[Federal Register Volume 82, Number 85 (Thursday, May 4, 2017)]
[Proposed Rules]
[Pages 20980-21012]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 2017-08519]



[[Page 20979]]

Vol. 82

Thursday,

No. 85

May 4, 2017

Part II





Department of Health and Human Services





-----------------------------------------------------------------------





Centers for Medicare & Medicaid Services





-----------------------------------------------------------------------





42 CFR Parts 409 and 488





Medicare Program; Prospective Payment System and Consolidated Billing 
for Skilled Nursing Facilities: Revisions to Case-Mix Methodology; 
Proposed Rule

  Federal Register / Vol. 82 , No. 85 / Thursday, May 4, 2017 / 
Proposed Rules  

[[Page 20980]]


-----------------------------------------------------------------------

DEPARTMENT OF HEALTH AND HUMAN SERVICES

Centers for Medicare & Medicaid Services

42 CFR Parts 409 and 488

[CMS-1686-ANPRM]
RIN 0938-AT17


Medicare Program; Prospective Payment System and Consolidated 
Billing for Skilled Nursing Facilities: Revisions to Case-Mix 
Methodology

AGENCY: Centers for Medicare & Medicaid Services (CMS), HHS.

ACTION: Advance notice of proposed rulemaking with comment.

-----------------------------------------------------------------------

SUMMARY: We are issuing this advance notice of proposed rulemaking 
(ANPRM) to solicit public comments on potential options we may consider 
for revising certain aspects of the existing skilled nursing facility 
(SNF) prospective payment system (PPS) payment methodology to improve 
its accuracy, based on the results of our SNF Payment Models Research 
(SNF PMR) project. In particular, we are seeking comments on the 
possibility of replacing the SNF PPS' existing case-mix classification 
model, the Resource Utilization Groups, Version 4 (RUG-IV), with a new 
model, the Resident Classification System, Version I (RCS-I). We also 
discuss options for how such a change could be implemented, as well as 
a number of other policy changes we may consider to complement 
implementation of RCS-I.

DATES: To be assured consideration, comments must be received at one of 
the addresses provided below, no later than 5 p.m. on June 26, 2017.

ADDRESSES: In commenting, please refer to file code CMS-1686-ANPRM. 
Because of staff and resource limitations, we cannot accept comments by 
facsimile (FAX) transmission.
    You may submit comments in one of four ways (please choose only one 
of the ways listed):
    1. Electronically. You may submit electronic comments on this 
regulation to http://www.regulations.gov. Within the search bar, enter 
the Regulation Identifier Number associated with this regulation, 0938-
AT17, and then click on the ``Comment Now'' box.
    2. By regular mail. You may mail written comments to the following 
address ONLY: Centers for Medicare & Medicaid Services, Department of 
Health and Human Services, Attention: CMS-1686-ANPRM, P.O. Box 8016, 
Baltimore, MD 21244-8016.
    Please allow sufficient time for mailed comments to be received 
before the close of the comment period.
    3. By express or overnight mail. You may send written comments to 
the following address ONLY: Centers for Medicare & Medicaid Services, 
Department of Health and Human Services, Attention: CMS-1686-ANPRM, 
Mail Stop C4-26-05, 7500 Security Boulevard, Baltimore, MD 21244-1850.
    4. By hand or courier. If you prefer, you may deliver (by hand or 
courier) your written comments before the close of the comment period 
to either of the following addresses:
    a. Centers for Medicare & Medicaid Services, Department of Health 
and Human Services, Room 445-G, Hubert H. Humphrey Building, 200 
Independence Avenue SW., Washington, DC 20201.
    (Because access to the interior of the Hubert H. Humphrey Building 
is not readily available to persons without Federal Government 
identification, commenters are encouraged to leave their comments in 
the CMS drop slots located in the main lobby of the building. A stamp-
in clock is available for persons wishing to retain a proof of filing 
by stamping in and retaining an extra copy of the comments being 
filed.)
    b. Centers for Medicare & Medicaid Services, Department of Health 
and Human Services, 7500 Security Boulevard, Baltimore, MD 21244-1850.
    If you intend to deliver your comments to the Baltimore address, 
please call telephone number (410) 786-7195 in advance to schedule your 
arrival with one of our staff members.
    Comments mailed to the addresses indicated as appropriate for hand 
or courier delivery may be delayed and received after the comment 
period.
    For information on viewing public comments, see the beginning of 
the SUPPLEMENTARY INFORMATION section.

FOR FURTHER INFORMATION CONTACT: John Kane, (410) 786-0557.

SUPPLEMENTARY INFORMATION: Inspection of Public Comments: All comments 
received before the close of the comment period are available for 
viewing by the public, including any personally identifiable or 
confidential business information that is included in a comment. We 
post all comments received before the close of the comment period on 
the following Web site as soon as possible after they have been 
received: http://www.regulations.gov. Follow the search instructions on 
that Web site to view public comments.
    Comments received timely will also be available for public 
inspection as they are received, generally beginning approximately 3 
weeks after publication of a document, at the headquarters of the 
Centers for Medicare & Medicaid Services, 7500 Security Boulevard, 
Baltimore, Maryland 21244, Monday through Friday of each week from 8:30 
a.m. to 4 p.m. To schedule an appointment to view public comments, 
phone 1-800-743-3951.
    To assist readers in referencing sections contained in this 
document, we are providing the following Table of Contents.

Table of Contents

I. Executive Summary
    A. Purpose
    B. Summary of Major Provisions
II. Background
    A. Issues Relating to the Current Case Mix System for Payment of 
Skilled Nursing Facility Services Under Part A of the Medicare 
Program
    B. Summary of the Skilled Nursing Facility Payment Models 
Research Project
III. Potential Revisions to SNF PPS Payment Methodology
    A. Revisions to SNF PPS Base Federal Payment Rate Components
    1. Background on SNF PPS Federal Base Payment Rates and 
Components
    2. Data Sources Utilized for Revision of Federal Base Payment 
Rate Components
    3. Methodology Used for the Calculation of Revised Federal Base 
Payment Rate Components
    4. Updates and Wage Adjustments of Revised Federal Base Payment 
Rate Components
    B. Potential Design and Methodology for Case-Mix Adjustment of 
Federal Rates
    1. Background on Resident Classification System, Version I
    2. Data Sources Utilized for Developing RCS-I
    a. Medicare Enrollment Data
    b. Medicare Claims Data
    c. Assessment Data
    d. Facility Data
    3. Resident Classification Under RCS-I
    a. Background
    b. Physical and Occupational Therapy Case-Mix Classification
    c. Speech-Language Pathology Case-Mix Classification
    d. Nursing Case-Mix Classification
    e. Non-Therapy Ancillary Case-Mix Classification
    f. Payment Classifications under RCS-I
    4. Variable Per Diem Adjustment Factors and Payment Schedule
    C. Use of the Resident Assessment Instrument--Minimum Data Set, 
Version 3
    1. Potential Revisions to Minimum Data Set (MDS) Completion 
Schedule
    2. Potential Revisions to Therapy Provision Policies Under the 
SNF PPS
    3. Interrupted Stay Policy
    D. Relationship of RCS-I to Existing Skilled Nursing Facility 
Level of Care Criteria
    E. Effect of RCS-I on Temporary AIDS Add-on Payment

[[Page 20981]]

    F. Potential Impacts of Implementing RCS-I
IV. Collection of Information Requirements
V. Response to Comments

Acronyms

    In addition, because of the many terms to which we refer by acronym 
in this ANPRM, we are listing these abbreviations and their 
corresponding terms in alphabetical order below:

AIDS Acquired Immune Deficiency Syndrome
ARD Assessment reference date
BBRA Medicare, Medicaid, and SCHIP Balanced Budget Refinement Act of 
1999, Public Law 106-113
CASPER Certification and Survey Provider Enhanced Reporting
CCN CMS Certification Number
CFR Code of Federal Regulations
CMI Case-mix index
CMS Centers for Medicare & Medicaid Services
FR Federal Register
FY Fiscal year
ICD-10-CM International Classification of Diseases, 10th Revision, 
Clinical Modification
IPPS Inpatient prospective payment system
IRF Inpatient Rehabilitation Facility
IRF-PAI Inpatient Rehabilitation Facility Patient Assessment 
Instrument
LTCH Long-term care hospital
MDS Minimum data set
MMA Medicare Prescription Drug, Improvement, and Modernization Act 
of 2003, Public Law 108-173
NF Nursing facility
NTA Non-therapy ancillary
OASIS Outcome and Assessment Information Set
OMB Office of Management and Budget
PAC Post-acute care
PPS Prospective Payment System
QIES Quality Improvement and Evaluation System
QIES ASAP Quality Improvement and Evaluation System Assessment 
Submission and Processing
RAI Resident assessment instrument
RCS-I Resident Classification System, Version I
RFA Regulatory Flexibility Act, Public Law 96-354
RIA Regulatory impact analysis
RUG-III Resource Utilization Groups, Version 3
RUG-IV Resource Utilization Groups, Version 4
RUG-53 Refined 53-Group RUG-III Case-Mix Classification System
SNF Skilled nursing facility
SNF PMR Skilled Nursing Facility Payment Models Research
STM Staff time measurement
STRIVE Staff time and resource intensity verification
TEP Technical expert panel

I. Executive Summary

A. Purpose

    This ANPRM solicits comments on options we may consider for 
revising certain aspects of the existing SNF PPS payment methodology, 
to improve its accuracy, based on the results of the SNF PMR project. 
In particular, we are seeking comments on the possibility of replacing 
the SNF PPS' existing case-mix classification model, RUG-IV, with the 
RCS-I case mix model developed during the SNF PMR project. We also 
discuss and seek comment on options for how such a change could be 
implemented, as well as a number of other policy changes we may 
consider to complement implementation of RCS-I. We would note that we 
intend to propose case-mix refinements in the FY 2019 SNF PPS proposed 
rule, and this ANPRM serves to solicit comments on potential revisions 
we are considering proposing in such rulemaking.

