ACCESS, Chronic Care, and the CMMI Savings Paradox
A New Medicare Payment Architecture Meets the Reality of Under-Managed Disease
###
SUMMARY
CMS’s ACCESS model is a bold attempt to create a new Medicare payment rail for technology-enabled chronic care, paying organizations for measurable improvement rather than visits, devices, app clicks, or care-management minutes. Its logic is strong: fee-for-service underpays continuous chronic care and over-rewards billable activity.
But ACCESS also exposes a deeper CMMI paradox. Under-managed chronic disease patients may need more drugs, PT, behavioral health, monitoring, and specialist care—not less. If ACCESS is judged too narrowly on savings, it may reward low-cost metric improvement while discouraging costly but ethically necessary care activation. That tension is the model’s central test.
###
CMS’s ACCESS model is one of the more intellectually coherent CMMI models in recent years. It is also one of the more revealing. On its surface, ACCESS is a voluntary demonstration for technology-supported chronic care in Original Medicare. It offers new payments to organizations that manage chronic conditions such as hypertension, diabetes, chronic kidney disease, atherosclerotic cardiovascular disease, chronic musculoskeletal pain, depression, and anxiety. The model is framed around digital tools, patient engagement, outcomes, and prevention.
But ACCESS is better understood as something more ambitious: an attempt to build a new Medicare payment architecture for modern chronic care. The model asks whether Medicare can pay for longitudinal, software-supported, asynchronous, remote, hybrid, and AI-enabled care without simply creating another fee-for-service volume engine. That is the fundamental design problem. If CMS creates a code for every remote-monitoring event, every digital touch, every coaching minute, every app interaction, every algorithmic review, and every clinician check-in, Medicare will reproduce the familiar pathologies of fee-for-service: more billable units, more documentation, more claims, and more spending. ACCESS tries to move the unit of payment away from activities and toward outcomes.
That is the conceptual leap. ACCESS is not primarily a digital health model. It is a new payment rail. It pays organizations recurring, condition-track payments tied to measurable clinical improvement or control. It is not paying for a visit, a device, a software license, a care-management minute, or a remote-monitoring transmission. It is paying for the production of measurable improvement in a defined chronic disease domain.
This makes ACCESS strategically important even if the model itself remains limited. It may become a prototype for how CMS thinks about AI-enabled diagnostics, connected devices, digital therapeutics, hybrid virtual care, and chronic disease management outside ordinary office visits. It is also a test of whether CMMI can create a model that is clinically modern, operationally scalable, politically durable, and fiscally defensible.
That last requirement is where the paradox begins.
CMMI models are not free-floating health policy experiments. They live under a statutory framework that expects models to reduce spending without reducing quality, or improve quality without increasing spending. That sounds reasonable in the abstract. Yet for under-managed chronic disease patients, better care often means spending more money, at least initially. A patient who is untreated, nonadherent, poorly monitored, socially unsupported, depressed, unable to access physical therapy, or not receiving indicated cardiometabolic therapy may be “cheap” only because the system is passively failing. If ACCESS identifies the failure and fixes it, Medicare spending may rise.
That is not a marginal issue. It goes to the heart of the model.
ACCESS is built on the hope that better chronic care will eventually reduce downstream costs: fewer admissions, fewer ED visits, fewer complications, fewer unnecessary specialty encounters, less avoidable progression. That may be true in some cases. But the first thing a competent chronic care organization may discover is unmet need. The patient has not filled a medication. The patient cannot afford a drug. The patient needs a higher-cost cardiometabolic therapy. The patient needs physical therapy. The patient needs behavioral health care. The patient needs labs, imaging, remote monitoring, home support, specialist input, or medication reconciliation. The morally correct intervention may be to activate care that Medicare was not previously paying for.
That is the ACCESS dilemma: the model is supposed to save money, but the patients who most need better chronic care may require more spending, not less.
What ACCESS Is Trying to Fix
Original Medicare has long had a structural problem with chronic disease management. Fee-for-service Medicare is built around discrete services: office visits, procedures, tests, imaging, drugs, devices, and time-defined care-management activities. This works reasonably well when the relevant clinical act is episodic and easily coded. It works poorly when the relevant intervention is continuous, longitudinal, remote, behavioral, digital, asynchronous, or distributed across a care team.
