Public Comments

Status: Original and derived artifacts are available for this document.

Download: pdf

DECISION ASSIST LLC
Healthcare Reimbursement & Coverage Strategy Consulting
Lexington, South Carolina

April 13, 2026
Centers for Medicare & Medicaid Services
Department of Health and Human Services
Attention: CMS–1851–P
P.O. Box 8010
Baltimore, MD 21244–1850
Re: Medicare Program; FY 2027 Hospice Wage Index and Payment Rate Update and
Hospice Quality Reporting Program Requirements (CMS–1851–P)
Submitted Electronically via: https://www.regulations.gov/docket/CMS-2026-1156

To Whom It May Concern:
Decision Assist LLC respectfully submits the following comments in response to the
above-referenced proposed rule published in the Federal Register on April 6, 2026 (91
FR 17338). These comments are submitted by Harry Feliciano, MD, MPH, Principal and
Medical Director of Decision Assist LLC, a healthcare reimbursement and coverage
strategy consulting firm. Dr. Feliciano served for 25 years as Contractor/Senior Medical
Director at a major Medicare Administrative Contractor (MAC), during which time he led
medical review operations, authored Local Coverage Determinations (LCDs) for the
hospice benefit, and oversaw claims processing operations across a 16-state jurisdiction.
That operational background informs the analytical perspective offered in these
comments.

1

These comments focus exclusively on Section III.B of the proposed rule, specifically the
analysis of Medicare non-hospice spending during a hospice election, the proposed
Service and Spending Variation Index (SSVI), and the proposed mandatory hospice
election statement addendum. While the Centers for Medicare & Medicaid Services
(CMS) is to be commended for the rigor and transparency of its non-hospice spending
analysis and for the programmatic intent behind both proposals, Decision Assist LLC
respectfully submits that neither intervention, individually or in combination, addresses
the root cause of the documented misclassification errors that are driving the growth in
non-hospice spending. This comment proposes a complementary programmatic
framework that, in the submitter's assessment, would provide CMS with a sustainable,
scalable, and prospective mechanism for reducing that variation at the claims processing
level.

I. AFFIRMING THE CMS EVIDENCE BASE AND DIAGNOSIS-GROUP
CONCENTRATION
CMS has documented a pattern of non-hospice Medicare spending during hospice
elections that is both substantial and diagnostically concentrated. The proposed rule
reports that Medicare payments for non-hospice Part A and Part B items and services
received by hospice beneficiaries increased from nearly $790 million in FY 2020 to over
$2 billion in FY 2024, representing a 160 percent increase over four years. Non-hospice
Part D spending increased from $552.9 million to $813.1 million over the same period, for
a combined non-hospice spending total exceeding $2.8 billion in FY 2024.
Tables 7 and 8 of the proposed rule are particularly instructive. Two diagnostic coding
groups — Neurological/Degenerative (ICD-10 codes G30, G31, G20) and
Heart/Cerebrovascular (ICD-10 codes I11, I25, I50, I63, I67, I69, I13) — account for
$575.6 million and $590.0 million respectively in DME and carrier claims (Table 7), and
$205.1 million and $276.3 million respectively in Part D payments (Table 8). When
respiratory disease claims (ICD-10 codes J44, J96) are added, these three diagnostic
clusters collectively represent approximately 75 percent of the total documented nonhospice spending exposure. This concentration is consistent with a Pareto distribution
and has direct implications for any proposed corrective mechanism. A pilot program
scoped to these three diagnostic clusters would address the substantial majority of the
documented problem, while limiting the operational complexity of initial implementation.

2

CMS's long-standing regulatory position further contextualizes the magnitude of this
problem. Since the original implementation of the Medicare hospice benefit in 1983, CMS
has maintained — and repeatedly reaffirmed across multiple Federal Register
rulemakings (48 FR 56010; 84 FR 38509; 85 FR 47091; 86 FR 19713; 88 FR 20032; 89
FR 64202) — that services unrelated to the terminal illness and related conditions should
be exceptional, unusual, and rare given the comprehensive nature of the services
covered under the hospice benefit. The current data are inconsistent with that regulatory
expectation, and the discrepancy warrants a structural rather than an administrative
response.