B. Summary of Major Provisions

    In section II of this ANPRM, we discuss the current SNF PPS, 
specifically the RUG-IV case-mix classification methodology that is 
used to assign SNF Part A residents to payment groups that reflect 
varying levels of resource intensity. We also discuss issues with the 
current system which prompted CMS to consider potential revisions to 
the existing case-mix methodology. Finally, we discuss the SNF PMR 
project, which was intended to develop a replacement for the RUG-IV 
case-mix classification model within our current statutory authority.
    In section III. of this ANPRM, we discuss the case-mix model that 
could serve to replace RUG-IV, which is the RCS-I model. We begin by 
discussing the revised base rate structure that would be used under 
RCS-I, based on certain changes to the existing SNF PPS case-mix 
adjusted components that we are considering, based on the findings from 
the SNF PMR project. Similar to the current system, RUG-IV, the revised 
model, the RCS-I, would case-mix adjust for the following major cost 
categories: Physical therapy (PT), occupational therapy (OT), speech-
language pathology (SLP) services, nursing services and non-therapy 
ancillaries (NTAs). However, where RUG-IV consists of two case-mix 
adjusted components (therapy and nursing), the RCS-I would create four 
(PT/OT, SLP, nursing, and NTA) for a more resident-centered case-mix 
adjustment. We then discuss each of the potential case-mix adjusted 
components under the RCS-I model, including how residents would be 
classified under each case-mix component and the resident-
characteristics that our research indicates could serve as appropriate 
predictors of varying resource intensity for each component. Finally, 
we also discuss and solicit public comments on other potential policy 
changes, developed under the SMF PMR project, to the SNF PPS payment 
methodology.

II. Background

A. Issues Relating to the Current Case-Mix System for Payment of 
Skilled Nursing Facility Services Under Part A of the Medicare Program

    Section 1888(e)(4)(G)(i) of the Act requires the Secretary to make 
an adjustment to the per diem rates to account for case-mix. The 
statute specifies that the adjustment is to be based on both a resident 
classification system that the Secretary establishes that accounts for 
the relative resource use of different resident types, as well as 
resident assessment and other data that the Secretary considers 
appropriate.
    In general, the case-mix classification system currently used under 
the SNF PPS classifies residents into payment classification groups, 
called RUGs, based on various resident characteristics and the type and 
intensity of therapy services provided to the resident. Each RUG is 
assigned a set of case-mix indexes (CMIs) that reflect relative 
differences in cost and resource intensity for each case-mix adjusted 
component. The higher the CMI, the higher the expected resource 
utilization and cost associated with that resident's care. Under the 
existing SNF PPS methodology, there are two case-mix components. The 
nursing component reflects relative differences in a resident's 
associated nursing and non-therapy ancillary (NTA) costs, based on 
various resident characteristics, such as resident comorbidities, and 
treatments. The therapy component reflects relative differences in a 
resident's associated therapy costs, which is based on a combination of 
PT, OT, and SLP services. Resident classification under the existing 
therapy component is based primarily on the amount of therapy the SNF 
chooses to provide to a SNF resident. Under the RUG-IV model, residents 
are classified into rehabilitation groups, where payment is determined 
primarily based on the intensity of therapy services received by the 
resident, and into nursing groups, based on the intensity of nursing 
services received by the resident and other aspects of the resident's 
care and condition. However, only the higher paying of these groups is 
used for payment purposes. For example, if a resident is classified 
into a both the RUA (Rehabilitation) and PA1 (Nursing) RUG-IV groups, 
where RUA has a higher per-diem payment rate than PA1,

[[Page 20982]]

the RUA group is used for payment purposes. It should be noted that the 
vast majority of Part A covered SNF days (over 90 percent) are paid 
using a rehabilitation RUG. A variety of concerns have been raised with 
the current SNF PPS, specifically the RUG-IV model, which we discuss 
below.
    When the SNF PPS was first implemented (63 FR 26252), we developed 
the RUG-III case-mix classification model, which tied the amount of 
payment to resident resource use in combination with resident 
characteristic information. Staff time measurement (STM) studies 
conducted in 1990, 1995, and 1997 provided information on resource use 
(time spent by staff members on residents) and resident characteristics 
that enabled us not only to establish RUG-III, but also to create CMIs. 
This initial RUG-III model was refined by changes finalized in the FY 
2006 SNF PPS final rule (70 FR 45032), which included adding nine case-
mix groups to the top of the original 44-group RUG-III hierarchy, which 
created the RUG-53 case-mix model.
    In the FY 2010 SNF PPS proposed rule (74 FR 22208), we proposed a 
revised RUG-IV model based on, among other reasons, concerns that 
incentives in the SNF PPS had changed the relative amount of nursing 
resources required to treat SNF residents (74 FR 22220). These concerns 
led us to conduct a new Staff Time Measurement (STM) study, the Staff 
Time and Resource Intensity Verification (STRIVE) project, which served 
as the basis for developing the current SNF PPS case-mix classification 
model, RUG-IV, which became effective in FY 2011. At that time, we 
considered alternative case mix models, including predictive models of 
therapy payment based on resident characteristics; however, we had a 
``great deal of concern that by separating payment from the actual 
provision of services, the system, and more importantly, the 
beneficiaries would be vulnerable to underutilization.'' (74 FR 22220). 
Other options considered at the time included a non-therapy ancillary 
(NTA) payment model based on resident characteristics (74 FR 22238) and 
a DRG-based payment model that relied on information from the prior 
inpatient stay (74 FR 22220); these and other options are discussed in 
detail in a CMS Report to Congress issued in December 2006 (available 
at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/Downloads/RC_2006_PC-PPSSNF.pdf).
    In the years since we implemented the SNF PPS, finalized RUG-IV, 
and made statements regarding our concerns about underutilization of 
services in previously considered models, we have witnessed a 
significant trend that has caused us to reconsider these concerns. More 
specifically, as discussed in section V.E. of the FY 2015 SNF PPS 
proposed rule (79 FR 25767), we documented and discussed trends 
observed in therapy utilization in a memo entitled ``Observations on 
Therapy Utilization Trends'' (which may be accessed at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/Downloads/Therapy_Trends_Memo_04212014.pdf). The two most notable trends 
discussed in that memo were that the percentage of residents 
classifying into the Ultra-High therapy category has increased steadily 
and, of greater concern, that the percentage of residents receiving 
just enough therapy to surpass the Ultra-High and Very-High therapy 
thresholds has also increased. In that memo, we state ``the percentage 
of claims-matched MDS assessments in the range of 720 minutes to 739 
minutes, which is just enough to surpass the 720 minute threshold for 
RU groups, has increased from 5 percent in FY 2005 to 33 percent in FY 
2013'' and this trend has continued since that time. While it might be 
possible to attribute the increasing share of residents in the Ultra-
High therapy category to increasing acuity within the SNF population, 
we believe the increase in ``thresholding'' (that is, of providing just 
enough therapy for residents to surpass the relevant therapy 
thresholds) is a strong indication of service provision predicated on 
financial considerations rather than resident need. We discussed this 
issue in response to comments in the FY 2015 SNF PPS final rule, where, 
in response to comments regarding the lack of ``current medical 
evidence related to how much therapy a given resident should receive,'' 
we stated the following:

    With regard to the comments which highlight the lack of existing 
medical evidence for how much therapy a given resident should 
receive, we would note that . . . the number of therapy minutes 
provided to SNF residents within certain therapy RUG categories is, 
in fact, clustered around the minimum thresholds for a given therapy 
RUG category. However, given the comments highlighting the lack of 
medical evidence related to the appropriate amount of therapy in a 
given situation, it is all the more concerning that practice 
patterns would appear to be as homogenized as the data would 
suggest. (79 FR 45651)

    In response to comments related to factors which may explain the 
observed trends, we stated the following:

    With regard to the comment which highlighted potential 
explanatory factors for the observed trends, such as internal 
pressure within SNFs that would override clinical judgment, we find 
these potential explanatory factors troubling and entirely 
inconsistent with the intended use of the SNF benefit. Specifically, 
the minimum therapy minute thresholds for each therapy RUG category 
are certainly not intended as ceilings or targets for therapy 
provision. As discussed in Chapter 8, Section 30 of the Medicare 
Benefit Policy Manual (Pub. 100-02), to be covered, the services 
provided to a SNF resident must be ``reasonable and necessary for 
the treatment of a patient's illness or injury, that is, are 
consistent with the nature and severity of the individual's illness 
or injury, the individual's particular medical needs, and accepted 
standards of medical practice.'' (emphasis added) Therefore, 
services which are not specifically tailored to meet the 
individualized needs and goals of the resident, based on the 
resident's condition and the evaluation and judgment of the 
resident's clinicians, may not meet this aspect of the definition 
for covered SNF care, and we believe that internal provider rules 
should not seek to circumvent the Medicare statute, regulations and 
policies, or the professional judgment of clinicians. (79 FR 45651 
through 45652)

    In addition to this discussion of observed trends, others have also 
identified potential areas of concern within the current SNF PPS. The 
two most notable sources are the Office of the Inspector General (OIG) 
and the Medicare Payment Advisory Commission (MedPAC).
    With regard to the OIG, three recent OIG reports describe the OIG's 
concerns with the current SNF PPS. In December 2010, the OIG released a 
report entitled ``Questionable Billing by Skilled Nursing Facilities'' 
(which may be accessed at https://oig.hhs.gov/oei/reports/oei-02-09-00202.pdf). In this report, among its findings, the OIG found that 
``from 2006 to 2008, SNFs increasingly billed for higher paying RUGs, 
even though beneficiary characteristics remained largely unchanged'' 
(OEI-02-09-00202, ii), and among other things, recommended that we 
should ``consider several options to ensure that the amount of therapy 
paid for by Medicare accurately reflects beneficiaries' needs'' (OEI-
02-09-00202, iii). Further, in November 2012, the OIG released a report 
entitled ``Inappropriate Payments to Skilled Nursing Facilities Cost 
Medicare More Than a Billion Dollars in 2009'' (which may be accessed 
at https://oig.hhs.gov/oei/reports/oei-02-09-00200.pdf). In this 
report, the OIG found that ``SNFs billed one-quarter of all claims in 
error in 2009'' and that the ``majority of the claims in error were 
upcoded; many of these claims were for ultrahigh