Chronic disease does not happen in fifteen-minute increments. Hypertension, obesity, diabetes, CKD, depression, anxiety, and chronic pain are not solved by isolated encounters. They require sustained engagement, behavior change, medication management, monitoring, coaching, escalation, de-escalation, and adaptation. Much of that work falls between visits. Much of it is not naturally captured by CPT units. Much of it is performed by teams, software, devices, and workflows rather than a single physician in a room.
CMS has tried to patch this problem with codes for chronic care management, remote physiologic monitoring, remote therapeutic monitoring, telehealth, and related services. Those codes have value, but they also illustrate the problem. Once CMS defines and prices an activity, the system learns to generate the activity. Time thresholds, documentation requirements, device transmissions, staff minutes, and monthly billing rules become the commercial object. The risk is that digital health companies and providers optimize around billable events rather than patient health.
The Shiff/Sutton JAMA Perspective is important because it states this critique from inside CMMI. It argues that technology adoption in health care is often pulled toward billing optimization, documentation, and shifting clinician work into reimbursable categories. That is not because technology is inherently low-value. It is because the payment environment rewards monetizable activity. ACCESS is intended to redirect technology toward measurable clinical improvement.
The model also tries to solve a second problem: shared-savings models do not provide a clean contracting pathway for many technology-enabled chronic care companies. ACOs and other risk-bearing entities may believe that a digital cardiometabolic, behavioral health, or MSK company can improve outcomes and reduce spending, but condition-specific savings attribution is difficult. Contracting can become bespoke, slow, expensive, and vulnerable to vendor lock-in. ACCESS creates a standardized Medicare pathway. The ACCESS organization bills Medicare directly under the model; ACO reconciliation can incorporate the spending; and clinicians can refer without negotiating a custom condition-specific risk contract.
This is why ACCESS may be a payment rail rather than a one-off model. If it works, CMS will have a repeatable template: define a condition track, define baseline measures, define outcome thresholds, define payment logic, require clinical accountability, require data submission, and let multiple care models compete.
The Basic Structure
ACCESS is a ten-year CMMI model in Original Medicare beginning in July 2026, with rolling participation. It initially includes four tracks.
The early cardio-kidney-metabolic track includes hypertension, dyslipidemia, obesity or overweight with central obesity, and prediabetes. The cardio-kidney-metabolic track includes diabetes, stage 3a or 3b chronic kidney disease, and atherosclerotic cardiovascular disease. The musculoskeletal track focuses on chronic musculoskeletal pain. The behavioral health track focuses on depression and anxiety.
The tracks are broad by design. CMS is not paying for a narrow digital therapeutic, a single device, or a single software function. It is paying organizations to manage clusters of related conditions that share intervention types, outcome measures, and care intensity. That matters. A narrow technology-specific payment would invite the old “one technology, one code” problem. A track-based payment invites organizations to build care systems.
Participating organizations must enroll in Medicare Part B as providers or suppliers, with exclusions for certain supplier types such as DMEPOS and laboratory suppliers. They must designate a physician clinical director or medical director. They must comply with applicable state licensure, HIPAA, and FDA requirements, or operate within applicable FDA enforcement discretion. Care may be delivered in person, virtually, asynchronously, or through technology-enabled means. ACCESS contemplates clinician consultations, lifestyle support, behavioral support, therapy and counseling, patient education, care coordination, medication management, diagnostic ordering and interpretation, and FDA-authorized or enforcement-discretion devices or software.
Patients voluntarily enroll, either directly or through referral. Medicare Advantage patients are not in ACCESS, although MA plans may create analogous arrangements. Some beneficiaries attempting to enroll may be randomized into a control group for evaluation. This randomized component is not a technical detail; it is central to the model’s eventual credibility. CMMI models need defensible evidence on quality and spending if they are to be expanded or made permanent.
ACCESS is designed to complement traditional care. Patients retain their ordinary Medicare rights and can continue to see any Medicare provider. ACCESS organizations must share care plans and updates with primary care or referring clinicians. Those clinicians may bill a small co-management payment for documented review and coordination work.
