Case Study 2 — The Capitated Practice: A Contract Structure That Failed, and Came Back

A composite built from documented industry patterns, clearly labeled as such. The regulatory and structural facts are Tier 1; the practice in the narrative is Tier 3 and constructed.

Background

Section 2.6 listed capitation as one of five contract structures and described it in a paragraph. It deserves a case study, because it is the structure that most completely changes what a provider organization is — and because American healthcare has now tried it twice, failed the first time, and is trying again with substantially different machinery.

Capitation pays a provider a fixed amount per member per month for an attributed population, regardless of whether or how often those members are seen. The primary care group described in §2.6 — 5,200 members at \$29.50 PMPM — receives \$153,400 a month whether it delivers 1,100 visits or 3,400.

The arithmetic inverts every incentive in fee-for-service. Under a fee schedule, a visit is revenue. Under capitation, a visit is cost. That is the entire point: capitation is designed to reward keeping a population healthy rather than treating it frequently, and to make prevention financially rational rather than financially punished.

It is also, and this is not a criticism, a transfer of insurance risk from the payer to the provider. A group accepting capitation has agreed to deliver whatever care its attributed members need for a fixed budget. If those members turn out to be sicker than the rate assumed, the group absorbs it.

The issue: what went wrong the first time

Capitation spread widely in the United States during the managed-care expansion of the 1990s, and in a substantial number of cases it failed — sometimes spectacularly, with physician groups and independent practice associations becoming insolvent. The documented failure modes are worth naming one by one, because every one of them is a lesson about contract structure that applies today.

1. The rate was set without adequate risk adjustment. Early capitation rates were frequently adjusted only for age and sex. A group whose attributed panel happened to include a disproportionate share of members with serious chronic illness received the same PMPM as a group with a healthy panel and delivered several times the care. There was no mechanism to correct it, because the mechanism did not yet exist. This is precisely the problem hierarchical condition categories were built to solve, and it is why Chapter 36 exists.

2. Adverse selection was not accounted for. Groups with strong reputations for treating complex conditions attracted complex patients, and were penalized for it by a payment system that did not recognize the difference. The better the group was at a specialty, the worse its capitation economics.

3. Groups accepted risk for services they did not control. A primary care group capitated for "professional services" might find itself financially responsible for specialist referrals, outpatient procedures, or pharmacy — categories it could influence but not determine. When a member required an expensive intervention appropriately ordered by someone else, the budget absorbed it.

4. There was no stop-loss. A single catastrophic case could consume months of capitation revenue. Sophisticated arrangements now include reinsurance or stop-loss provisions above a threshold; early ones frequently did not.

5. The financial reporting did not exist. A group paid per service knows its revenue from its charge file. A group paid per member per month needs to know its cost per member, which requires data infrastructure that most practices did not have and many still do not. Groups discovered they were losing money on capitation months or quarters after it started.

6. And the incentive worked in both directions. If a visit is a cost, then withholding a visit is a saving. The 1990s produced a durable public backlash against managed care built substantially on this perception — that a payment structure designed to reward prevention could also reward denial of care. Whether and how often that actually occurred is genuinely contested; that the perception ended the political viability of the arrangement is not.

The composite

The following is constructed from these documented patterns. It is not a real organization.

A twelve-physician primary care group signs a capitation agreement covering roughly 9,000 attributed members at a rate negotiated from the payer's historical claims experience for a similar population. The rate is adjusted for age and sex only.

Year one goes well. Visit volume is close to the assumption. The group's monthly revenue is predictable for the first time in its history, which the administrator finds genuinely wonderful, and it invests in evening hours and a nurse-triage line — both of which are pure cost under fee-for-service and pure investment under capitation.

Year two goes badly, for three reasons that compound.

The payer's marketing succeeds and enrollment grows, but the new members skew older and sicker than the existing panel. The rate does not change, because nothing in the contract adjusts it for acuity.

Second, the contract makes the group financially responsible for outpatient specialty referrals. A new orthopedic practice opens nearby and referral volume rises. The group's physicians are not ordering inappropriately; the referrals are indicated. The budget does not care.

