI models found roughly one-third produced substantial net savings, one-third substantial net losses, and one-third nominal effects, with mixed quality impacts. Avalere Health, via TechTarget and Healthcare Dive.

5. The Medicare Shared Savings Program saves on the order of $2 billion a year (more than $1.8 billion in 2022), with primary-care-led accountable care organizations consistently the strongest performers. CMS; Niskanen Center.

6. Quality measurement under value-based programs  ...  including the Merit-based Incentive Payment System (MIPS)  ...  relies on extracting coded clinical measures from provider records and has been widely criticized as an onerous, costly reporting burden. Health-policy commentary on MIPS and clinical quality reporting.

The Healthcode System  ...  Chapter 8

The previous chapter ended on a phrase: not a bonus at the margin, but the main event. This chapter is about what changes when the main event is the only thing that pays.

Begin with the premise that follows from everything before it. If a provider's livelihood depended not on the procedures performed but on whether the people in their care got well and stayed well, every incentive this series has catalogued would invert. The code would stop being the thing chased and become, at most, the thing recorded. Recall the sixth chapter's hardest line: that you cannot code an absence, that the disaster averted leaves no trace the apparatus can read. The proposal here is the attempt to pay, at last, for that absence  ...  for the stent that was never needed, the admission that never happened, the population quietly kept out of the hospital.

The mechanism has two parts, and the first is a matter of proportion. The majority of a provider's income  ...  not a sliver, not a year-end bonus, the majority  ...  would be staggered against the totality of patient and population outcomes over time. Each word carries weight. Totality means the whole arc of a panel's health, not a single encounter or a single tidy metric. Staggered means paid in tranches as those outcomes accrue and hold, rather than settled the moment a service is rendered. Majority means this is the structure of the income, not a garnish on fee-for-service. When most of the money rides on whether people are well, the daily question in the exam room shifts from what can I bill to what will actually help, and the code subsides from master to clerk.

The second part answers the objection the first invites. Outcomes depend on the patients one begins with. A clinic serving the poor, the sick, and the unhoused will post worse raw numbers than a concierge practice in a wealthy suburb, through no fault of its medicine. A model that paid on absolute outcomes would simply reward doctors for choosing easy patients. So the incremental metrics are benchmarked not against an ideal but against the system-wide average for comparable populations: a provider is paid for moving their panel's health relative to where similar panels sit, and for improving it over time. The reward goes to the clinic that takes the hardest cases and moves them, not to the one that started downhill.

That is the proposal. The rest of this chapter is the part most proposals skip  ...  the reckoning with what it would demand and where it would strain, because a reform sold without its difficulties is the kind that gets oversold and then abandoned.

The first difficulty is attribution. Health is multi-causal and care is fragmented across many hands. When a diabetic patient improves, the credit belongs in some unknowable proportion to the primary physician, the endocrinologist, the pharmacist, the patient's own resolve, and the grocery store that finally opened nearby. No formula assigns that credit cleanly, and any model claiming to is lying. The honest response is to attribute at the level of the team and the population rather than the single encounter, to accept imprecision, and to let the benchmarking wash the noise out across a large enough panel. Imperfect, and still an improvement on pretending the procedure code was ever the unit of value.

The second difficulty is the one this entire series should teach us to expect: the moment you pay for a number, the number becomes a target, and in this system the number is still a code. The cautionary tale is already running. Medicare Advantage pays health plans according to each enrollee's risk score, a number built from diagnosis codes  ...  and the result is the textbook abuse. The Medicare Payment Advisory Commission projects that the program will be overpaid by roughly eighty-four billion dollars in 2025, of which some forty billion traces to "coding intensity," the practice of recording extra diagnoses to make patients appear sicker and lift the score; much of the remainder comes from quietly enrolling the healthier.[1] Nearly every large plan has faced allegations of this upcoding.[2] The lesson is not that paying for the patient is hopeless. The lesson is that any such model must be built from the start in the knowledge that it will be gamed  ...  with auditing, with risk adjustment that resists inflation, with benchmarks that penalize cherry-picking  ...  and that even then it will be gamed at the edges. The aim is not a system that cannot be gamed. The aim is one in which the honest path pays better than the gamed one.

The third difficulty is time, and it presses toward the next chapter. Outcomes accrue over years; prevention's dividend may arrive a decade later and land in another payer's budget, after the patient has changed jobs, switched plans, or aged into Medicare. A payment cycle runs a year. That mismatch is real, and it is why even this model, generous as it is toward prevention, strains at prevention's longest horizons.

