When Cost Control Meets Clinical Reality: The Design Problem in the CMS GUARD Model
A design analysis of pricing reform, protected classes, and treatment stability for older adults living with HIV in Medicare Part D
Medicare drug pricing reform is entering a new phase through the Centers for Medicare & Medicaid Services (CMS) proposed GUARD Model. The model introduces international reference pricing into Medicare Part D, with the stated goal of reducing drug costs and improving affordability. CMS has framed the model as a way to lower out-of-pocket costs and support better adherence, an objective that is both clear and necessary.
The GUARD Model is a design intervention rather than a marginal adjustment. It alters how price signals are set, how plans respond, and how access is structured. The analytic question is not whether costs should be reduced, but whether the design choices embedded in the model account for the conditions under which care is actually delivered, particularly for populations whose treatment depends on long-term stability.
The Design Problem
The GUARD Model is structured to align U.S. drug prices with those in other countries through a system of benchmarks and rebates. This approach assumes that price can be adjusted without materially affecting access, so long as coverage remains formally intact.
At the same time, Medicare maintains a separate design feature: the six protected classes policy. This policy requires Part D plans to cover substantially all drugs in specific therapeutic categories, including HIV, based on the premise that treatments within these classes are not interchangeable.
These two design logics now operate in parallel: one changes incentives across plans and manufacturers, while the other assumes access to a full range of therapies must be preserved. The model does not explicitly specify how these design features will interact in practice and proceeds as if both can operate without affecting the other.
Design Fictions
Two design fictions are evident in this structure.
The first is a form of neutrality. The model assumes that price adjustments can be applied uniformly across therapeutic areas without producing uneven effects. This holds only if medications within those areas are substitutable in practice. For protected classes, that assumption does not hold.
The second is a form of stability. The model treats access as stable so long as coverage categories remain intact. In practice, access is mediated through formularies, utilization management, and plan behavior. Stability is not defined by formal inclusion but by whether individuals can continue to obtain the specific medications that work for them without disruption.
These fictions do not arise from intent. They persist because systems continue to function until design constraints force visible failure.
Where HIV Care Does Not Conform to the Model
HIV treatment does not operate under conditions of interchangeability. Medication regimens are individualized over time, with clinical decisions reflecting prior treatment history, resistance patterns, comorbid conditions, and patient-specific tolerability. For individuals who have achieved viral suppression, maintaining a stable regimen is the basis of effective care.
Among older adults living with HIV, the conditions become more complex. Polypharmacy is common, and drug interactions must be managed across multiple chronic conditions. Adjustments to treatment are not limited to HIV alone but intersect with cardiovascular, metabolic, and behavioral health considerations.
Within this context, a change in access is not limited to a change in price. It can alter whether a regimen remains viable. The protected classes policy was designed to account for this reality, while the GUARD Model introduces a pricing structure that may alter how that policy functions in practice.
The Gap in Current Analysis
Policy discussion of the GUARD Model has focused on projected savings, rebate structures, and potential effects on innovation. These are valid considerations, but they operate at a level of abstraction. That abstraction assumes uniform patient experience.
What is not consistently examined is how design choices affect populations whose care depends on continuity rather than substitution. Older adults living with HIV are one such population. More than half of people living with HIV in the United States are age 50 and older, a proportion that continues to increase. Their care reflects long-term engagement with treatment systems, not episodic use.
When policy analysis does not name these populations, it defaults to a generalized patient model. That model does not account for conditions where stability is the primary requirement. The absence is not simply descriptive. It shapes how success is defined.
Implementation as the Test
The interaction between pricing models and access protections will not be determined at the level of policy intent. It will be determined through implementation.
Part D plans operate through formularies, tiering structures, and utilization management tools. Changes in pricing incentives influence how these tools are applied. Even where coverage requirements remain, access can be altered through prior authorization, step therapy, or placement within cost-sharing tiers.
If a model introduces financial pressure without specifying safeguards for protected classes beyond existing requirements, plan behavior becomes the site where design is resolved. Equity is not established through general statements of protection but when populations are explicitly accounted for in how systems are required to function.
If older adults living with HIV are not named in the design of safeguards, their outcomes will follow from default plan behavior. Default behavior is not a failure of implementation. It is the execution of design.
Closing
The GUARD Model presents a clear policy objective: reduce drug costs within Medicare. That objective is not in question. What remains unresolved is whether the model’s design accounts for populations whose care depends on maintaining access to specific, non-interchangeable treatments over time.
Making that population visible is not a request for exception. It is a requirement for accurate policy design. Where stability is a clinical condition, not a preference, access cannot be inferred from coverage alone. The distinction becomes visible only when design is examined at the point where policy meets use.