[[Page 20983]]

therapy.'' (OEI-02-09-00200, Executive Summary). Among its 
recommendations, the OIG stated that ``the findings of this report 
provide further evidence that CMS needs to change how it pays for 
therapy'' (OEI-02-09-00200, 15). Finally, in September 2015, the OIG 
released a report entitled ``The Medicare Payment System for Skilled 
Nursing Facilities Needs to be Reevaluated'' (which may be accessed at 
https://oig.hhs.gov/oei/reports/oei-02-13-00610.pdf). Among its 
findings, the OIG found that ``Medicare payments for therapy greatly 
exceed SNFs' costs for therapy,'' further noting that ``the difference 
between Medicare payments and SNFs' costs for therapy, combined with 
the current payment method, creates an incentive for SNFs to bill for 
higher levels of therapy than necessary'' (OEI-02-13-00610, 7). Among 
its recommendations, the OIG stated that CMS should ``change the method 
of paying for therapy,'' further stating that ``CMS should accelerate 
its efforts to develop and implement a new method of paying for therapy 
that relies on beneficiary characteristics or care needs.'' (OEI-02-13-
00610, 12).
    With regard to MedPAC's recommendations in this area, Chapter 8 of 
MedPAC's March 2017 Report to Congress (available at http://www.medpac.gov/docs/default-source/reports/mar17_medpac_ch8.pdf) 
includes the following recommendation: ``The Congress should . . . 
direct the Secretary to revise the prospective payment system (PPS) for 
skilled nursing facilities'' and ``. . . make any additional 
adjustments to payments needed to more closely align payment with 
costs.'' (March 2017 MedPAC Report to Congress, 220). This 
recommendation is seemingly predicated on MedPAC's own analysis of the 
current SNF PPS, where they state that ``almost since its inception the 
SNF PPS has been criticized for encouraging the provision of excessive 
rehabilitation therapy services and not accurately targeting payments 
for nontherapy ancillaries'' (March 2017 MedPAC Report to Congress, 
202). Finally, with regard to the possibility of changing the existing 
SNF payment system, MedPAC stated that ``since 2015, [CMS] has gathered 
four expert panels to receive input on aspects of possible design 
features before it proposes a revised PPS'' and further that ``the 
designs under consideration are consistent with those recommended by 
the Commission'' (March 2017 MedPAC Report to Congress, 203).
    The combination of the observed trends in the current SNF PPS 
discussed above (which strongly suggest that providers may be basing 
service provision on financial reasons rather than resident need), the 
issues raised in the OIG reports discussed above, and the issues raised 
by MedPAC, has caused us to consider significant revisions to the 
existing SNF PPS, in keeping with our overall responsibility to ensure 
that payments under the SNF PPS accurately reflect both resident needs 
and resource utilization.
    Under the RUG-IV system, therapy service provision determines not 
only therapy payments, but also nursing payments. This is because, as 
noted above, only one of a resident's assigned RUG groups, 
rehabilitation or nursing, is used for payment purposes. Each 
rehabilitation group is assigned a nursing CMI to reflect relative 
differences in nursing costs for residents in those rehabilitation 
groups, which is less specifically tailored to the individual nursing 
costs for a given resident than the nursing CMIs assigned for the 
nursing RUGs. Given that, as mentioned above, most resident days are 
paid using a rehabilitation RUG, and since assignment into a 
rehabilitation RUG is based on therapy service provision, this means 
that therapy service provision effectively determines nursing payments 
for those residents who are assigned to a rehabilitation RUG. Thus, we 
believe any attempts to revise the SNF PPS payment methodology to 
better account for therapy service provision under the SNF PPS would 
need to be comprehensive and affect both the therapy and nursing case-
mix components. Moreover, in the FY 2015 SNF PPS final rule, in 
response to comments regarding access for certain ``specialty'' 
populations (such as those with complex nursing needs), we stated the 
following:

    With regard to the comment on specialty populations, we agree 
with the commenter that access must be preserved for all categories 
of SNF residents, particularly those with complex medical and 
nursing needs. As appropriate, we will examine our current 
monitoring efforts to identify any revisions which may be necessary 
to account appropriately for these populations. (79 FR 45651)

    In addition, MedPAC, in their March 2017 Report to Congress, stated 
that they have previously recommended that we revise the current SNF 
PPS to ``base therapy payments on patient characteristics (not service 
provision), remove payments for NTA services from the nursing 
component, [and] establish a separate component within the PPS that 
adjusts payments for NTA services'' (March 2017 MedPAC Report to 
Congress, 202). Accordingly, we note that included among the potential 
revisions we discuss in this ANPRM, are revisions to the SNF PPS to 
address longstanding concerns regarding the ability of the RUG-IV 
system to account for variation in nursing and NTA services, as 
described in sections III.D.3.d and III.D.3.e. of this ANPRM.
    In the sections that follow, we solicit comments on comprehensive 
revisions to the current SNF PPS case-mix classification system. 
Specifically, we discuss a potential alternative to the existing RUG-
IV, called RCS-I, which we are considering. We solicit comment on the 
extent to which RCS-I addresses the issues we outline above. As further 
discussed below, we believe that the RCS-I model represents an 
improvement over the RUG-IV model because it would better account for 
resident characteristics and care needs, thus better aligning SNF PPS 
payments with resource use and eliminating therapy provision-related 
financial incentives inherent in the current payment model used in the 
SNF PPS. To better ensure that resident care decisions appropriately 
reflect each resident's actual care needs, we believe it is important 
to remove, to the extent possible, service-based metrics from the SNF 
PPS and derive payment from objective resident characteristics.

B. Summary of the Skilled Nursing Facility Payment Models Research 
Project

    As noted above, since 1998, Medicare Part A has paid for SNF 
services on a per diem basis through the SNF PPS. Currently, therapy 
payments under the SNF PPS are based primarily on the amount of therapy 
furnished to a patient, regardless of that patient's specific 
characteristics and care needs. Beginning in 2013, we contracted with 
Acumen, LLC to identify potential alternatives to the existing 
methodology used to pay for services under the SNF PPS. The 
recommendations developed under this contract, entitled the SNF PMR 
project, form the basis of the ideas contained in the sections below.
    The SNF PMR operated in three phases. In the first phase of the 
project, which focused exclusively on therapy payment issues, Acumen 
reviewed past research studies and policy issues related to SNF PPS 
therapy payment and options for improving or replacing the current 
therapy payment methodology. After consideration of multiple potential 
alternatives, such as competitive bidding and a hybrid model combining 
resource-based pricing (for example, how therapy payments are made 
under the current SNF PPS) with resident characteristics, we identified 
a model that relies on resident

[[Page 20984]]

characteristics rather than the amount of therapy received as the most 
appropriate replacement for the existing therapy payment model. As 
stated above, we believe that relying on resident characteristics would 
improve the resident-centeredness of the model and discourage resident 
care decisions predicated on service-based financial incentives. A 
report summarizing Acumen's activities and recommendations during the 
first phase of the SNF PMR contract, the SNF Therapy Payment Models 
Base Year Final Summary Report, is available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/Downloads/Summary_Report_20140501.pdf.
    In the second phase of the project, Acumen used the findings from 
the Base Year Final Summary Report as a guide to identify potential 
models suitable for further analysis. During this phase of the project, 
in an effort to establish a comprehensive approach to Medicare Part A 
SNF payment reform, we expanded the scope of the SNF PMR to encompass 
other aspects of the SNF PPS beyond therapy. Although we always 
intended to ensure that any revisions specific to therapy payment would 
be considered as part of an integrated approach with the remaining 
payment methodology, we felt it prudent to examine potential 
improvements and refinements to the overall SNF PPS payment system as 
well.
    During this phase of the SNF PMR, Acumen hosted four Technical 
Expert Panels (TEPs), which brought together industry experts, 
stakeholders, and clinicians with the research team to discuss 
different topics within the overall analytic framework. In February 
2015, Acumen hosted a TEP to discuss questions and issues related to 
therapy case-mix classification. In November 2015, Acumen hosted a 
second TEP focused on questions and issues related to nursing case-mix 
classification, as well as to discuss issues related to payment for 
NTAs. In June 2016, Acumen hosted a third TEP to provide stakeholders 
with an outline of a potential revised SNF PPS payment structure, 
including new case-mix adjusted components and potential companion 
policies, such as variable per diem payment adjustments. Finally, in 
October 2016, Acumen hosted a fourth TEP, during which Acumen presented 
the case-mix components for a potential revised SNF PPS, as well as an 
initial impact analysis associated with the potential revised SNF PPS 
payment model. The presentation slides used during each of the TEPs, as 
well as a summary report for each TEP, is available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.
    In the final phase of the contract, which is ongoing, we tasked 
Acumen to assist in developing supporting language and documentation, 
most notably a technical report, related to the alternative SNF PPS 
case-mix classification model we are considering, which we have named 
the RCS-I.
    This ANPRM solicits comments on the issues with the current SNF 
PPS, and what steps should be taken to refine the existing SNF PPS in 
response to those issues. In particular, in this ANPRM, we discuss and 
are soliciting comments regarding how we could replace the existing 
RUG-IV case-mix classification model with a potential alternative such 
as the RCS-I case-mix classification model. We solicit comments on the 
adequacy and appropriateness of the RCS-I case-mix model to serve as a 
replacement for the RUG-IV model. Our goals in developing a potential 
alternative are as follows:
     To create a model that compensates SNFs accurately based 
on the complexity of the particular beneficiaries they serve and the 
resources necessary in caring for those beneficiaries; and
     To address our concerns, along with those of OIG and 
MedPAC, about current incentives for SNFs to deliver therapy to 
beneficiaries based on financial considerations, rather than the most 
effective course of treatment for beneficiaries; and
     To maintain simplicity by, to the extent possible, 
limiting the number and type of elements we use to determine case-mix, 
as well as limiting the number of assessments necessary under the 
payment system.
    We solicit comment on the goals outlined above and how effective 
the RCS-I system we outline below is at addressing those goals.
    In addition to the general discussion of RCS-I, we also discuss and 
are soliciting public comment on certain complementary policies that we 
believe could also serve to improve the SNF PPS. To provide commenters 
with an appropriate basis for comment on RCS-I, we also discuss the 
potential impact to providers of implementing this type of model. We 
also solicit public comment on certain logistical aspects of 
implementing revisions to the current SNF PPS, such as whether those 
revisions should be implemented in a budget neutral manner, and how 
much lead time providers and other stakeholders should receive before 
any finalized changes would be implemented. Finally, we are soliciting 
public comment on other potential issues CMS should consider in 
implementing revisions to the current SNF PPS, such as potential 
effects on state Medicaid programs, potential behavioral changes, and 
the type of education and training that would be necessary to implement 
successfully any changes to the SNF PPS.
    In the sections below, we outline each aspect of the RCS-I case-mix 
classification model we are considering, as well as additional 
revisions to the SNF PPS which may be considered along with potential 
implementation of the RCS-I classification model. We invite comments on 
any and all aspects of the RCS-I case-mix model, including the research 
analyses described in this ANPRM and in the SNF PMR Technical Report 
(available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html), as well as on any of the other 
considerations discussed in this ANPRM.

