Every Practice You Acquire Is Already Doing the Work

If you are consolidating neurology practices right now, you are early to something the rest of healthcare has not fully noticed yet. You are also buying a very specific asset, and it is worth being precise about what it is.

You are not buying visit volume. Visit volume is capped by the scarcest input in the entire specialty — a neurologist's hours — and it does not get cheaper with scale. What you are actually buying is panels: thousands of patients with progressive, multi-year, comorbidity-dense neurological disease, each of whom generates a continuous stream of coordination work between visits.

That coordination work is real. Your practices are doing it today. In almost every practice we have looked at, they are doing it for free.

You Bought Panels. Panels Have Two Revenue Lines.

There are two ways a neurology practice earns. Most MSO operating models are built almost entirely around the first one, and treat the second as a rounding error or a someday project. In neurology specifically, that ordering is backwards.

CMS has been paying for line two for years. Four distinct programs cover it — Chronic Care Management for patients with two or more chronic conditions, Remote Patient Monitoring for device-based physiologic data, Principal Care Management for a single high-complexity condition, and Remote Therapeutic Monitoring for therapy adherence and response data. Each one has published codes, published time thresholds and published documentation requirements. None of them is a pilot, and under Medicare none of them requires a contract to be negotiated first — though coverage across commercial and Medicare Advantage plans does vary, so it is worth testing against your own payer mix.

What each of them does require is proof: documented consent, a documented care plan, and documented clinical staff time attributable to a specific program in a specific month. That is the whole game. The clinical work is not the constraint. The evidence of the clinical work is the constraint.

Dementia Is the Best Care Management Population in American Medicine

Take any list of the attributes that make a population ideal for chronic care management, and the dementia panel scores near the top of every one.

• Eligibility is nearly automatic. CCM requires two or more chronic conditions expected to last at least twelve months. A dementia diagnosis is one; the hypertension, diabetes, depression, hyperlipidemia or atrial fibrillation sitting alongside it in the chart supply the rest. Very few patients on a memory clinic panel fail the test.
• The coordination burden is genuinely high. Medication reconciliation, caregiver instruction, fall risk, behavioral changes, transitions after every hospitalization, driving and safety conversations, advance planning. This is not manufactured activity to justify a code.
• The horizon is long. Neurodegenerative disease does not resolve. A patient enrolled today is very often still enrolled thirty-six months from now — which is precisely what makes the revenue recurring rather than episodic.
• The clinical case and the billing case point the same direction. These programs reimburse exactly the proactive coordination that reduces crises, ER visits and avoidable admissions in this population. You are not gaming anything. You are being paid for the thing you would want done anyway.

Sources listed at the end of this article. We have used only figures we could trace to a primary source, and labelled the rest as what it is: modelling and judgment.

So if the population is this good and the programs have existed this long, the obvious question is the interesting one.

And the Worst Fit for Care Management Software Built for Anyone Else

Here is the paradox at the centre of this entire opportunity, and it is the reason the revenue is still sitting there for you to take rather than already competed away.

The patients who need care management most are the least able to participate in it.

Every mainstream care management platform in the market makes one foundational assumption: that the patient is the participant. The patient consents. The patient answers the outreach call. The patient downloads the app, pairs the blood pressure cuff, reports the symptom, remembers the appointment. Strip that assumption out and most of these systems have nothing left to stand on.

A patient with moderate Alzheimer's disease cannot reliably do any of it. In neurology, care management is not a two-way relationship between practice and patient. It is a three-way relationship between the practice, the patient and the caregiver — and the caregiver is frequently the daughter, three states away, who works full time and is the only person who actually knows which medications were taken this week.

Software that treats the caregiver as an emergency contact instead of a primary user does not merely perform worse in this population. It fails structurally, in five specific and predictable places.

What “Tailor-Made for Dementia” Has to Mean

Every vendor in this market will tell you they support neurology. Ask them what they changed. In most cases the answer is a specialty dropdown and a different set of ICD-10 codes on the same primary-care chassis.

We built the model the other way around — starting from the dementia patient and the person who actually manages their life, and working outward. Four things follow from that starting point, and they are not cosmetic.