The fourth difficulty is the transition itself. A sudden switch would break the providers running on the thinnest margins, the rural hospitals and safety-net clinics that serve precisely the populations hardest to move. The change must be staged: a rising floor of outcome-based payment over years, money advanced for the care coordinators and data systems that outcomes demand, and protection for those carrying the sickest panels while they build the capacity to succeed. Done carelessly, the cure bankrupts the patients who need it most.

A fair reader will notice that none of this escapes measurement, and the seventh chapter warned exactly that. The defense is precise. This model does not abolish the code; it aims it. A procedure code asks what was done to the patient. An outcome measure asks whether the patient is well. Both are imperfect proxies and both can be gamed, yet one points toward the goal and the other points away from it. Moving the majority of the money from the first question to the second does not free medicine from codes... it points the codes, for the first time, at health.

Even aimed perfectly, the model reaches only as far as a clinician's panel can plausibly influence. It can pay for the population that never needed the stent. It cannot pay for the housing that kept the asthma at bay, the wage that bought the vegetables, the schooling that let a patient read the label and follow the dose. Those belong to no provider's panel and answer to no payment cycle. This is the necessary half  ...  the largest, most honestly decoupled half a payment model can deliver. The other half was never medicine's to bill, and building it is the work of the final chapter.

Notes

1. The Medicare Payment Advisory Commission (MedPAC) projected, in its March 2025 report, that Medicare would spend roughly 20% more on Medicare Advantage enrollees than on comparable traditional-Medicare beneficiaries  ...  about $84 billion in 2025  ...  with approximately $40 billion attributable to "coding intensity" (recording additional diagnoses to raise risk scores) and much of the remainder to favorable selection. MedPAC, Report to Congress (March 2025); Becker's Payer Issues; RISE Health.

2. Nearly every major Medicare Advantage plan has faced or settled allegations of upcoding, and the HHS Office of Inspector General identified billions of dollars in questionable risk-adjustment payments tied to in-home assessments and chart reviews. Becker's Payer Issues; HHS Office of Inspector General.

The Healthcode System  ...  Chapter 9

The previous chapter aimed the codes at health, then admitted how far that aim could carry: as far as a clinician's panel can reach, and no farther. This chapter is about everything past that range, which turns out to be most of what keeps a person well.

There is an old parable about a village beside a river. People keep washing downstream, drowning, and the villagers grow expert at pulling them out  ...  better nets, faster boats, a whole heroic apparatus of rescue. No one walks upstream to ask why so many keep falling in. American medicine is that village, and it has built the finest rescue operation in the history of the world at the exact point where people are already in the water. The question this chapter asks is why we send almost nothing to the headwaters.

The case for the headwaters is not sentimental; it is measured. When researchers weigh what actually determines a population's length and quality of life, medical care accounts for only about a fifth of the result.[1] The largest share, roughly forty percent, belongs to social and economic conditions: income, education, employment, the safety of a street. Another thirty percent tracks behavior, and ten percent the physical environment.[1] The physician, custodian of the most expensive lever, commands the shortest one... Four-fifths of what makes us well is decided before anyone reaches an exam room.

We have arranged our spending in near-perfect opposition to that fact. The United States pays more for medical care than any nation on earth, close to a fifth of its entire economy; yet when medical and social spending are summed together, it sits near the middle of its peers, with middling outcomes to match.[2] The reason is the ratio. The country holds the lowest proportion of social-to-medical spending in the developed world  ...  where comparable nations spend roughly two dollars on social supports for every medical dollar, the United States spends about one.[2] That ratio predicts health: nations and states that lean toward social spending post better outcomes, a relationship that holds even when the United States is dropped from the data.[3] Meanwhile public health and prevention, the work aimed directly at the headwaters, receives roughly three cents of every health dollar, a share that has fallen for two decades.[4] We built a five-trillion-dollar rescue and starved the work that would keep people out of the river.

No payment model reaches the headwaters, and the previous chapter's cannot either. At its best it pays for the population that never needed the stent  ...  real, and still downstream. The determinants live beyond every payment model for reasons that are finally one reason. They belong to no provider's panel: a landlord, a school board, an employer, a zoning commission shape them, and none of them bills Medicare. They ripen across decades, the lead paint scraped from a child's wall paying its dividend forty years on, in a body the original payer will never meet. They span sectors that do not answer to the health system at all. They resist attribution by their very nature  ...  which is the same reason the code was always blind to them, because their product is the negative space of the sixth chapter, the crisis that never arrived, spread thin across a whole neighborhood and a whole life. No claim, no panel, no annual cycle can hold a thing like that.