III. Potential Revisions to SNF PPS Payment Methodology

A. Revisions to SNF PPS Federal Base Payment Rate Components

1. Background on SNF PPS Federal Base Payment Rates and Components
    Section 1888(e)(4) of the Act requires that the SNF PPS per diem 
federal payment rates be based on FY 1995 costs, updated for inflation. 
These base rates are then required to be adjusted to reflect 
differences in patient case-mix. In keeping with this statutory 
requirement, the base per diem payment rates were set in 1998 and 
reflect average SNF costs in a base year (FY 1995), updated for 
inflation to the first period of the SNF PPS, which was the 15-month 
period beginning on July 1, 1998. The federal base payment rates were 
calculated separately for urban and rural facilities and based on 
allowable costs from the FY 1995 cost reports of hospital-based and 
freestanding SNFs, where allowable costs included all routine, 
ancillary, and capital-related costs (excluding those related to 
approved educational activities) associated with SNF services provided 
under Part A, and all services and items for which payment could be 
made under Part B prior to July 1, 1998.
    In general, routine costs are those included by SNFs in a daily 
service charge and include regular room, dietary, and nursing services, 
medical social services and psychiatric social services, as well as the 
use of certain facilities and equipment for which a separate charge is 
not made. Ancillary

[[Page 20985]]

costs are directly identifiable to residents and cover specialized 
services, including therapy, drugs, and laboratory services. Lastly, 
capital-related costs include the costs of land, building, and 
equipment and the interest incurred in financing the acquisition of 
such items. (63 FR 26253)
    There are four federal base payment rate components which may 
factor into SNF PPS payment. Two of these components, ``nursing case-
mix'' and ``therapy case-mix,'' are case-mix adjusted components, while 
the remaining two components, ``therapy non-case-mix'' and ``non-case-
mix,'' are not case-mix adjusted. While we discuss the details of the 
RCS-I payment model and justifications for certain associated policies 
we are considering in section III.D. of this ANPRM, we note that, as 
part of the RCS-I case-mix model under consideration, we would 
bifurcate both the ``nursing case-mix'' and ``therapy case-mix'' 
components of the federal base payment rate into two components each, 
thereby creating four case-mix adjusted components. More specifically, 
we would separate the ``therapy case-mix'' rate component into a 
``Physical Therapy/Occupational Therapy'' (PT/OT) component and a 
``Speech-Language Pathology'' (SLP) component. Our rationale for 
bifurcating the therapy case-mix component in this manner is presented 
in section III.D.3.b. of this ANPRM. Based on the results of the SNF 
PMR, we would also separate the ``nursing case-mix'' rate component 
into a ``nursing'' component and a ``Non-Therapy Ancillary'' (NTA) 
component. Our rationale for bifurcating the nursing case-mix component 
in this manner is presented in section III.D.3.e. of this ANPRM. Given 
that all SNF residents, under the RCS-I model, would be assigned to a 
classification group for each of the two therapy-related case-mix 
adjusted components as further discussed below, we believe that we 
could eliminate the ``therapy non-case-mix'' rate component under the 
RCS-I model. The existing non-case-mix component could be maintained as 
it is currently constituted under the existing SNF PPS. Although the 
case-mix components of the RCS-I case-mix classification system would 
address costs associated with individual resident care based on an 
individual's specific needs and characteristics, the non-case-mix 
component addresses consistent costs that are incurred for all 
residents, such as room and board and various capital-related expenses. 
As these costs are not likely to change, regardless of what changes we 
might make to the SNF PPS, we believe it would be appropriate to 
continue using the non-case-mix component as it is currently used.
    In the next section, we discuss the methodology we used to 
bifurcate the federal base payment rates for each of the two existing 
case-mix adjusted components, as well as the data sources used in this 
calculation. The methodology does not calculate new federal base 
payment rates, but simply splits the existing base rate case-mix 
components for therapy and nursing. The methodology and data used in 
this calculation are based on the data and methodology used in the 
calculation of the original federal payment rates in 1998, as further 
discussed below.
2. Data Sources Utilized for Revision of Federal Base Payment Rate 
Components
    Section II.A.2. of the interim final rule with comment period that 
initially implemented the SNF PPS (63 FR 26256 through 26260) provides 
a detailed discussion of the data sources used to calculate the 
original federal base payment rates in 1998. We are considering using 
the same data sources to determine the portion of the therapy case-mix 
component base rate that would be assigned to the SLP component base 
rate. As described in section III.C.3. of this ANPRM, the methodology 
for bifurcating the nursing component base rate is different than the 
methodology used for bifurcating the therapy component base rate, 
despite using the same data sources. The portion of the nursing 
component base rate that corresponds to NTA costs was already 
calculated using the same data source used to calculate the federal 
base payment rates in 1998. As explained below, we used the previously 
calculated percentage of the nursing component base rate corresponding 
to NTA costs to set the NTA base rate, and verified this calculation 
with the analysis described in section III.C.3 of this ANPRM. 
Therefore, the steps described below address the calculations performed 
to bifurcate the therapy base rate alone.
    The percentage of the current therapy case-mix component of the 
federal base payment rates that would be assigned to the SLP component 
of the federal base payment rates was determined using cost information 
from FY 1995 cost reports, after making the following exclusions and 
adjustments: First, only settled and as-submitted cost reports for 
hospital-based and freestanding SNFs for periods beginning in FY 1995 
and spanning 10 to 13 months were included. This set of restrictions 
replicates the restrictions used to derive the original federal base 
payment rates as set forth in the 1998 interim final rule with comment 
period (63 FR 26256). Following the methodology used to derive the SNF 
PPS base rates, routine and ancillary costs from ``as submitted'' cost 
reports were adjusted down by 1.31 and 3.26 percent, respectively. As 
discussed in the 1998 interim final rule with comment period, the 
specific adjustment factors were chosen to reflect average adjustments 
resulting from cost report settlement and were based on a comparison of 
as-submitted and settled reports from FY 1992 to FY 1994 (63 FR 26256); 
these adjustments are in accordance with section 1888(e)(4)(A)(i) of 
the Act. We used similar data, exclusions, and adjustments as in the 
original base rates calculation so the resulting base rates for the 
components would resemble as closely as possible what they would have 
been had they been established in 1998. However, there were two ways in 
which the SLP percentage calculation deviates from the 1998 base rates 
calculation. First, the 1998 calculation of the base rates excluded 
reports for facilities exempted from cost limits in the base year. The 
available data do not identify which facilities were exempted from cost 
limits in the base year, so this restriction was not implemented. We do 
not believe this had a notable impact on our estimate of the SLP 
percentage, because only a small fraction of facilities were exempted 
from cost limits. Consistent with the 1998 base rates calculation, we 
excluded facilities with per diem costs more than three standard 
deviations higher than the geometric mean across facilities. Therefore, 
facilities with unusually high costs did not influence our estimate. 
Second, the 1998 calculation of the base rates excluded costs related 
to exceptions payments and costs related to approved educational 
activities. The available cost report data did not identify costs 
related to exceptions payments nor indicate what percentage of overall 
therapy costs or costs by therapy discipline were related to approved 
educational activities, so these costs are not excluded from the SLP 
percentage calculation. Because exceptions were only granted for 
routine costs, we believe the inability to exclude these costs should 
not affect our estimate of the SLP percentage (as exceptions would not 
apply to therapy costs). Additionally, the data indicate that 
educational costs made up less than one-hundredth of 1 percent of 
overall SNF costs. If the proportion of educational costs is relatively 
uniform across cost categories, the inability to