II. THE ROOT CAUSE: A CLAIMS SYSTEM PROCESS GATE, NOT A
DISCLOSURE GAP
CMS has identified four potential sources of the documented non-hospice spending
growth: hospices misclassifying conditions, referring patients to non-hospice providers,
failing to coordinate care, or deliberately avoiding costs. These are substantively different
failure modes with different corrective requirements, but they share a common enabling
architecture: the Medicare claims processing system's current handling of the Condition
Code 07 (CC07) bypass mechanism for institutional claims and the GW modifier bypass
mechanism for professional and supplier claims.
These bypass mechanisms were deliberately designed as a provider trust architecture,
consistent with CMS's longstanding deference to physician and provider clinical
judgment. The underlying premise — that hospice providers would responsibly distinguish
related from unrelated services because such determinations would be rare and
exceptional — was reasonable at the time the hospice benefit was designed, when cancer
accounted for the majority of hospice cases and the clinical profile of hospice patients
was relatively uniform. That premise is no longer operationally sustainable.
The current claims processing workflow in the Fiscal Intermediary Shared System (FISS),
the Multi-Carrier System (MCS), and the ViPS Medicare System (VMS) functions as a
swinging door rather than a bona fide process gate. When a non-hospice provider
appends CC07 or the GW modifier to a claim for a Medicare beneficiary with an active
hospice election, the claims system processes that attestation without semantic
interrogation. It does not evaluate whether the service claimed is clinically distinguishable
from the palliation and management responsibilities the hospice assumed at the time of
3

election. It does not cross-reference the service against the hospice's documentation of
the patient's terminal diagnosis, comorbidity profile, or functional status. It processes the
bypass attestation as a binary instruction and pays accordingly.
This architecture means that the root cause of misclassification-driven improper
payments does not reside in a failure of beneficiary disclosure, a failure of provider
education, or a failure of post-payment auditing capacity. It resides in the absence of a
semantic validation function at the claims submission level — a function that would
evaluate whether a given service, item, or drug claimed as unrelated to the terminal illness
and related conditions is, in fact, clinically distinguishable from the scope of care the
hospice per diem is designed to cover.

III. ASSESSMENT OF THE PROPOSED INTERVENTIONS
A. The Mandatory Hospice Election Statement Addendum
The proposal to require hospices to provide the election statement addendum to all
Medicare beneficiaries at the time of hospice election, effective October 1, 2026, is an
appropriate and overdue transparency measure. The regulatory requirement that
hospices document which conditions, items, services, and drugs they have determined
to be unrelated to the terminal illness and related conditions — with supporting clinical
explanation and reference to relevant clinical practice guidelines — creates a structured,
provider-authored document that did not previously exist as a universal admission-time
record.
However, the mandatory addendum is a disclosure instrument, not a validation
instrument. It formalizes and standardizes the communication of existing hospice
determinations but does not evaluate whether those determinations are clinically
defensible. A hospice that systematically misclassifies conditions as unrelated — whether
through clinical error, inadequate documentation processes, or deliberate cost avoidance
— will produce a mandatory addendum that reflects and perpetuates those
misclassifications. The addendum requirement does not contain a mechanism for
detecting that failure at the point of documentation.

4

The significance of the mandatory addendum lies not primarily in its disclosure function
but in the documentary substrate it creates for a downstream semantic validation process.
Beginning October 1, 2026, every eligible hospice election will generate a structured,
standardized document listing the specific conditions, items, services, and drugs the
hospice has determined to be unrelated. That document, if made available to the
processing MAC at the time a non-hospice claim with CC07 or GW modifier is received,
would provide the semantic input necessary for a claims-level validation process of the
kind described in Section IV of these comments. The mandatory addendum is, in this
sense, a necessary but not sufficient precondition for the corrective architecture that
would actually prevent the improper payments CMS has documented.