What a month actually looks like

The reason this matters commercially is volume. A dementia household is not a low-touch relationship — it is one of the highest-touch relationships in ambulatory medicine, and almost all of it happens between visits. Here is a single patient in a single ordinary month.

That is the whole thesis in one graphic. The base threshold was cleared by the second entry, and three times over by month end. Whether any of it is worth anything depends entirely on whether the caregiver was a first-class user of a system that was counting.

The AI Layer: Three Users, Not One

The volume above cuts both ways. Seven meaningful interactions per patient per month is excellent revenue and an impossible workload — multiply it by a panel of three hundred and you have simply moved the bottleneck from the neurologist to the care manager. Enrollment depth stalls not because patients are unwilling but because your staff runs out of hours in the third month.

This is where the AI layer earns its place, and it has to serve all three corners of the triad — not just the one that shows well in a demo.

One point of candour, because it matters for how you model this: an AI interaction is not billable clinical staff time, and we will never tell you otherwise. What the AI layer does is more valuable and less exciting than that. It absorbs the volume that would otherwise consume your staff, it captures and structures every interaction as it happens so the qualifying human minutes are documented rather than remembered, and it lets one care manager hold a materially larger panel without the quality falling over. In the model further down, that is the difference between enrolling twenty-five patients per physician and enrolling fifty.

Underneath all of it sits the same protocol engine that runs our infusion and oncology work — because a dementia panel is rarely only a dementia panel any more. Which brings us to the second, larger bet.

The Second Bet: Anti-Amyloid Readiness Is an MSO-Scale Problem

Two disease-modifying anti-amyloid therapies are now FDA-approved and commercially available. Real-world uptake has been slow, and the comfortable explanation — that neurologists are clinically skeptical — does not survive contact with the practices themselves.

The barrier is operational. Putting one patient on therapy requires eligibility confirmation and amyloid testing, APOE genotyping, a baseline MRI, prior authorization against payer-specific criteria, an infusion schedule that holds for months, a surveillance MRI schedule for ARIA that is not optional, symptom vigilance between infusions, and a caregiver who understands what to watch for. Miss one gate and the protocol is broken. We wrote about this at length in The Real Barrier to Anti-Amyloid Adoption.

A solo practice cannot build that capability. A ten-practice platform can build it once and deploy it ten times — and that asymmetry is, in plain terms, the reason MSOs exist. This is one of the few places where the holding company is genuinely the right unit of solution rather than an overhead layer.

The other burden nobody demos: getting paid for it

Everything above is about the clinical protocol. There is a second protocol running alongside it that gets far less attention and destroys far more money, and it lands on exactly the same people. These are five- and six-figure claims. When one goes wrong, it does not bounce back as a polite question — it becomes a denial, an appeal, a write-off, and a very awkward line in a monthly review.

Consider everything that has to be simultaneously true before a single anti-amyloid dose becomes a clean claim.

This is the part of the platform that is least interesting to demo and most valuable to own. The system holds the state, not the staff. Authorisation has an expiry and a next action weeks ahead of it. Units derive from the dose that was actually administered rather than from what someone remembers. The imaging gate is linked to the dose it gates, so the evidence assembles itself. The site-of-care code set follows the site. And the exception — the one dose out of sixty where something is missing — surfaces before the appointment rather than in a denial letter eleven weeks later.

And note where this class is heading, because it changes the shape of the problem rather than the size of it. With subcutaneous lecanemab dosing now approved, part of anti-amyloid therapy is moving out of your chair altogether. None of the surrounding discipline moves with it: the baseline and surveillance MRI gates, the ARIA vigilance, the coverage criteria and the caregiver education are all exactly where they were. What changes is that more of it now happens in a house you cannot see into. A network whose anti-amyloid capability is really just an infusion suite gets weaker as this shifts; a network whose capability is a monitoring and protocol standard gets stronger, because the part it owns is the part that just became harder.

There is a strategic dimension here too. Therapy manufacturers have a direct commercial interest in removing the operational friction that keeps eligible patients off treatment — that is exactly what hub and practice-enablement programs exist to do. A neurology platform that can demonstrate reliable protocol execution across dozens of sites is a materially more interesting partner to them than any single practice can ever be.