[[Page 20986]]

exclude these costs should have a negligible impact on our estimate.
    In addition to Part A costs from the cost report data, the 1998 
federal base rates calculation incorporated estimates of amounts 
payable under Part B for covered SNF services provided to Part A SNF 
residents, as required by section 1888(e)(4)(A)(ii) of the Act. In 
calculating the SLP percentage, we also estimated the amounts payable 
under Part B for covered SNF services provided to Part A residents. All 
Part B claims associated with Part A SNF claims overlapping with FY 
1995 cost reports were matched to the corresponding facility's cost 
report. For each cost center (for example, SLP, PT, OT) in each cost 
report, a ratio was calculated to determine the amount by which Part A 
costs needed to be increased to account for the portion of costs 
payable under Part B. This ratio for each cost center was determined by 
dividing the total charges from the matched Part B claims by the total 
charges from the Part A SNF claims overlapping with the cost report.
    Finally, the 1998 federal base rates calculation standardized the 
cost data for each facility to control for the effects of case-mix and 
geographic-related wage differences, as required by section 
1888(e)(4)(C) of the Act. When calculating the SLP share of the current 
therapy base rate, we replicated the method used in 1998 to standardize 
for wage differences, as described in the 1998 interim final rule with 
comment period (63 FR 26259 through 26260). We applied a hospital wage 
index to the labor-related share of costs, estimated at 75.888 percent, 
and used an index composed of hospital wages from FY 1994. The SLP 
percentage calculation did not include the case-mix adjustment used in 
the 1998 calculation because the 1998 adjustment relied on the obsolete 
RUG-III classification system. In the 1998 federal base rates 
calculation, information from SNF and inpatient claims was mapped to 
RUG-III clinical categories at the resident level to case-mix adjust 
facility per diem costs. However, the 1998 interim final rule did not 
document this mapping, and the data used as the basis for this 
adjustment are no longer available, and therefore this step could not 
be replicated. Because the case-mix adjustment was applied at the 
facility level, the inability to replicate this step should not impact 
our estimate of the SLP percentage, as we expect the case-mix 
adjustment would affect the estimates of SLP and total therapy per diem 
costs to the same degree.
3. Methodology Used for the Calculation of Revised Federal Base Payment 
Rate Components
    As discussed above, we are considering separating the current 
therapy components into a PT/OT component and an SLP component. To do 
this, we considered calculating the percentage of the current therapy 
component of the federal base rate that corresponds to each of the two 
RCS-I components (PT/OT and SLP) in accordance with the methodology set 
forth below.
    The data described in section III.C.2. of this ANPRM provides cost 
estimates for the Medicare Part A SNF population for each cost report 
that met the inclusion criteria. Cost reports stratify costs by a 
number of cost centers that indicate different types of services. For 
instance, costs are reported separately for each of the three therapy 
disciplines (PT, OT, and SLP). Cost reports also include the number of 
Medicare Part A utilization days during the cost reporting period. This 
allows us to calculate both average SLP costs per day and average 
therapy costs per day in the facility during the cost reporting period. 
Therapy costs are defined as the sum of costs for the three therapy 
disciplines.
    The goal of this methodology is to estimate the fraction of therapy 
costs that corresponds to SLP costs. We use the facility-level averages 
developed from cost reports to derive a federal average for both 
therapy costs and SLP costs. To do this, we followed the methodology 
outlined in section II.A.3 of the 1998 interim final rule with comment 
period (63 FR 26260), which was used by CMS (then known as HCFA) to 
create the federal base payment rates:
    (1) For each of the two measures of cost (SLP costs per day and 
total therapy costs per day), we computed the mean based on data from 
freestanding SNFs only. This mean was weighted by the total number of 
Medicare days of the facility.
    (2) For each of the two measures of cost (SLP costs per day and 
total therapy costs per day), we computed the mean based on data from 
both hospital-based and freestanding SNFs. This mean was weighted by 
the total number of Medicare days of the facility.
    (3) For each of the two measures of cost (SLP costs per day and 
total therapy costs per day), we calculated the arithmetic mean of the 
amounts determined under steps (1) and (2) above.
    In section 3.11.3 of the SNF PMR Technical Report (available at 
https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html), we show the results of each of these 
calculations.
    The three steps outlined above produce a measure of SLP costs per 
day and a measure of therapy costs per day. We divided the SLP cost 
measure by the therapy cost measure to obtain the percentage of the 
therapy component that corresponds to SLP costs. We believe that 
following a methodology to derive the SLP percentage that is consistent 
with the methodology used to determine the base rates in the 1998 
interim final rule with comment period is appropriate because a 
consistent methodology helps to ensure that the resulting base rates 
for the components resemble what they would be had they been 
established in 1998 and that the methodology is as consistent as 
possible with the relevant statutory requirements, as discussed in 
section III.A.1 above. We found that 16 percent of the therapy 
component of the base rate for urban SNFs and 18 percent of the therapy 
component of the base rate for rural SNFs correspond to SLP costs. 
Under the RCS-I model we are considering, the current therapy case-mix 
component would be separated into a Physical Therapy/Occupational 
Therapy component and a Speech-Language Pathology component using the 
percentages derived above. This process is done separately for urban 
and for rural facilities. In section 3.11.3 of the SNF PMR Technical 
Report (available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html), we provide the specific 
cost centers used to identify SLP costs and total therapy costs.
    In addition, we are considering separating the current nursing 
case-mix component into a nursing case-mix component and an NTA 
component. Similar to the therapy component, we are considering 
calculating the percentage of the current nursing component of the 
federal base rates that corresponds to each of the two RCS-I components 
(NTA and nursing). The 1998 reopening of the comment period for the 
interim final rule (63 FR 65561, November 27, 1998) states that NTA 
costs comprise 43.4 percent of the current nursing component of the 
urban federal base rate, and the remaining 56.6 percent accounts for 
nursing and social services salary costs. These percentages for the 
nursing component of the federal base rate for rural facilities are 
42.7 percent and 57.3 percent, respectively (63 FR 65561). Therefore, 
we are considering assigning 43 percent of the current nursing 
component of the federal base rates to the new NTA

[[Page 20987]]

component of the federal base rate, and to assign the remaining 57 
percent to the new nursing component of the federal base rate.
    We verified the 1998 calculation of the percentages of the nursing 
component federal base rates that correspond to NTA costs by developing 
a measure of NTA costs per day for urban and rural facilities. We used 
the same data and followed the same methodology described above to 
develop measures of SLP costs per day and total therapy costs per day. 
The measure of NTA costs per day produced by this analysis is $47.70 
for urban facilities and $47.30 for rural facilities. The original 1998 
federal base rates for the nursing component, which relied on a similar 
methodology, were $109.48 for urban facilities and $104.88 for rural 
facilities. Therefore, our measure of NTA costs in urban facilities was 
equivalent to 43.6 percent of the urban 1998 federal nursing base rate, 
and our measure of NTA costs in rural facilities was equivalent to 45.1 
percent of the rural 1998 federal nursing base rate. These results are 
similar to the estimates published in the 1998 reopening of the comment 
period for the interim final rule (63 FR 65561, November 27, 1998), 
which we believe supports the validity of the 43 percent figure stated 
above.
    For illustration purposes, Tables 1 and 2 set forth what the 
unadjusted federal per diem rates would be for each of the case-mix 
adjusted components if we were to apply the RCS-I case-mix 
classification model to the proposed FY 2018 base rates (as set forth 
in the FY 2018 SNF PPS proposed rule. These are derived by dividing the 
proposed FY 2018 SNF PPS base rates according to the percentages 
described above. Tables 1 and 2 also show what the unadjusted federal 
per diem rates for the non-case-mix component would be, which are not 
affected by the change in case-mix methodology from the RUG-IV to the 
RCS-I. We use these unadjusted federal per diem rates in calculating 
the impact analysis discussed in section III.H. of this ANPRM.

                                                 Table 1--RCS-I Unadjusted Federal Rate Per Diem--Urban
--------------------------------------------------------------------------------------------------------------------------------------------------------
                           Rate component                                Nursing            NTA             PT/OT             SLP          Non-case-mix
--------------------------------------------------------------------------------------------------------------------------------------------------------
Per Diem Amount....................................................         $100.91           $76.12          $126.76           $24.14           $90.35
--------------------------------------------------------------------------------------------------------------------------------------------------------


                                                 Table 2--RCS-I Unadjusted Federal Rate Per Diem--Rural
--------------------------------------------------------------------------------------------------------------------------------------------------------
                           Rate component                                Nursing            NTA             PT/OT             SLP          Non-case-mix
--------------------------------------------------------------------------------------------------------------------------------------------------------
Per Diem Amount....................................................          $96.40           $72.72          $141.47           $31.06           $92.02
--------------------------------------------------------------------------------------------------------------------------------------------------------

    We invite comments on the data sources and methodology we are 
considering for calculating the unadjusted federal per diem rates and 
components that would be used in conjunction with the RCS-I case-mix 
classification model.
4. Updates and Wage Adjustments of Revised Federal Base Payment Rate 
Components
    In section III.B. of the FY 2017 SNF PPS final rule (81 FR 51972), 
we describe the process used to update the federal per diem rates each 
year. Additionally, as discussed in section III.B.4 of the FY 2017 SNF 
PPS final rule (81 FR 51978), SNF PPS rates are adjusted for geographic 
differences in wages using the most recent hospital wage index. Under 
the RCS-I case-mix model we are considering, we would continue to 
update the federal base payment rates and adjust for geographic 
differences in wages following the current methodology used for such 
updates and wage index adjustments under the SNF PPS. Specifically, 
under the RCS-I case-mix model, we would continue the practice of using 
the SNF market basket, adjusted as described in section III.B. of the 
FY 2017 SNF PPS final rule, and of adjusting for geographic differences 
in wages as described in section III.B.4 of the FY 2017 SNF PPS final 
rule. We invite comments on these ideas.