Outcome-Aligned Payments
The core payment mechanism is the Outcome-Aligned Payment. Participating organizations receive recurring payments for managing a qualifying track. Full payment depends on achieving measurable clinical improvement or control, generally relative to the patient’s baseline. Examples include lowering blood pressure by a clinically meaningful amount, achieving guideline-informed biomarker control, improving pain and function, or improving depression and anxiety scores.
This is not outcome payment in the abstract. It is baseline-relative and measure-specific. That distinction matters. A model that pays only for absolute control would reward organizations for enrolling easy patients. A model that ignores baseline would punish organizations for taking harder patients. ACCESS attempts to split the difference by making improvement from baseline count, while also requiring that at least one relevant measure be uncontrolled to qualify for full initial payment. The goal is to avoid paying organizations merely to enroll already-controlled patients.
The payment logic is also portfolio-based. CMS determines payment based on the share of an organization’s patients who meet outcome targets, compared with a minimum threshold. An organization can perform well overall even if some patients fail to improve. This matters operationally. Chronic care is messy; individual patients may not respond, may disengage, may face social barriers, may have side effects, or may deteriorate for reasons beyond the care organization’s control. A model that demanded success for every patient would be unworkable. A model that pays for aggregate performance is more realistic.
ACCESS is not full capitation. It is not total-cost-of-care risk. It is condition-track, outcome-contingent payment. CMS is trying to capture some of the benefits of value-based care without requiring every digital chronic care organization to become an insurer, ACO, or capitated medical group.
The Fee-for-Service Exclusion
The fee-for-service exclusion is one of the model’s most important guardrails. ACCESS is explicitly an alternative to ordinary FFS billing by the ACCESS participant for a given aligned beneficiary during the care period. It is not an add-on. The ACCESS participant and its affiliates generally may not submit ordinary Medicare FFS claims for other services furnished to that beneficiary during the ACCESS care period; they may bill only ACCESS-specific codes.
Without this exclusion, ACCESS could become another payment layer on top of existing fee-for-service claims. A company could collect outcome-aligned payments while also billing remote monitoring, chronic care management, telehealth, evaluation and management, or other FFS services for the same patient. CMS appears to understand that this would undermine the model from the start.
The exclusion is beneficiary-specific and participant-specific. An organization can treat one patient under ACCESS and another under traditional FFS. But it cannot treat the same patient under ACCESS while also billing traditional FFS for other services to that same patient during the care period. Beneficiaries remain free to see other providers, who may continue billing Medicare normally. This preserves patient choice, but it also introduces one of the model’s harder questions: when is outside care complementary, and when is it duplicative?
CMS addresses this partly through substitute-spend adjustments for certain overlapping services. For example, if an ACCESS MSK organization is managing a chronic pain condition and the patient receives duplicative outside services in a narrowly defined code set, the ACCESS payment may be reduced. But this line will be hard to police. Patients with chronic disease often need multiple forms of care. A PT evaluation, specialist visit, medication adjustment, imaging study, or behavioral health referral may be duplicative in one case and clinically essential in another.
The Co-Management Payment
The co-management payment is small but symbolically important. Primary care and referring clinicians can bill for reviewing ACCESS updates and documenting coordination activities such as medication adjustments or problem-list updates. The dollars are modest, but the policy signal is large: CMS recognizes that the traditional clinician’s attention is necessary infrastructure.
This is a practical attempt to prevent ACCESS from becoming a parallel digital health universe. Digital health companies often struggle because they sit outside the patient’s ordinary care team. They generate data, recommendations, nudges, and care plans, but the primary care doctor may not know what is happening, may not trust the vendor, may not have time to review the material, or may not be paid to act on it. ACCESS tries to build a bridge. The ACCESS organization must send updates; the clinician is paid, modestly, to review and coordinate.
Whether this will be enough is uncertain. A $30-ish co-management payment will not transform primary care capacity. But it may make a difference at the margin, and it acknowledges the right problem: technology-enabled chronic care cannot succeed if it floats above the real clinical system.
ACCESS and FDA: A Regulatory Sandbox
The FDA connection is unusually interesting. ACCESS participants must comply with applicable FDA requirements or operate within FDA enforcement discretion. CMS also links ACCESS to FDA’s TEMPO pilot, which offers a pathway for certain technology-enabled products to operate under enforcement discretion while real-world performance data are collected in the model context.