One Infusion Standard, Every Chair, Every Site

And anti-amyloid is only the newest arrival. Look at what is already flowing through the infusion suites in a neurology portfolio: MS biologics, immunoglobulin for CIDP and myasthenia gravis, CGRP therapies for chronic migraine, ALS infusions, and now anti-amyloid on top. Your infusion footprint is quietly the most capital-intensive, most protocol-dependent and most margin-sensitive asset you own — chairs, pharmacy, nurses and a schedule that has to hold.

Here is the operational reality in almost every multi-site group we have looked at. Ask four sites how they run the same therapy and you get four answers, all of them defensible, none of them written down in the same place:

• Different pre-medication and pre-check practice, because each site's protocol reflects whoever set it up and what they were taught.
• Different reaction response — observation periods, rate adjustments and escalation thresholds that vary by nurse rather than by policy.
• Different chair-time assumptions, so the same therapy is booked for ninety minutes at one site and three hours at another, and nobody can compare utilisation.
• Different prior authorisation playbooks, re-learned per payer, per site, per patient.
• Different documentation, which means portfolio-level reporting is an exercise in reconciling four spreadsheets rather than reading one view.

Every one of those is survivable inside a single practice. None of them is survivable as an acquisition strategy, because the divergence compounds with every site you add — and it is exactly the thing a holding company is supposed to eliminate.

The rails are constant. Only the payload changes.

The good news is that infusion standardisation is far more tractable than it looks, because the therapies differ enormously while the operational discipline barely differs at all. Six rails carry every infusion your network delivers. What changes between therapies is what gets loaded onto them.

That is what makes this a platform problem rather than a practice problem. If the rails live in a system, a new therapy is a configuration — a protocol pack loaded onto rails that already exist, deployed to every site at once. If the rails live in people, a new therapy is eleven separate implementation projects, run at eleven different speeds, with eleven different results.

This is the same engine described earlier, pointed at a different problem. Pre-checks as enforced gates, surveillance that continues between visits, authorisation and documentation riding along with the protocol — and one portfolio-level view of which patients at which sites are on protocol, off protocol, or about to be. We wrote about why fragmenting that across six vendors becomes a safety problem rather than merely an inconvenience in Six Vendors, One Patient.

The Integration Clock: Where This Actually Shows Up

Care management revenue is not really a clinical initiative in an MSO. It is an integration capability. The question is not whether CCM works — it demonstrably does — but how many days after close it starts working, and whether the answer is the same for practice eleven as it was for practice one.

That last row is the one worth dwelling on. A capability that has to be rebuilt at every site is an operating expense. A capability that deploys by configuration is a reason the platform is worth more than the sum of the practices in it — which is, presumably, the entire premise you are underwriting.

“Fine — but every practice we buy has a different EHR.”

This is the first question we get from anyone who has actually run an integration, and it deserves a straight answer rather than the word “interoperability.”

We do not replace the EHR. The platform runs alongside it and the system of record stays the system of record — which is the only version of this that survives contact with a physician who has spent four years learning their chart. What we integrate is a deliberately narrow surface, because the surface that matters for care management revenue is genuinely small: get the panel in, and get the monthly documentation back out to where the practice already works.

Concretely, at the neurology group described above we are integrated with Epic, exporting each patient's monthly CCM summary directly into the EHR. The documentation that substantiates the claim lands in the chart automatically — nobody re-keys it, nobody rebuilds it at month end, and the record a payer would eventually ask to see is sitting where an auditor would expect to find it. That is a narrow integration and we would rather describe it accurately than dress it up as something larger.

The narrowness is the point. A migration is a quarter-long project with a change-management problem attached; a summary export is not. It is the reason a new site can be counted in weeks rather than in budget cycles — and it is also the honest limit of the claim. Each EHR is its own piece of work, we scope it per site, and in any conversation with you we will be specific about what we have built already and what we would be building for the first time.

The Portfolio Math, With the Assumptions Showing

We are not going to hand you a number without showing how it was built. Here is the model, the inputs, and where it is fragile.