Third — and this is the failure that is entirely the group's own — nobody is measuring cost per member per month. The practice management system reports charges and collections, and under capitation both are nearly meaningless. The group discovers the problem when its cash position deteriorates, which is roughly two quarters after the problem began.

The group renegotiates: risk adjustment based on documented diagnoses, a carve-out removing specialty referrals from the capitated budget, and stop-loss above a per-member threshold. All three of those concessions are things it should have negotiated at the start, and it obtained them only after demonstrating a loss.

What it shows

First, contract structure is not a detail; it is the business model. Section 2.6 presented five structures as a list. This case study is the argument that the list is the most consequential page in the chapter for anyone running an organization. A group that switches from fee schedule to capitation has not changed how it gets paid — it has changed what business it is in.

Second, the failure was a data failure as much as a contract failure. The group could not see its own position. Under fee-for-service, revenue is legible from the charge file and every practice management system reports it. Under capitation, the number that matters — cost per member per month — is not produced by any standard billing report. A payment model that a practice cannot measure is a payment model that practice should not sign.

Third, the diagnosis codes suddenly matter for a completely different reason. Under fee-for-service, the diagnosis justifies the service. Under a risk-adjusted capitated arrangement, the documented diagnoses set the payment for the entire population for the following year. The unspecified code that was perfectly adequate on a fee-for-service claim can be materially incomplete here. That is Chapter 36, and it is also the reason Account 10-4471's diabetes line is going to come back.

Fourth: the incentive to under-deliver is real and must be managed explicitly, not denied. The honest position is that every payment structure has a distortion. Fee-for-service rewards volume, and the documented consequences of that are also substantial. Capitation rewards restraint. Neither is neutral. What separates a defensible arrangement from an indefensible one is whether the distortion is acknowledged and countered — with quality measures, with access standards, with external review, and with transparency about what is being measured and why. Chapter 36 §36.9 and §36.10 cover the machinery.

Outcome

Capitation in its 1990s form largely receded. Its successor arrangements — accountable care organizations, Medicare Advantage risk-sharing, various shared-savings and full-risk contracts — are built on the machinery that the first attempt lacked:

• Risk adjustment using hierarchical condition categories rather than age and sex.
• Quality measurement tied to payment, so that restraint cannot be the whole strategy.
• Stop-loss and reinsurance provisions.
• Data infrastructure and attribution methodologies that let a group see its own position, though the timeliness of that data remains a live complaint.
• Graduated risk — arrangements that begin with upside-only shared savings and progress to downside risk, so an organization can learn before it is exposed.

Whether the second attempt works better is an open empirical question and this book does not have an answer. What is not open is that the arrangements exist, are growing, and change what a coder's work is for. Chapter 36 is written for that reader.

The lesson

Read the reimbursement exhibit before the rest of the contract, and then ask what you would have to measure to know whether you were winning.

For a fee schedule, that measurement is easy and every system produces it. For a case rate, it is cost per episode. For capitation, it is cost per member per month, and almost no practice management system reports it without work.

And a second lesson, which is the through-line of this chapter: every contract structure moves risk from one party to another, and the party accepting risk should be the one best positioned to manage it. A primary care group can manage the risk of how often it sees its own patients. It cannot manage the risk that a new specialty practice opens across the street. A contract that assigns the second is badly drawn no matter how attractive the rate looks.

Discussion questions

1. Under capitation, a visit is a cost. Under a fee schedule, a visit is revenue. Name three specific operational decisions a practice would make differently under each, and identify which of the three is best for patients under each structure. (The answer is not the same one twice.)

2. The composite group failed partly because it could not measure cost per member per month. Design the minimum report it needed. What data would it require, and where would each field come from?

3. The case study says every payment structure has a distortion. Name the distortion in each of the five structures in §2.6, and say which one you think is most dangerous and why.

4. §2.5 distinguished who bears population risk (employer versus insurer) from how a provider is paid (§2.6). Construct the four combinations of self-funded/fully insured with fee-schedule/capitated, and say which combination puts the most parties at risk simultaneously.

5. A payer offers your practice a capitation contract at a rate 8% above your current fee-for-service collections for the same population. Write the five questions you would ask before the rate is even discussed.