This creates a potential CMS-FDA bridge. Payment, clinical deployment, real-world evidence, and regulatory learning are usually separate. ACCESS could connect them. A digital health tool, connected device, or AI diagnostic may not need its own Medicare payment code if it becomes part of a broader outcome-producing care model. The economic unit is not the algorithm; it is the chronic care outcome.
This could be especially important for early detection and AI diagnostics. Standalone detection has a business model problem if finding a condition does not trigger a reimbursable downstream pathway. ACCESS changes that logic. Identifying uncontrolled hypertension, prediabetes, early CKD, obesity, depression, or chronic pain may become economically useful if the organization can then manage the condition under an outcome-aligned track. In that sense, ACCESS may stimulate diagnostics indirectly, not by paying for diagnosis, but by creating a payment pathway for acting on diagnosis.
The Accepted Applicants: Revealed Market Interest
The accepted-applicant list is useful less as a set of tea leaves than as revealed-market evidence. CMS says more than 150 organizations were accepted for the ACCESS launch, and the list includes a broad range of entities: digital-first chronic care companies, behavioral health platforms, MSK companies, cardiometabolic firms, remote-monitoring organizations, provider-affiliated groups, nutrition and weight-management brands, virtual care companies, nephrology practices, community organizations, and multi-track platforms.
This matters because it suggests ACCESS created a commercially legible pathway. CMS itself notes that many accepted organizations have not previously served Medicare beneficiaries. That is a striking statement. ACCESS is not merely giving existing Medicare providers a new code. It is inviting nontraditional chronic care organizations into Original Medicare, but under provider/supplier enrollment, physician oversight, data reporting, privacy, licensure, and quality requirements.
The applicant mix also reveals where the market sees opportunity. Cardiometabolic care appears dominant, which makes sense. Hypertension, obesity, dyslipidemia, prediabetes, diabetes, CKD, and ASCVD are enormous Medicare problems. They have measurable biomarkers. They map naturally to remote monitoring, medication management, nutrition support, coaching, adherence programs, and connected devices. Behavioral health is also heavily represented, reflecting the scalability of virtual behavioral health and the availability of validated measures such as PHQ-9 and GAD-7. MSK is present, though perhaps with a more complex savings story because it overlaps with PT, imaging, orthopedics, pain management, and procedural care.
The broad applicant list is not proof that ACCESS will work. CMS cautions that inclusion is not endorsement and does not guarantee final participation. Medicare enrollment, participation agreements, and final approval remain required. But the applicant list does show that ACCESS has already created a credible market signal.
Lessons from the Oncology Care Model
The Oncology Care Model is an important comparison because it shows how hard it is for CMMI models to produce net Medicare savings even when they produce measurable gross reductions in spending.
OCM ran from 2016 to 2022 and was CMS’s first major oncology-focused alternative payment model. It paid participating oncology practices monthly enhanced oncology services payments, while also allowing performance-based payments for reducing spending and meeting quality benchmarks. A 2026 JAMA evaluation found that OCM was associated with a modest reduction in total episode payments: about $616 per six-month episode, or 2.1 percent of baseline episode spending. Savings grew over time and were larger in later performance periods. But once monthly enhanced oncology services payments and performance-based payments were included, the model produced an estimated net loss to Medicare of $639 million over six years. Most quality and utilization measures did not significantly improve.
OCM is not ACCESS, but it offers several lessons.
First, gross savings are not net savings. A care model can reduce claims spending and still lose money after model payments are included. ACCESS will face the same arithmetic. Outcome-aligned payments are not free. If downstream savings are modest, delayed, or offset by increased appropriate care, net savings may be elusive.
Second, transformation takes time. OCM savings improved in later years. This matters for ACCESS because chronic disease prevention may take even longer. A model that improves blood pressure, weight, depression, or pain may not produce large utilization savings inside a short measurement window.
Third, quality measures may fail to capture clinically meaningful change. OCM may have changed care patterns without producing broad measurable quality differences. ACCESS will rely on more direct clinical measures, which is an advantage, but it still faces the challenge of whether measured improvement corresponds to broader patient benefit and lower spending.
Fourth, voluntary models select participants. OCM participants were not random oncology practices. ACCESS participants will likewise be self-selected organizations that believe they can win under the model. That may be good for early performance but limits generalizability.