The reflex of a system made of codes is to meet this with more codes. There is already a movement to give social needs their own diagnostic codes, to have hospitals screen for hunger and housing and bill for the screening, to medicalize the determinants so the apparatus can finally see them. The instinct is generous and the error is precisely the one this series has followed from its first page. To code housing as a health intervention is to subordinate the home to the logic of the claim  ...  to make shelter legible only as a line item, fundable only once a clinician has documented a patient's need, governed again by the machinery whose blindness created the problem. The headwaters cannot be saved by being poured into the code. They have to be funded and governed on their own terms, which is to say outside the health-code system altogether.

What that outside looks like has a shape, and the shape is not medicine's. It rests on three supports. The first is dedicated public health funding  ...  a stream that does not run through the claims system and does not compete, dollar against dollar, with the next imaging suite; sustained rather than surged in a crisis and abandoned after it; aimed at conditions instead of encounters. The second is community-led power, because the determinants are local and lived, and the people inside a place, not a distant payer, are the ones who can say what its health requires. Prevention done to a community tends to fail where prevention directed by one tends to hold. The third is cross-sector accountability, because health is manufactured in housing and schools and workplaces and air, and the responsibility for it must span those sectors rather than sit orphaned inside a starved health department  ...  health counted as a result of housing policy, education policy, labor and environmental policy, and not quarantined as the private business of medicine.

Honesty requires admitting this is the harder of the two reforms, because pretending otherwise would repeat exactly the overselling the seventh chapter warned against. It is harder politically: the headwaters have no lobby the size of the medical industry, no charge that posts automatically, no constituency organized around a billing code. It is harder to measure: the returns are diffuse, delayed, and stubborn to attribute, and the very qualities that keep them off the claim form keep them off the budget in a country that asks each dollar to prove itself by next quarter. The returns are real all the same  ...  the spending-ratio studies are not subtle, and well-built community programs can return many dollars for each one spent. The fair comparison was never against perfection. It is against the status quo, which is the most expensive failure available: a five-trillion-dollar machine, outgrowing the economy that feeds it, buying outcomes its peers reach for half the price.

The argument now stands at its full height, and it has two hands. With one, aim the codes at health, so that the majority of medicine's money rewards the patient who got well rather than the procedure performed. With the other, build  ...  outside the codes, on its own terms  ...  the public health and social infrastructure that produces health long before a patient is ever made. One reform turns the apparatus toward the right target. The other funds the work the apparatus was never able to see. Neither alone is enough: the first without the second is a sick-care system with better aim; the second without the first leaves the old machine standing to swallow the savings. Together, they would amount at last to the thing this country keeps claiming to possess and has never once built.

What that thing should be called, and what it would ask of us to mean the word honestly, is the subject of the last word that follows.

Notes

1. The County Health Rankings model weights modifiable determinants of health as approximately 40% social and economic factors, 30% health behaviors, 20% clinical care, and 10% physical environment; an empirical validation of the model placed clinical care's contribution near 16%. University of Wisconsin Population Health Institute and the Robert Wood Johnson Foundation, County Health Rankings model; Hood, Gennuso, Swain & Catlin, "County Health Rankings: Relationships Between Determinant Factors and Health Outcomes," American Journal of Preventive Medicine (2016).

2. The U.S. spends more on medical care than any nation (about 18% of GDP) but ranks near the OECD average once medical and social-service spending are summed; it has the lowest ratio of social-service to health-care spending in the OECD, where peer nations spend roughly two dollars on social services for each medical dollar and the U.S. spends about one. Elizabeth H. Bradley and Lauren A. Taylor, The American Health Care Paradox (2013); Bradley, Sipsma & Taylor, QJM (2017).

3. The ratio of social-service to health-care spending is significantly associated with better health outcomes across OECD countries and across U.S. states  ...  a relationship that held even when the United States was excluded from the analysis. Bradley et al., Health Affairs (2011 and 2016); Robert Wood Johnson Foundation.

Government public health and prevention account for roughly 3% of U.S. health spending (about 3.3%, or $160 billion, in 2023; preventive care about 2.9% in 2018), a share that has declined since roughly 2002. Trust for America's Health, "The Impact of Chronic Underfunding on America's Public Health System" (2020); Bishai, Leider & Resnick, "Public Health's Falling Share of US Health Spending," American Journal of Public Health (2016); CMS National Health Expenditure Accounts.