B. Potential Design and Methodology for Case-Mix Adjustment of Federal 
Rates

1. Background on Resident Classification System, Version I
    Section 1888(e)(4)(G)(i) of the Act requires that the Secretary 
provide for an appropriate adjustment to account for case mix and that 
such an adjustment shall be based on a resident classification system 
that accounts for the relative resource utilization of different 
patient types. The current case-mix classification system uses a 
combination of resident characteristics and service intensity metrics 
(for example, therapy minutes) to assign residents to one of 66 RUGs, 
each of which has a set of CMIs indicative of the relative cost to a 
SNF of treating residents within that classification category. However, 
as noted in section III.A. of this ANPRM, incorporating service-based 
metrics into the payment system can incentivize the provision of 
services based on a facility's financial considerations rather than 
resident needs. To better ensure that resident care decisions 
appropriately reflect each resident's actual care needs, we believe it 
is important to remove, to the extent possible, service-based metrics 
from the SNF PPS and derive payment from objective resident 
characteristics that are resident, and not facility, centered. To that 
end, RCS-I was developed to be a payment model which derives almost 
exclusively from verifiable resident characteristics.
    Additionally, the current RUG-IV case-mix classification system 
reduces the varied needs and characteristics of a resident into a 
single RUG-IV group that is used for payment. As of FY 2016, of the 66 
possible RUG classifications, over 90 percent of covered SNF PPS days 
are billed using one of the 23 Rehabilitation RUGs, with over 60 
percent of covered SNF PPS days billed using one of the three Ultra-
High Rehabilitation RUGs. The implication of this pattern is that more 
than half of the days billed under the SNF PPS effectively utilize only 
a resident's therapy minutes and Activities of Daily Living (ADL) score 
to determine the appropriate payment for all aspects of a resident's 
care. Both of these metrics, more notably a resident's therapy minutes, 
may derive not so much from the resident's own characteristics, but 
rather, from the type and amount of care the SNF decides to provide to 
the resident. Even assuming that the facility takes the resident's 
needs and unique characteristics into account in making these service 
decisions, the focus of payment remains centered, to a potentially 
great extent, on the facility's

[[Page 20988]]

own decision making and not on the resident's needs.
    While the RUG-IV model utilizes a host of service-based metrics 
(type and amount of care the SNF decides to provide) to classify the 
resident into a single RUG-IV group, the RCS-I model under 
consideration would separately identify and adjust for the varied needs 
and characteristics of a resident's care and then combine them 
together. We believe that the RCS-I classification model could improve 
the SNF PPS by basing payments predominantly on clinical 
characteristics rather than service provision, thereby enhancing 
payment accuracy and strengthening incentives for appropriate care.
2. Data Sources Utilized for Developing RCS-I
    To understand, research, and analyze the costs of providing Part A 
services to SNF residents, Acumen utilized a variety of data sources in 
the course of their research. In this section, we discuss these sources 
and how they were used in the SNF PMR in developing the RCS-I case-mix 
classification model. A more thorough discussion of the data sources 
used during the SNF PMR is available in section 3.1 of the SNF PMR 
Technical Report (available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html).
a. Medicare Enrollment Data
    Beneficiary enrollment and demographic information was pulled from 
the CMS enrollment database (EDB) and Common Medicare Environment 
(CME). Beneficiaries' Medicare enrollment was used to apply 
restrictions to create a study population for analysis. For example, 
beneficiaries were required to have continuous Medicare Part A 
enrollment during a stay. Demographic characteristics (for example, 
age) were incorporated as being predictive of resource use. 
Furthermore, enrollment and demographic information from these data 
sources were used to assess the impact of the RCS-I model under 
consideration on subpopulations of interest. In particular, the EDB and 
CME include indicators for potentially vulnerable subpopulations, such 
as those dually-enrolled in Medicaid.
b. Medicare Claims Data
    Medicare Parts A and B claims from the CMS Common Working Files 
(CWF) and Prescription Drug Event (PDE) claims from the PDE database 
were used to conduct claims analyses as part of the SNF PMR. The claims 
data analyzed derived from SNF claims. SNF claims (CMS-1450 form, OMB 
control number 0938-0997), including type of bill (TOB) 21x (SNF 
Inpatient Part A) and 18x (hospital swing bed), were used to identify 
Medicare Part A stays paid under the SNF PPS. Part A stays were 
constructed by linking claims that share the same beneficiary 
identifier, facility CMS Certification Number (CCN), and admission 
date. Information from the claims, such as RUGs, diagnoses, and 
assessment dates, were aggregated across a stay. Stays created from SNF 
claims were linked to other claims data and assessment data via 
beneficiary identifiers.
    Acute care hospital stays that qualified the beneficiary for the 
SNF benefit were identified using Medicare inpatient hospital claims. 
More specifically, the dates of the qualifying hospital stay listed in 
the span codes of the SNF claim were used, connecting inpatient claims 
with those dates listed as the admission and discharge dates. Although 
there are exceptions, the claims from the preceding inpatient 
hospitalization commonly contain clinical and service information 
relevant to the care administered during a SNF stay. Components of this 
information were used in the regression models predicting therapy and 
NTA costs or to better understand patterns of post-acute care referrals 
for patients requiring SNF services. Additionally, the most recent 
hospital stay was matched to the SNF stay, which often (though not 
always) was the same as the preceding inpatient hospitalization, and 
used in the regression models.
    Other Medicare claims, including outpatient hospital, physician, 
home health, hospice, durable medical equipment, and drug 
prescriptions, were incorporated, as necessary, into the analysis in 
one of three ways: (i) To verify information found on assessment and 
SNF or inpatient claims data; (ii) to provide additional resident 
characteristics to test outside of those found in assessment and SNF 
and inpatient claims data; and (iii) to stratify modeling results to 
identify effects of the system on beneficiary subpopulations. These 
claims were linked to SNF claims using beneficiary identifiers.
c. Assessment Data
    MDS assessments were the primary source of resident characteristics 
used to explain service use and payment in the SNF setting. Acumen's 
data repositories include MDS assessments submitted by SNFs and swing-
bed hospitals. MDS version 2.0 assessments were submitted until October 
2010, at which point MDS version 3.0 assessments began. MDS data were 
extracted from the Quality Improvement Evaluation System (QIES). MDS 
assessments were then matched to SNF claims data using the beneficiary 
identifier, assessment indicator, assessment date, and Resource 
Utilization Group (RUG).
    The SNF PMR also used assessment data not available in the SNF 
setting. Data from the IRF Patient Assessment Instrument (IRF-PAI) and 
Outcome and Assessment Information Set (OASIS) were used to identify 
characteristics that are predictive of service use and costs in the IRF 
and home health settings, to consider potential similarities with 
service use in the SNF setting. IRF-PAI and OASIS include assessments 
for all Medicare IRF and home health patients, regardless of fee-for-
service or Medicare Advantage enrollment. While the care furnished in 
the IRF and home health settings may differ from that furnished in a 
SNF, there are similarities in the patient populations across PAC 
settings. IRF-PAI and OASIS data were used for exploratory analyses but 
were not used to develop RCS-I payment components.
d. Facility Data
    Facility characteristics, while not considered as explanatory 
variables when modeling service use, were used for impact analyses. By 
incorporating this facility-level information, we could identify any 
disproportionate effects of the new case-mix classification system on 
different types of facilities.
    Facility-level characteristics were taken from the Certification 
and Survey Provider Enhanced Reports (CASPER). From CASPER, we draw 
facility-level characteristics such as ownership, chain affiliation, 
facility size, and staffing levels. CASPER data were supplemented with 
information from publicly available data sources. The principal data 
sources that are publicly available include the Medicare Cost Reports 
(Form 2540-10, 2540-96, and 2540-92) extracted from the Healthcare Cost 
Report Information System (HCRIS) files, Provider-Specific Files (PSF), 
Provider of Service files (POS), and Nursing Home Compare (NHC). These 
data sources have information on facility costs and payment and 
characteristics that directly affect PPS calculations.
3. Resident Classification Under RCS-I
a. Background
    As noted above, section 1888(e)(4)(G)(i) of the Act requires that 
the Secretary provide for an appropriate adjustment to account for case 
mix and that such an adjustment shall be based

[[Page 20989]]