The broader OCM lesson is sobering: even sophisticated, well-intentioned models can be clinically plausible, operationally serious, and still fail the net-savings test.
Liao’s Critique: Complexity, Distortion, and Realism
Joshua Liao’s Health Affairs critique is useful because it avoids both cheerleading and cynicism. He acknowledges the core problem ACCESS is trying to solve: chronic care is poorly matched to traditional FFS reimbursement. He also emphasizes that payment models predictably create organizational responses. FFS rewards volume. Capitation rewards restraint. Documentation-based quality measurement rewards documentation optimization. Risk adjustment rewards coding. Outcome-based models reward selection, measure optimization, and denominator management.
The right question, in Liao’s framing, is not whether ACCESS will generate distortions. It will. The question is whether the model can generate enough clinical improvement to justify those distortions.
That is the proper standard. No payment system is distortion-free. The status quo is not neutral. Fragmented, encounter-based FFS chronic care is itself a policy choice, and a costly one. ACCESS should not be rejected merely because it will produce strategic behavior. But it should be evaluated with clear eyes.
Liao also points to a critical distinction between scalable chronic care and intensive multimorbidity management. ACCESS may work best for a middle population: patients with measurable, improvable chronic conditions who can engage with a technology-supported model and whose care can be standardized to some degree. It may be less suitable for the sickest, most socially complex, least connected, least adherent, or most multimorbid beneficiaries. That distinction matters because the highest-need patients may also be the least profitable or least measurable.
This is where the ACCESS optimism meets the ACCESS paradox.
The Savings Paradox: Under-Managed Patients May Need More Spending
The CMMI statute expects savings or budget neutrality with quality improvement. But under-managed chronic disease is often not a clean savings opportunity. It is a backlog of unmet need.
A patient with uncontrolled diabetes may need medication intensification, CGM, nutrition counseling, renal testing, eye exams, foot care, and closer follow-up. A patient with CKD may need appropriate labs, SGLT2 therapy if indicated, nephrology input, medication review, and blood pressure optimization. A patient with chronic pain may need PT, behavioral support, functional assessment, careful medication changes, and perhaps imaging or specialist evaluation. A patient with depression may need therapy, medication management, safety assessment, and repeated follow-up. A patient with cardiovascular disease may need high-intensity lipid therapy, blood pressure medication changes, adherence support, and labs.
Some of this care may reduce downstream spending. Some may improve quality of life without saving money. Some may increase spending indefinitely because the patient is finally receiving appropriate care.
This is not failure. It is medicine.
Consider the nonadherence case. ACCESS may identify that a patient is not taking a $5,000 medication that is clinically indicated. Perhaps the patient cannot afford it, does not understand it, experienced side effects, or never completed a refill. The best clinical intervention may be to fix the adherence problem. But if adherence improves, Medicare spending rises. Pharmacy costs rise. Monitoring may rise. Follow-up may rise. The model may have improved care while worsening measured spending.
The ACCESS organization therefore faces an uncomfortable incentive. It is paid to improve outcomes at a fixed payment, and the broader model is expected to reduce Medicare costs. It is not necessarily rewarded for discovering and activating expensive care. The company may be strongly motivated to find low-cost, measurable improvements: nudges, reminders, coaching, dashboard review, medication reconciliation, lifestyle support, app engagement, protocolized outreach. Those can be valuable. But the harder ethical question is whether the organization is equally motivated to trigger expensive downstream care when that is what the patient needs.
This is the central paradox: ACCESS is designed to find under-managed patients, but truly fixing under-management may increase Medicare spending.
Medicare Advantage as a Mirror Image
Medicare Advantage provides a useful comparison. MA plans are famously incentivized to find and document diagnoses because risk-adjusted payment rises when patients appear sicker. MA has built a large infrastructure around coding intensity: chart reviews, health risk assessments, home visits, suspect-condition analytics, and documentation capture. The plan is rewarded when illness is found and coded.
But MA is not symmetrically rewarded for doing everything clinically appropriate in response. A plan may document diabetes with complications, vascular disease, depression, COPD, CKD, frailty, or pressure ulcers. The diagnosis increases payment. But the payment system does not automatically require or reward optimal treatment follow-through: the right drugs, PT, behavioral health, wound care, nutrition, specialist care, home support, or adherence work. Star ratings and quality measures provide some counterweight, but the coding incentive is more direct and more comprehensive than the treatment obligation.