B. The Service and Spending Variation Index (SSVI)
The SSVI represents a meaningful advance in CMS's capacity to identify hospice-level
outliers across multiple utilization and spending dimensions. The nine-metric scoring
architecture, with particular weight given to total non-hospice spending through an eighttier point allocation, will allow CMS and its contractors to rank hospices by composite risk
profile and focus program integrity resources — medical reviews, targeted education, and
investigations — on the highest-concern providers.
Notwithstanding this advance, the SSVI is, by design, a lagging aggregate outcome
measure. It is calculated from prior fiscal year claims data, published retrospectively, and
applied to focus downstream review activity on already-paid claims. As an analytical tool
in a Define-Measure-Analyze-Improve-Control (DMAIC) quality framework, the SSVI
operationalizes the Measure phase: it quantifies and ranks the output variance that has
already occurred. It does not address the Improve or Control phases, which require
intervening in the process that generates the variance before payment is made.
The SSVI will meaningfully improve the targeting efficiency of the Supplemental Medical
Review Contractor (SMRC) by concentrating retrospective review resources on higherscoring hospices. However, SMRC review is a statistically sampled, retrospective, recordreview process. It recovers a fraction of already-made improper payments and generates
corrective feedback that may influence future provider behavior. It cannot prevent a CC07
or GW modifier bypass claim from being paid tomorrow, next week, or next year. The
SSVI, even fully implemented and accompanied by the mandatory addendum, does not
close the process gate that is currently enabling the scale of non-hospice spending CMS
has documented.

5

Decision Assist LLC respectfully suggests that future iterations of the SSVI would benefit
from disaggregating the non-hospice spending component by the principal hospice
diagnosis coding groups identified in Tables 7 and 8 of the proposed rule. A diagnosisgroup-level SSVI metric would allow CMS and its contractors to distinguish between
hospices whose non-hospice spending is concentrated in the highest-prevalence
diagnostic clusters — Neurological/Degenerative, Heart/Cerebrovascular, and
Respiratory — and hospices whose spending patterns reflect a different profile. This
refinement would improve the SSVI's utility as a Control-phase feedback mechanism and
would align the measure with the Pareto-based pilot scope described below.

IV. A DMAIC-STRUCTURED PILOT: THE INTEGRATED MULTISYSTEM MAC ARCHITECTURE
The following proposal describes an operational framework that would use the
documentary infrastructure created by the mandatory addendum to implement a
prospective, claims-level semantic validation process for CC07 and GW modifier bypass
claims. The framework is grounded in existing MAC infrastructure and does not require
new statutory authority, new claims system development at the national level, or new data
sharing arrangements between otherwise independent contractor entities.

A. The Integrated Multi-System MAC Architecture
The six non-hospice claim types documented in Table 5 of the proposed rule are
processed across three distinct Medicare claims processing systems, each with different
contractor access rights. FISS processes all Part A and institutional Part B claims —
inpatient, outpatient, home health, and skilled nursing facility — making such claims
accessible to any A/B MAC holding concurrent HHH jurisdiction. MCS processes
professional and carrier claims and is similarly accessible to any A/B MAC. VMS — the
ViPS Medicare System — processes DME supplier claims and is accessible exclusively
to DME MACs; A/B MACs do not have VMS access regardless of their HHH scope. This
three-system reality means that no single contractor can access all six claim type
channels unless it holds concurrent A/B MAC HHH and DME MAC jurisdiction over the
same beneficiary geography. That describes a unique position in the Medicare contractor
landscape, and it is what makes a collaborative pilot under a shared corporate parent
6