Now the honest part, because it is the part that determines whether any of the above happens: the fragile input is B. The rate is published and the physician count is on your cap table. Enrollment depth is where every one of these programs lives or dies — and enrollment depth in a dementia panel is precisely what breaks when the consent workflow cannot handle a surrogate and the outreach is addressed to a patient who cannot respond.

Which is the whole argument of this piece, expressed as arithmetic. The dementia-native design is not a nicer user experience. It is the variable that decides whether column one or column three is the one you actually get.

Battle-Tested — and the Honest Size of It

None of this is a whiteboard exercise for us. Our care management model was built for dementia and hardened in a real neurology practice, in production, with real patients and real caregivers — not adapted from a primary-care product after the fact.

The Questions We Would Ask If We Were You

You will be pitched by several vendors on this, including us. Rather than ask you to take our word for anything, here is the list we would use. It is deliberately unkind to weak answers, ours included.

The Work Is Already Being Done

Somewhere in your portfolio this week, a nurse will spend twenty-five minutes on the phone with a daughter in another state, walking her through why her father is suddenly agitated in the evenings, what to try tonight, and what would warrant a call back. That conversation is skilled clinical work. It probably prevents an ER visit. It is exactly what CMS designed these programs to pay for.

Whether it becomes revenue depends entirely on whether the system your practices use was built for the person who was actually on that call.

Sources & Further Reading

• Alzheimer's Association. 2026 Alzheimer's Disease Facts and Figures — an estimated 7.4 million Americans age 65 and older are living with Alzheimer's in 2026, projected to reach 13.8 million by 2060; the great majority of day-to-day care is provided by unpaid family caregivers, which is the structural fact underneath the caregiver-first argument in this article.
• Centers for Medicare & Medicaid Services. Chronic Care Management Services and Care Management Services booklets (Medicare Learning Network) — source for the four program families referenced here (CCM, PCM, RPM, RTM), the 20-minute base CCM clinical staff time threshold, and the requirements for documented patient consent, a comprehensive care plan and 24/7 access to the care team. National average payment figures referenced in the model are drawn from the CY2026 Medicare Physician Fee Schedule; base CCM (99490) is approximately $66 nationally in 2026, non-facility.
• Prescribing information for lecanemab (Leqembi) and donanemab (Kisunla) — source for infusion cadence, MRI surveillance for amyloid-related imaging abnormalities (ARIA), and pre-treatment evaluation including APOE genotyping. These remain the two FDA-approved anti-amyloid therapies as of this writing; lecanemab has since added intravenous maintenance dosing every four weeks and, in 2026, a subcutaneous option. Characterised here at a summary level; this article is not clinical guidance.
• FDA-approved prescribing information for the neurology infusion therapies referenced in the standardisation matrix — including ocrelizumab and natalizumab (multiple sclerosis), immunoglobulin products used in CIDP and myasthenia gravis, eptinezumab (migraine, 30-minute infusion every three months) and intravenous edaravone (ALS, 60-minute infusions given daily for 14 days in the initial cycle and 10 of every 14 days thereafter — the chair-load contrast drawn in the matrix). Dosing cadence, screening requirements and monitoring are characterised at a deliberately summary level; the matrix is illustrative and simplified, and is not clinical guidance or a substitute for the label.
• American Medical Association. 2025 AMA Prior Authorization Physician Survey — context for the administrative burden that accompanies high-cost specialty therapies, including reported care delays.
• MemberCare internal figures — the $235 per-enrolled-patient-per-month blended CCM/RPM rate used in the model, and the protocol compliance, missed-checkpoint and patient-count figures reported by the 20-physician neurology group referenced above. These are our own numbers and the group's, not independently audited, and we present them as such.

A note on sourcing: figures widely repeated in this market — adoption percentages, revenue-per-patient benchmarks, staffing multipliers — frequently cannot be traced to a primary source. We have omitted them rather than repeat them, and where we have used our own modelling we have shown the inputs so you can replace them with yours. The portfolio table is an illustrative model, not a forecast, and contains no patient data.