Thus MA’s pathology can be summarized as: find the illness, code the illness, get paid for the illness; treatment follow-through is only partially and indirectly incentivized.
ACCESS is different but related. ACCESS does not monetize diagnosis capture in the same way. It monetizes chronic care engagement and measurable outcome improvement. The ACCESS organization is encouraged to study the patient, monitor the patient, engage the patient, and improve track-specific measures. But it is not necessarily encouraged to raise spending, even if raising spending is the best ethical care.
Thus ACCESS’s pathology may be: study the patient, improve measurable outcomes, avoid triggering costly care that makes the model look financially unattractive.
The contrast is illuminating.
Medicare Advantage risks diagnosis capture without proportional care. ACCESS risks care-management capture without full care activation.
MA is rewarded for knowing that the patient is sick. ACCESS is rewarded for making selected measurable problems better at reasonable cost. In both cases, the intermediary is rewarded for information and management activity, but the patient’s real interest is not merely being known or monitored. The patient’s interest is receiving the right care.
The common flaw is the gap between information and obligation. In MA, the plan may know more because knowing more raises revenue. In ACCESS, the organization may know more because knowing more allows outcome management. But the act of knowing does not guarantee that the system will spend what is necessary to treat the patient properly.
Selection and the Ideal ACCESS Patient
Every payment model creates an ideal customer. For ACCESS, the ideal patient is likely to be sick enough to qualify, uncontrolled enough to improve, reachable enough to engage, stable enough to manage remotely, and not so complex that improvement requires large spending outside the ACCESS payment.
That is a narrow sweet spot.
A patient already controlled may not qualify for full initial payment. A patient far from target may be difficult to improve. A patient with severe social barriers may not engage. A patient needing extensive specialist care may generate costly outside utilization. A patient requiring expensive drugs may improve clinically but worsen spending. A patient with multiple overlapping conditions may require more individualized care than a scalable protocol can provide.
CMS tries to mitigate selection through baseline-relative targets, uncontrolled-measure requirements, limits on participant-initiated disenrollment, risk-adjusted public reporting, and rural adjustments. These are important. But no guardrail eliminates selection pressure. Organizations will still learn which patients are likely to succeed under the payment rules.
The most attractive ACCESS patient may not be the neediest patient. It may be the improvable patient.
That does not invalidate the model. Many improvable patients need help. But it should shape expectations. ACCESS may be strongest as a scalable chronic-care improvement model, not as a solution for the most complex Medicare beneficiaries.
Outcome Measurement: Strength and Vulnerability
ACCESS’s greatest strength is that it pays for outcomes rather than activities. Its greatest vulnerability is that measured outcomes can become the object of optimization.
Blood pressure, HbA1c, lipids, weight, pain PROMs, PHQ-9, and GAD-7 are clinically meaningful. They are far better payment targets than app clicks or minutes. But they are still measures. Measures can be selected around, documented around, repeated strategically, and improved narrowly. An organization may focus on the outcome threshold rather than the patient’s broader health. It may prioritize patients most likely to cross the line. It may optimize data capture. It may focus on short-term score movement rather than durable improvement.
Behavioral health illustrates the problem. PHQ-9 and GAD-7 are validated and useful, but mood and anxiety scores are noisy. They fluctuate with life events, comorbidities, loneliness, cognition, substance use, sleep, pain, and social stress. A digital behavioral health model may improve scores for some patients while missing deeper needs. Public risk-adjusted outcomes may help, but risk adjustment for social and behavioral complexity is notoriously difficult.
MSK has a similar problem. Pain intensity, interference, and function are important, but chronic pain is heterogeneous. A patient may need PT, behavioral support, medication changes, imaging, injections, surgery, weight loss, workplace modification, or social support. A model focused on PROM improvement may be valuable, but it may also favor patients whose pain is most amenable to low-cost digital or hybrid intervention.
Cardiometabolic tracks have more objective biomarkers, but they raise the overtreatment and under-treatment questions. Aggressive medication intensification can improve measures but create polypharmacy or side effects. Conversely, avoidance of expensive medication may preserve spending but fail the patient.