uniquely feasible without requiring inter-contractor data sharing agreements between
independent organizations.
Decision Assist LLC proposes a bifurcated pilot architecture organized along claims
system access lines, with two forks operating collaboratively under a shared corporate
parent:
Fork 1 — HHH/A/B MAC Scope (FISS and MCS): Covers five of the six Table 5 claim
types — inpatient CC07 ($193.3 million), SNF CC07 ($17.7 million), home health CC07
($20.6 million), outpatient CC07/GW ($202.9 million), and professional/carrier GW
($1.563 billion) — representing approximately $1.997 billion of the $2.066 billion Table 5
total, or approximately 97 percent of the documented non-hospice spending problem.
Fork 1 is executable by any A/B MAC that also holds HHH jurisdiction. The four FISSprocessed claim types benefit from single-system trigger architecture: both the hospice
election record and the triggering CC07 claim reside in FISS, requiring no cross-system
coordination. Professional and carrier GW modifier claims require a straightforward FISSto-MCS cross-reference within the same A/B MAC HHH operational environment.
Fork 2 — Full-Spectrum Scope (FISS, MCS, and VMS): Extends Fork 1 by adding the
DME supplier GW modifier channel ($68.3 million in FY 2024), processed through VMS,
to achieve complete coverage of all six Table 5 claim types. Fork 2 requires concurrent
A/B MAC HHH and DME MAC jurisdiction over the same beneficiary geography — a
configuration that exists for only one Medicare contractor — and can therefore only be
executed by that contractor or through a collaborative arrangement between an A/B MAC
HHH contractor and a DME MAC contractor operating under a shared corporate parent
within the same geographic footprint.
A collaborative pilot under a shared corporate parent — one subsidiary executing Fork 1
across its HHH jurisdiction using FISS and MCS, the other executing Fork 2 within its
overlapping HHH and DME MAC jurisdiction using FISS, MCS, and VMS — would
achieve comprehensive coverage of all six Table 5 claim types across a contiguous
beneficiary geography without requiring any inter-contractor data sharing agreement
between independent entities.
In both forks, the trigger logic is operationally consistent: when a non-hospice claim is
received bearing Condition Code 07 or a GW modifier for a beneficiary with an active
hospice election on file within the designated geographic footprint, the receiving
contractor triggers an automated ADR to the hospice of record for the election statement
addendum. The mandatory addendum, once effective October 1, 2026, will be on file at
the hospice within five days of election and updated within three days of any plan of care
change affecting its content. It is therefore available as a contemporaneous clinical
document at the time most non-hospice claims would be received. The ADR-triggered
retrieval of this document creates the semantic input that a structured clinical validation
7

methodology can then interrogate: does the condition, item, or drug on the non-hospice
claim appear on the addendum as unrelated? If yes, is the hospice's clinical explanation
for that unrelated determination defensible against the diagnostic and functional profile of
the patient's terminal illness and related conditions? If no, the claim warrants either denial
or escalation to a human medical reviewer.

B. The DMAIC Framework Applied to the Pilot
The pilot architecture maps directly onto the DMAIC quality improvement methodology:
1. Define: The problem is defined by CMS's own data in Sections III.B.1 and III.B.2
of the proposed rule — $2.8 billion in annual non-hospice spending during
hospice elections, concentrated in three diagnostic clusters, driven in part by
misclassification errors enabled by a systematic bypass architecture.
2. Measure: The SSVI, once implemented, operationalizes this phase by quantifying
and ranking hospice-level non-hospice spending. Diagnosis-group
disaggregation, as recommended above, would further refine the measurement
instrument.
3. Analyze: SSVI-driven outlier reports enable the SMRC and MACs to identify
which hospices and which diagnostic categories warrant targeted review. The
mandatory addendum creates a new analytical input by providing a structured
record of hospice-level unrelated determinations that can be compared against
the non-hospice claims data for the same beneficiary population.
4. Improve: An ADR-triggered, addendum-based semantic validation process at the
integrated multi-system MAC level, scoped to the three highest-volume
diagnostic clusters, implements a prospective process intervention at the claims
submission level. This is the phase the proposed rule’s current interventions do
not reach.
5. Control: SSVI trend data, disaggregated by diagnosis group, provides the
longitudinal feedback mechanism to evaluate whether the semantic validation
process is reducing unwarranted CC07 and GW modifier bypass claim volume in
the targeted diagnostic clusters over time, and to identify any shift in the problem
to other diagnostic categories.
A further advantage of the proposed geographic scoping is that it creates a natural quasiexperimental design for evaluating pilot effectiveness. Because the integrated multisystem MAC architecture would be implemented within a defined contractor geographic
footprint, Medicare beneficiaries enrolled in hospice in states outside that footprint would
8

constitute a natural control population. Their SSVI scores and non-hospice spending
patterns — driven by the same administrative interventions (mandatory addendum and
SSVI publication) but without the benefit of the claims-level semantic validation process
gate — would provide a contemporaneous comparator against which CMS could
measure the incremental effectiveness of the structural intervention. This quasiexperimental framework would allow CMS to isolate the marginal impact of the process
gate from the background effect of the administrative interventions, producing rigorous,
publishable evidence on whether a claims-level semantic validation methodology
meaningfully reduces unwarranted CC07 and GW modifier bypass volume beyond what
the mandatory addendum and SSVI achieve alone. Decision Assist LLC respectfully
submits that this natural control structure should be explicitly incorporated into the pilot
evaluation design from the outset, and that CMS should commit to publishing the
comparative results in conjunction with the SSVI annual publication for the first two fiscal
years following pilot implementation.