on a resident classification system that accounts for the relative 
resource utilization of different patient types. RCS-I was developed to 
be a model of payment which derives almost exclusively from resident 
characteristics. More specifically, the RCS-I model under consideration 
separately identifies and adjusts four different case-mix components 
for the varied needs and characteristics of a resident's care and then 
combines these together with the non-case-mix component to form the 
full SNF PPS per diem rate for that resident.
    As with any case-mix classification system, the predictors that 
were found to be part of case-mix classification under RCS-I are those 
which our analysis associated with variation in the costs for the given 
case-mix component. The federal per diem rates discussed above serve as 
``base rates'' specifically because they set the basic average cost of 
treating a typical SNF resident. Based on the presence of certain needs 
or characteristics, caring for certain residents may cost more or less 
than that average cost. A case-mix system identifies certain aspects of 
a resident or of a resident's care which, when present, lead to average 
costs for that group being higher or lower than the average cost of 
treating a typical SNF resident. For example, if we found that therapy 
costs were the same for two residents regardless of having a particular 
condition, then that condition would not be relevant in predicting 
increases in therapy costs. If, however, we found that, holding all 
else constant, the presence of a given condition was correlated with an 
increase in therapy costs for residents with that condition over those 
without that condition, then this could mean that this condition is 
indicative, or predictive, of increased costs relative to the average 
cost of treating SNF residents generally.
    In the subsections that follow, we describe each of the four case-
mix adjusted components under the RCS-I classification model we are 
considering, and the basis for each of the predictors that would be 
used within the RCS-I model to classify residents for payment purposes. 
In the final subsection under this section of the ANPRM, we outline two 
hypothetical payment scenarios utilizing the same set of resident 
characteristics, one using the existing RUG-IV classification model and 
one using the RCS-I classification model, to demonstrate the increased 
flexibility and resident-focused approach of the RCS-I model.
b. Physical and Occupational Therapy Case-Mix Classification
    A fundamental aspect of the RCS-I case-mix classification model is 
to use resident characteristics to predict the costs of furnishing 
similarly situated residents with SNF care. Costs derived from the 
charges on claims and CCRs on facility cost reports were used as the 
measure of resource use to develop the RCS-I system. Costs better 
reflect differences in the relative resource use of residents as 
opposed to charges, which partly reflect decisions made by providers 
about how much to charge payers for certain services. Costs derived 
from charges are reflective of therapy utilization as they are 
correlated to therapy minutes recorded for each therapy discipline. 
Under the current RUG-IV case-mix model, therapy minutes for all three 
therapy disciplines (physical therapy (PT), occupational therapy (OT), 
and speech-language pathology (SLP)) are added together to determine 
the appropriate case-mix classification for the resident. However, when 
we began to investigate resident characteristics predictive of therapy 
costs for each therapy discipline, summary statistics revealed that 
there exists little correlation between PT and OT costs per day with 
SLP costs per day (correlation coefficient of 0.04). The set of 
resident characteristics from the MDS that predicted PT and OT 
utilization was different than the set of characteristics predicting 
SLP utilization. Additionally, many predictors of high PT and OT costs 
per day predicted lower SLP costs per day, and vice versa. For example, 
residents with cognitive impairments receive less physical and 
occupational therapy but receive more speech-language pathology. As a 
result of this analysis, we found that isolating predictors of total 
therapy costs per day obscured differences in the determinants of PT/OT 
and SLP utilization.
    In contrast, the correlation coefficient between PT and OT costs 
per day was high (0.62), and regression analyses found that predictors 
of high PT costs per day were also predictive of high OT costs per day. 
For example, the analyses found that late-loss ADLs are strong 
predictors of both PT and OT costs per day. Acumen then ran regression 
analyses of a range of resident characteristics on PT and OT costs per 
day separately and found that the coefficients in both models followed 
similar patterns. Finally, resident characteristics were found to be 
better predictors of the sum of PT and OT costs per day than for either 
PT or OT costs separately. These analyses used a variety of variables 
from the MDS, as well as PT, OT, and SLP costs per day. More 
information on these analyses can be found in section 3.3.1 of the SNF 
PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.
    Given the results of this analytic work, we are considering 
combining PT and OT costs under a single case-mix adjusted component, 
while addressing SLP costs through a separate case-mix adjusted 
component. The next step in our analysis was to identify resident 
characteristics that were best predictive of PT/OT costs per day. To 
accomplish this, we conducted cost regressions with a host of variables 
from the MDS assessment, the prior inpatient claims, and the SNF claims 
that may have been predictive of relative increases in PT/OT costs. The 
variables were selected with the goal of being as inclusive as possible 
of the characteristics recorded on the MDS assessment, and also 
included information from the prior inpatient stay. The selection also 
incorporated clinical input. These initial costs regressions were 
exploratory and meant to identify a broad set of resident 
characteristics that are predictive of PT/OT resource utilization. The 
results were used to inform which variables should be investigated 
further and ultimately included in the payment system. A table of all 
of the variables considered as part of this analysis appears in the 
Appendix of the SNF PMR Technical Report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. Based on our regression analyses, we found that 
the three most relevant predictors of PT/OT costs per day were the 
clinical reasons for the SNF stay, the resident's functional status, 
and the presence of a cognitive impairment. More information on this 
analysis can be found in section 3.4.1 of the SNF PMR technical report 
available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.
    Under the RUG-IV case-mix model, residents are first categorized 
based on being a rehabilitation resident or a non-rehabilitation 
resident, and then categorized further based on additional aspects of 
the resident's care. Under the RCS-I case-mix model, for the purposes 
of determining the resident's PT/OT group and, as will be discussed 
below, the resident's SLP group, the resident is first categorized 
based on the clinical reasons for the resident's SNF stay. Empirical 
analyses demonstrated that the clinical basis for the resident's stay

[[Page 20990]]

(that is, the primary reason the resident is in the SNF) proved a 
strong predictor of therapy costs. More detail on these analyses can be 
found in section 3.4.1 of the SNF PMR Technical Report. In consultation 
with stakeholders (industry representatives, beneficiary 
representatives, clinicians, and payment policy experts) at multiple 
technical expert panels (TEPs), we created a set of ten inpatient 
clinical categories that we believe capture the range of general 
resident types which may be found in a SNF. These clinical categories 
are provided in Table 3.

                      Table 3--Clinical Categories
------------------------------------------------------------------------
 
------------------------------------------------------------------------
Major Joint Replacement or Spinal Surgery.  Cancer.
Non-Surgical Orthopedic/Musculoskeletal...  Pulmonary.
Orthopedic Surgery (Except Major Joint)...  Cardiovascular and
                                             Coagulations.
Acute Infections..........................  Acute Neurologic.
Medical Management........................  Non-Orthopedic Surgery.
------------------------------------------------------------------------

    Once we identified these clinical categories as being generally 
predictive of resource utilization in a SNF, we then undertook the 
necessary work to identify those categories predictive of PT/OT costs 
specifically. We conducted additional regression analyses to determine 
if any of these categories predicted similar levels of PT/OT as other 
categories, which may provide a basis for combining categories together 
where similar resident costs were predicted. As a result of this 
analysis, we found that the ten inpatient clinical categories could be 
collapsed into five clinical categories, which predict varying degrees 
of PT/OT costs. Acute infections, cancer, pulmonary, cardiovascular and 
coagulations, and medical management were collapsed into one clinical 
category entitled ``Medical Management'' because their residents had 
similar PT/OT costs. Similarly, orthopedic surgery (except major joint) 
and non-surgical orthopedic/musculoskeletal were collapsed into a new 
``Other Orthopedic'' category for equivalent reasons. The remaining 
three categories (Acute Neurologic, Non-Orthopedic Surgery, and Major 
Joint Replacement or Spinal Surgery) showed distinct PT/OT cost 
profiles and were thus retained as independent categories. More 
information on this analysis can be found in section 3.4.2 of the SNF 
PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. These 
collapsed categories, which would be used to categorize a resident 
initially under the PT/OT case-mix component, are presented in Table 4.

                   Table 4--PT/OT Clinical Categories
------------------------------------------------------------------------
 
-------------------------------------------------------------------------
Major Joint Replacement or Spinal Surgery.
Other Orthopedic.
Non-Orthopedic Surgery.
Acute Neurologic.
Medical Management.
------------------------------------------------------------------------

    With regard to operationalizing this categorization, we are 
considering using item I8000 on the MDS 3.0 to allow providers to 
report the resident's primary diagnosis. More specifically, the first 
line in item I8000 would be used by providers to report the ICD-10-CM 
code which represents the primary reason for the resident's SNF Part A 
stay.
    In addition to the resident's initial clinical categorization, as 
discussed previously in this section, regression analyses demonstrated 
that the resident's functional status is also predictive of PT/OT 
costs. However, the existing ADL scale used to classify residents into 
a RUG-IV group captures little variation in PT/OT costs, though this is 
unsurprising as the existing ADL scale was never intended for this 
purpose. Therefore, we found it appropriate to consider revisions to 
the ADL scale used to categorize the functional status of residents 
under the PT/OT component in a manner that is predictive of PT/OT 
costs.
    Under the RUG-IV case-mix system, a resident's ADL or functional 
score is calculated based on a combination of self-performance and 
support items coded by SNFs in Section G of the MDS 3.0 for four ADL 
areas: Transfers; eating; toileting; and bed mobility. Each ADL may be 
scored for four points, with a potential total score as high as 16 
points. Under the RCS-I case-mix model, a resident would be 
categorized, as it pertains to function, using only three of these ADL 
areas, specifically transfers, eating, and toileting. We removed bed 
mobility from this list, based on feedback we received from clinicians 
working on the research project and verified through presentation to 
stakeholders during our TEPs, that bed mobility depends partly on the 
type of bed, and therefore it is likely confounded by facility 
procedures, rather than exclusively providing information about the 
resident's function. Therefore, to help eliminate potential 
determinants of a resident's functional level which may be related to 
facility decisions on support provided to a resident regardless of 
need, we believe it would be more appropriate to focus on those ADL 
areas which are most relevant to the resident's actual capabilities and 
needs. To this end, the functional score used as part of the RCS-I 
case-mix model for purposes of categorizing residents under the PT/OT 
case-mix component would only use the self-performance items for these 
three ADL areas and ignore the support items coded for these areas. We 
believe that the self-performance items are a closer reflection of the 
resident's ability to perform a task, while the support items are more 
descriptive of the staff's practices and level of effort, which may not 
be consistent across facilities. We believe that the self-performance 
items better represent the actual needs of the resident, while the 
support items represent facility resource decisions. Therefore, we 
believe that a resident's ADL score, which would be used to categorize 
a resident under RCS-I's PT/OT case-mix component, should be based on 
only the self-performance items for the transfer, eating, and toileting 
areas in Section G of the MDS 3.0.
    In addition to these changes, we also are considering that, for 
purposes of classifying a resident under RCS-I's PT/OT case-mix 
component, each of these ADL areas would be scored for a total of 6 
points, rather than the current 4 points under the RUG-IV model, where 
the number of points increases with predicted increases in the 
resident's PT/OT costs. Using 6 points would allow us to consider the 
impact on PT/OT costs for each of the 6 possible performance levels in 
the ADL self-performance items. Under the RUG-IV model, if the SNF 
codes that the ``activity did not occur'' or ``occurred only once'', 
then these items are ignored for purposes of categorizing the resident 
for ADL purposes. However, cost regressions revealed that these two 
codes can predict lower costs for PT/OT services, which we believe is 
an important aspect of generally predicting PT/OT costs. Therefore, 
these two codes would be incorporated into the scoring for a resident's 
ADL score under the PT/OT component of the RCS-I case-mix model. In 
Table 5, we provide the scoring algorithm used for each of the three 
ADL areas and how many points would be scored for each potential 
response for each area. We determined the ADL scoring scale by first 
testing the relationship between each possible response to the three 
selected ADL items and PT/OT costs per day. This investigation revealed 
that therapy costs

[[Page 20991]]

first increase, then decrease with increasing dependence on the 
transfer and toileting items. Residents who require assistance to 
perform these ADLs tend to have higher PT/OT costs than both residents 
who are completely independent and residents who are completely 
dependent. However, costs consistently decrease with increasing 
dependence on the eating item. The points are assigned to each possible 
response to the three selected ADL items based on the observed cost 
patterns. As Table 5 shows, the points assigned to each response mirror 
the inverse U-shape of the dependence-cost curve for the transfer and 
toileting items and the monotonic decrease in costs associated with 
increasing dependence on the eating item. This produces a functional 
score that ranges from 0 to 18. As opposed to the ADL score used in 
RUG-IV, the functional score has a linear relationship with PT/OT 
costs: As the score increases, PT/OT costs per day also increase. In 
section 3.4.1 of the SNF PMR Technical report, we provide additional 
information on the analyses that led to the construction of this ADL 
score.