The Ethical Problem of “Savings”
The language of savings is too crude for chronic disease. In a well-managed system, savings may come from reducing waste, avoiding complications, substituting lower-cost care, preventing admissions, and stopping unnecessary services. But in an under-managed system, better care may increase utilization by revealing needs that had been neglected.
A useful distinction is between waste reduction and need activation.
Waste reduction saves money by eliminating low-value or duplicative care. Need activation increases spending by delivering high-value care that was missing. Chronic disease management requires both. A good model should reduce unnecessary imaging, avoidable ED visits, duplicative therapy, and low-value medications. But it should also increase appropriate drugs, labs, PT, behavioral health, monitoring, nutrition support, and specialist care when needed.
If CMS evaluates ACCESS only through net spending, the model may be penalized for need activation. If CMS evaluates only outcome measures, the model may ignore total cost. The hard policy task is to distinguish bad spending growth from good spending growth.
For example, if ACCESS increases SGLT2 use in appropriate CKD or diabetes patients, Part D spending may rise. If it increases PT in patients with fall risk or chronic pain, Part B spending may rise. If it increases behavioral health care for depressed patients, outpatient spending may rise. If it improves adherence to expensive medications, pharmacy spending may rise. These increases may be exactly what good care requires.
The question should not be, “Did spending rise?” It should be, “Did spending rise because the model induced low-value utilization, or because it corrected underuse of high-value care?”
That is harder to measure. But it is the question that matters.
A Sidebar: Two Mirror-Image Incentive Failures
Medicare Advantage and ACCESS reveal two different ways payment can separate knowledge from care.
In Medicare Advantage, the plan is incentivized to find diagnoses because diagnoses increase risk-adjusted payment. The danger is diagnosis capture without proportional treatment. The plan knows more about the patient, codes more about the patient, and gets paid more for the patient, but the system does not automatically require that the patient receive the right drug, PT, behavioral health, wound care, specialist follow-up, or home support.
In ACCESS, the organization is incentivized to find manageable problems and improve measurable outcomes at fixed payment. The danger is care-management capture without full care activation. The organization may know that the patient is undertreated, nonadherent, depressed, unstable, or in need of costly therapy, but may have weak incentives to trigger spending that makes the model less favorable.
MA monetizes illness recognition. ACCESS monetizes measurable improvement. MA may over-code without enough care. ACCESS may over-monitor or under-escalate when care is expensive.
Both models reward knowing more. Neither fully solves the ethical question: once you know, what are you obligated to do?
Political and Administrative Durability
ACCESS has a politically interesting profile. It is voluntary. It is outcome-based. It emphasizes prevention, chronic disease, patient choice, rural access, technology, and private-sector innovation. It invites digital health companies into Medicare, but under provider/supplier enrollment and compliance obligations. It is not a simple benefit expansion, and it is not a pure austerity model. This gives it some cross-administration durability.
It can be described in Republican-friendly language: innovation, competition, technology, private-sector solutions, outcomes, lower costs, and less fee-for-service bureaucracy. It can also be described in Democratic-friendly language: access, chronic care, behavioral health, prevention, patient support, rural care, and modernization of Medicare. That does not guarantee survival, but it gives ACCESS more political flexibility than many models.
The Shiff/Sutton article also signals that ACCESS is not a stray technical experiment. It reflects a broader CMMI payment philosophy: do not pay for every activity, do not create endless technology-specific codes, do not rely entirely on bespoke shared-savings contracts, but instead create standardized payment structures around outcomes.
The risk is that ACCESS will take years to evaluate. A ten-year model beginning in 2026 will cross administrations. Early operational problems, fraud concerns, weak savings, poor engagement, or visible patient-selection stories could damage it before mature results appear. Conversely, a strong early applicant pool, supportive clinicians, private-payer alignment, and credible outcome data could help sustain it.
Likely Winners and Losers
The likely winners are organizations that already combine clinical operations, patient engagement, software, measurement infrastructure, and disease-specific workflows. Pure software companies may struggle unless they partner with or become Medicare-enrolled clinical entities. Traditional provider groups may struggle if they lack digital infrastructure. The sweet spot is hybrid: clinically accountable, technologically scalable, operationally disciplined.