C. Recommended Pilot Scope
Consistent with the Pareto concentration documented in Tables 7 and 8 of the proposed
rule, Decision Assist LLC recommends that the pilot be scoped to the three diagnostic
clusters identified below. The dollar figures cited reflect DME/carrier (Part B) and Part D
non-hospice spending as reported by diagnostic cluster in Tables 7 and 8 — the only
tables in the proposed rule that disaggregate non-hospice spending by diagnosis group.
The pilot scope encompasses all six Table 5 claim types for each cluster, including
inpatient, outpatient, home health, SNF, professional/carrier, and DME claims, consistent
with the Fork 1 and Fork 2 architecture described above. The three clusters collectively
represent approximately 75 percent of the documented non-hospice spending problem:
• Neurological and Degenerative Diseases (ICD-10-CM codes G30, G31, G20):
approximately $780 million in combined DME/carrier (Part B) and Part D nonhospice spending in FY 2024 per Tables 7 and 8 of the proposed rule. Additional
non-hospice spending attributable to this cluster from inpatient, outpatient, home
health, and SNF claim types (documented in aggregate in Table 5 but not
disaggregated by diagnostic cluster in the proposed rule) increases the total
exposure for this cluster beyond the figure cited.
• Circulatory and Cerebrovascular Diseases (ICD-10-CM codes I11, I25, I50, I63,
I67, I69, I13): approximately $866 million in combined DME/carrier (Part B) and
Part D non-hospice spending in FY 2024 per Tables 7 and 8. As with the

9

Neurological/Degenerative cluster, the full cross-claim-type exposure including
inpatient, outpatient, home health, and SNF claims exceeds this figure.
• Respiratory Diseases (ICD-10-CM codes J44, J96): approximately $195 million in
combined DME/carrier (Part B) and Part D non-hospice spending in FY 2024 per
Tables 7 and 8. As with the other two clusters, the full cross-claim-type exposure
including inpatient, outpatient, home health, and SNF claims exceeds this figure.
These three clusters are not selected arbitrarily. They represent the diagnostic categories
in which the clinical boundary between palliation and management of the terminal
condition and care of an unrelated condition is most frequently questioned in medical
review, most susceptible to misclassification through inattentive documentation, and most
directly addressed by existing LCD guidance. A pilot constrained to these clusters is
operationally bounded, evidence-anchored, and clinically coherent. It is also consistent
with CMS's deregulatory posture, as it concentrates pilot resources on the highest-return
opportunity rather than attempting a program-wide implementation before the underlying
methodology has been validated.

V. THE ROLE OF STRUCTURED CLINICAL DOCUMENTATION
METHODOLOGY
The semantic validation function described above requires a clinical reference framework
that can evaluate whether a given service, item, or drug is distinguishable from the
palliation and management responsibilities associated with a specific terminal illness,
related conditions, and comorbidity profile. In Decision Assist LLC's assessment, the
World Health Organization's International Classification of Functioning, Disability and
Health (ICF) provides the most appropriate structural foundation for this function.
The ICF framework characterizes the relationship between a health condition and its
functional consequences across body functions and structures, activities, and
participation, in the context of the individual's environmental and personal factors. Applied
to the hospice related/unrelated determination, the ICF framework enables a clinically
grounded distinction between: (a) services that address the functional consequences of
the terminal illness and related conditions, which are the hospice's coverage
responsibility; and (b) services that address conditions whose functional profile is
genuinely independent of the terminal illness and related conditions, which may
appropriately be classified as unrelated.
10