Likely winners include cardiometabolic management platforms, digital-first or hybrid behavioral health firms, MSK companies with clinical staffing, nutrition and obesity-management organizations, remote-monitoring companies with care teams, nephrology or cardiology-adjacent groups, ACO-adjacent service organizations, and provider groups that can wrap technology around chronic care.
The model may also benefit AI diagnostic and device companies, but indirectly. The diagnostic or device is not the payment unit. It must be part of an outcome-production system. This is a major strategic point for digital health companies: the question is not “what code pays for our product?” The question is “what measurable outcome can our product help a Medicare-enrolled care organization achieve?”
Likely losers are companies built around billable activity rather than outcomes, companies without clinical accountability, companies that cannot handle Medicare compliance, and organizations whose economics depend on high-touch labor that exceeds ACCESS payment levels. Also at risk are firms whose patients require expensive downstream care that the model does not adequately reward.
What CMS Should Watch
CMS should watch at least six issues.
First, selection. Are ACCESS organizations enrolling the patients most likely to benefit, or the patients most likely to be profitable under the model?
Second, under-escalation. When ACCESS organizations identify unmet clinical need, do they trigger appropriate care even when it increases spending?
Third, outside spending. Are organizations shifting costs to unaffiliated providers, or are outside services clinically appropriate complements?
Fourth, primary care integration. Are PCPs and referring clinicians meaningfully engaged, or are they receiving electronic clutter and small payments without real coordination?
Fifth, durability of outcomes. Are improvements sustained, or are they short-term threshold effects?
Sixth, equity. Are rural, low-income, cognitively impaired, socially complex, digitally limited, and multimorbid patients being served, or are they being left behind?
The under-escalation issue deserves special attention because it may be less visible than overuse. Claims data can show increased imaging, PT, drug use, or specialist visits. It is harder to see the expensive care that should have happened but did not. A model that avoids necessary care may look financially disciplined. That is why outcome measures, patient safety monitoring, audits, and clinical review will matter.
Bottom Line
ACCESS is CMS’s attempt to build a new Medicare payment rail for technology-enabled chronic care. Its strongest idea is simple and powerful: pay for measurable improvement or control, not for clicks, devices, minutes, visits, or billing activity. Its most important guardrails include the fee-for-service exclusion, physician oversight, Medicare enrollment, FDA/HIPAA compliance, care coordination requirements, public outcome reporting, substitute-spend adjustments, and randomized evaluation.
The model is sophisticated. It addresses a real problem. It may create a new category of Medicare-participating chronic care organizations. It may become a reference architecture for private payers. It may help CMS and FDA learn how to regulate and pay for AI-enabled and software-supported care. It may shift digital health strategy away from code capture and toward outcome production.
But ACCESS also exposes the deep contradiction in CMMI’s statutory savings logic. Chronic disease patients are often under-managed. Under-management may look cheap until the system starts doing its job. Better care may require more drugs, more PT, more behavioral health, more labs, more monitoring, more specialist input, and more adherence support. Some of that spending may eventually prevent complications. Some may improve life without saving money. Some may simply be the ethically correct cost of treating Medicare beneficiaries properly.
This is the fundamental ACCESS paradox: the model is designed to find and manage chronic disease, but the act of finding and properly managing chronic disease may increase spending.
Medicare Advantage offers the mirror image. MA is rewarded for finding illness diagnoses, but not always equivalently rewarded for doing everything needed to treat them. ACCESS is rewarded for studying patients and improving measurable outcomes, but may be financially uncomfortable when the best care requires expensive activation. MA risks coding without care. ACCESS risks management without escalation.
The patient does not need merely to be coded, monitored, engaged, or measured. The patient needs the right care.
That should be the standard by which ACCESS is judged. Not whether it is digitally elegant. Not whether it creates a new market. Not even whether it saves money in a narrow window. The real question is whether ACCESS can improve chronic care while resisting the predictable pressure to select easy patients, optimize measures, avoid costly escalation, and confuse low spending with good care.
If ACCESS succeeds, it could be one of the most important Medicare payment innovations of the decade. If it fails, its failure will still be instructive. It will show that outcome-based payment is not enough unless the model can distinguish waste from unmet need, savings from underuse, and measurable improvement from ethical care.