This distinction is precisely what the current CC07 and GW modifier bypass architecture
cannot make and what the mandatory addendum alone cannot validate. It requires a
structured clinical documentation methodology that maps the beneficiary’s principal
diagnosis, comorbidity profile, and functional status onto an explicit framework for
evaluating the clinical relationship between conditions — and that produces a
documented, auditable basis for the hospice's related/unrelated determination.
The ICF framework's particular suitability for this application is reinforced by its
established presence in Medicare Hospice LCD guidance. HHH MAC-issued LCDs for
coverage of Hospice Alzheimer’s Disease and Related Disorders, Hospice Neurological
Conditions and Hospice Cardiopulmonary Conditions have incorporated ICF-based
conceptual frameworks for evaluating the functional trajectory of terminally ill Medicare
beneficiaries. The application of ICF-grounded documentation methodology to the
related/unrelated determination is therefore not a departure from existing Medicare
coverage policy; it is a natural extension of an analytical approach already embedded in
the coverage landscape.
Decision Assist LLC is engaged in the development and validation of a structured clinical
documentation platform grounded in the ICF framework, designed to support defensible,
auditable related/unrelated determinations at the point of hospice admission and plan of
care update. The platform is currently in proof-of-concept development. These comments
are submitted in the interest of informing CMS's programmatic framework, not in
promotion of any specific commercial product. Decision Assist LLC welcomes the
opportunity to discuss the methodology underlying this platform with CMS and its
contractors as the mandatory addendum implementation and SSVI pilot are
operationalized.

VI. SUMMARY OF RECOMMENDATIONS
Decision Assist LLC respectfully offers the following recommendations for CMS's
consideration in the FY 2027 Hospice Wage Index final rule and in subsequent
programmatic guidance:

11

1. Affirm the mandatory addendum as a necessary precondition for claims-level
semantic validation and, in final rule guidance, explicitly anticipate its role as a
documentary input to MAC-level medical review processes triggered by CC07
and GW modifier claims.
2. Direct the SSVI's non-hospice spending component to be disaggregated by the
principal hospice diagnosis coding groups represented in Tables 7 and 8 of the
proposed rule in future SSVI publications, enabling diagnosis-group-specific
benchmarking and Control-phase feedback.
3. Initiate a DMAIC-structured pilot implementing an Integrated Multi-System MAC
Architecture, bifurcated into Fork 1 (FISS/MCS, executable by any A/B MAC with
concurrent HHH jurisdiction, covering five of six Table 5 claim types representing
approximately 97 percent of documented non-hospice spending) and Fork 2
(FISS/MCS/VMS, requiring concurrent HHH and DME MAC jurisdiction, closing
the DME supplier channel to achieve complete coverage of all six claim types).
The pilot should be scoped to the three diagnostic clusters —
Neurological/Degenerative, Circulatory/Cerebrovascular, and Respiratory —
representing approximately 75 percent of documented non-hospice spending,
and should be implemented through a collaborative arrangement between
contractors sharing a common corporate parent with the requisite overlapping
HHH and DME MAC jurisdictions, thereby avoiding the need for inter-contractor
data sharing agreements between independent entities.
4. Engage with stakeholders developing ICF-framework-grounded structured clinical
documentation methodologies as potential inputs to the semantic validation
function, consistent with CMS's stated interest in clinical practice guidelines and
relevant research as the appropriate basis for related/unrelated determinations.
5. Publish proposed process metrics for the Improve and Control phases of the
DMAIC framework — including pilot-period changes in CC07 and GW modifier
claim volume by diagnosis group — in conjunction with the SSVI annual
publication, to enable transparent assessment of pilot effectiveness.

VII. CONCLUSION
The FY 2027 proposed rule reflects a serious and data-driven effort by CMS to address
a documented and growing program integrity vulnerability. The mandatory addendum and
the SSVI are constructive steps that create, respectively, the documentary substrate and
the measurement infrastructure that a more complete corrective architecture requires.
12

Decision Assist LLC respectfully submits that the next step — the claims-level process
gate that would prevent rather than measure improper payments — is operationally
feasible within existing MAC infrastructure, logically sequenced to follow the mandatory
addendum's October 1, 2026 effective date, and proportionate in scope to the Pareto
concentration of the documented problem.
CMS's own regulatory history on the related/unrelated determination provides the policy
foundation. The 1983 declaration that virtually all care needed by terminally ill patients
would be provided by the hospice was not aspirational; it was a structural description of
the benefit's design. Restoring operational fidelity to that design requires addressing the
process architecture that has allowed systematic deviation from it. Decision Assist LLC
respectfully urges CMS to include the pilot framework described in these comments in its
program integrity planning for the post-addendum implementation period.
Thank you for the opportunity to submit these comments. Decision Assist LLC welcomes
further engagement with CMS, the Center for Program Integrity, and MAC contractors on
the technical and operational aspects of the proposed framework.

Respectfully submitted,

Harry Feliciano, MD, MPH
Principal and Medical Director
Decision Assist LLC

13