                                        Table 5--PT/OT ADL Scoring Scale
----------------------------------------------------------------------------------------------------------------
                   ADL self-performance score                        Transfer        Toileting        Eating
----------------------------------------------------------------------------------------------------------------
Independent.....................................................              +3              +3              +6
Supervision.....................................................              +4              +4              +5
Limited Assistance..............................................              +6              +6              +4
Extensive Assistance............................................              +5              +5              +3
Total Dependence................................................              +2              +2              +2
Activity Occurred only Once or Twice............................              +1              +1              +1
Activity did not Occur..........................................              +0              +0              +0
----------------------------------------------------------------------------------------------------------------

    The final aspect of categorizing a resident under the PT/OT 
component of the RCS-I case-mix model is related to the resident's 
cognitive status. Currently under the SNF PPS, cognitive status is used 
to classify a small portion of residents that fall into the Behavioral 
Symptoms and Cognitive Performance RUG-IV category. For all other 
residents, cognitive status is not used in determining the appropriate 
payment for a resident's care. However, industry representatives and 
clinicians at multiple TEPs suggested that a resident's cognitive 
status can have a significant impact on a resident's predicted PT/OT 
costs. This was reinforced by empirical analyses conducted by Acumen. 
Sections 3.3.1, 3.4.1, and 3.4.2 of the SNF PMR Technical report 
contains more information on these analyses (available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html). Therefore, we believe that a resident's 
cognitive status should be considered as a predictor of PT/OT costs.
    Under the RUG-IV model, cognitive status is assessed using the 
Brief Interview for Mental Status (BIMS) on the MDS 3.0. The BIMS is 
based on three items: ``Repetition of three words;'' ``temporal 
orientation;'' and ``recall.'' The sum of these numbers is the BIMS 
summary score. The BIMS score is from 0 to 15, with 0 assigned to 
residents with the worst cognitive performance and 15 assigned to 
residents with the highest performance. Residents with a BIMS score 
less than or equal to 9 classify for the Behavioral Symptoms and 
Cognitive Performance category.
    However, in approximately 15 percent of 5-day MDS assessments, a 
BIMS is not completed: In 12 percent of cases the interview is not 
attempted, and for 3 percent of cases the interview is attempted but 
cannot be completed. The MDS directs assessors to skip the BIMS if the 
resident is rarely or never understood (this is scored as ``skipped''). 
In these cases, the MDS requires assessors to complete the Staff 
Assessment for Mental Status (items C0700-C1000). The Cognitive 
Performance Scale (CPS) is used to assess cognitive function based on 
the Staff Assessment for Mental Status. The Staff Assessment for Mental 
Status consists of four items: ``Short-term Memory OK,'' ``Long-term 
Memory OK,'' ``Memory/Recall Ability,'' and ``Cognitive Skills for 
Daily Decision Making.'' However, only ``Short-term Memory OK'' and 
``Cognitive Skills for Daily Decision Making'' are currently used for 
payment. In MDS 2.0, the CPS was used as the sole measure of cognitive 
status. A resident was assigned a CPS score from 0 to 6 based on 
responses to several items on the MDS, with 0 indicating the resident 
was cognitively intact and 6 indicating the highest level of cognitive 
impairment. Any score of 3 or above was considered cognitively 
impaired. The CPS on the current version of the MDS (3.0) functions 
very similarly. Instead of assigning a score to each resident, a 
resident is determined to be cognitively impaired if he or she meets 
the criteria to receive a score of 3 or above on the CPS. Residents who 
meet this criteria are classified in the Behavioral Symptoms and 
Cognitive Performance category under RUG-IV, if they do not meet the 
criteria for a higher-paying category.
    Given that the 15 percent of residents who are not assessed on the 
BIMS must be assessed using a different scale that relies on a 
different set of MDS items, there is currently no single measure of 
cognitive status that allows comparability across all residents. To 
address this issue, Thomas et al., in a 2015 paper, proposed use of a 
new cognitive measure, the Cognitive Function Scale (CFS), which 
combines scores from the BIMS and CPS into one scale that can be used 
to compare cognitive function across all residents (Thomas KS, Dosa D, 
Wysocki A, Mor V; The Minimum Data Set 3.0 Cognitive Function Scale. 
Med Care. https://www.ncbi.nlm.nih.gov/pubmed/?term=25763665). 
Following a suggestion from the June 2016 TEP, we explored using the 
CFS as a measure of cognition, and found that there is a relationship 
between the different levels of the cognitive scale and resident costs. 
More information on this analysis can be found in section 3.4.1 of the 
SNF PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. 
Therefore, we are considering using the CFS as a cognitive measure in 
the RCS-I system. The RUG-IV system also incorporates both the BIMS and 
CPS score, but the CFS blends them together into one measure of 
cognitive status. Details on how the BIMS score and CPS score are 
determined using the MDS assessment are described above. The CFS places 
residents into one of four cognitive performance categories based on 
their score on either the BIMS or CPS, as shown in Table 6.

[[Page 20992]]



                 Table 6--CFS Classification Methodology
------------------------------------------------------------------------
                                                          BIMS     CPS
                  CFS cognitive scale                    score    score
------------------------------------------------------------------------
Cognitively Intact....................................    13-15  .......
Mildly Impaired.......................................     8-12      0-2
Moderately Impaired...................................      0-7      3-4
Severely Impaired.....................................  .......      5-6
------------------------------------------------------------------------

    Once each of these variables--clinical reasons for the SNF stay, 
the resident's functional status, and the presence of a cognitive 
impairment--in predicting resident PT/OT costs was identified, we then 
used a statistical regression technique called the Classification and 
Regression Tree (CART) to determine the most appropriate splits in 
resident PT/OT case-mix groups using these three variables. In other 
words, CART was used to determine how many PT/OT case-mix groups should 
exist under the RCS-I model under consideration and what types of 
residents or score ranges should be combined to form each of those PT/
OT case-mix groups. CART is a non-parametric decision tree learning 
technique that produces either classification or regression trees, 
depending on whether the dependent variable is categorical or numeric, 
respectively. Using the CART technique to create payment groups is 
advantageous because it is both immune to outliers and resistant to 
irrelevant parameters. The CART was used to create payment groups in 
other Medicare settings. For example, it determined Case Mix Groups 
(CMGs) splits within rehabilitation impairment groups (RICs) when the 
inpatient rehabilitation facilities (IRF) PPS was developed. This 
methodology is more thoroughly explained in section 3.4.2 of the SNF 
PMR Technical Report (available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html).
    Based on the CART algorithm, we determined that 30 case-mix groups 
would be necessary to classify residents adequately in terms of their 
PT/OT costs, in a manner that captures sufficient variation in PT/OT 
costs without creating unnecessarily granular separations. In addition, 
the PT/OT case-mix groups also reflect certain administrative decisions 
made by our project team. For example, while CART may have created 
different breakpoints for the functional score in different clinical 
categories, we believed that using a consistent split in scores across 
clinical categories would improve the simplicity of the case-mix model 
without compromising its accuracy. Therefore, we used the splits 
created by the CART algorithm as the basis for the consistent splits 
selected for the case-mix groups, simplifying the CART output while 
retaining important features of the CART-generated splits. 
Characteristics such as age, which CART did not select as an important 
criterion for classifying residents, were dropped, while splits that 
recurred across clinical categories, such as dividing residents into 
cognitively intact (CFS=1,2) and cognitively impaired (CFS=3,4) were 
retained. To confirm that the consistent splits approach did not 
require a notable sacrifice in payment accuracy, we used regression 
analysis to test the ability of the CART-generated splits and the 
consistent splits to predict PT/OT costs per day. We found that using 
the consistent splits resulted in only a minor reduction in predictive 
ability (a decrease of 0.004 in the R-squared). Section 3.4.2 of the 
SNF PMR Technical Report contains more details on these analyses 
(available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html).
    We provide the criteria for each of these groups, along with the 
CMI for each group, in Table 7. As shown in the table, three factors 
are used to classify each resident for PT/OT payment: Clinical 
category, function score, and the presence of moderate or severe 
cognitive impairment. Each case-mix group corresponds to one clinical 
category, one function score range, and the presence or absence of 
moderate/severe cognitive impairment. Based on these three factors, we 
are considering classifying a resident into one of the 30 groups shown 
in Table 7.
    To help ensure that payment reflects the average relative resource 
use at the per diem level, CMIs would be set to reflect relative case-
mix related differences in costs across groups. CMIs for the PT/OT 
component would be calculated based on two factors. One factor is the 
average per diem costs of a case-mix group relative to the population 
average. Relative differences in costs due to different length of stay 
distribution across groups are removed from this calculation (as 
further discussed in the description of variable per diem payments in 
section III.D.4 of this ANPRM). The other factor is the average 
variable per diem adjustment factor of the group relative to the 
population average. In this calculation, average per diem costs equal 
total PT/OT costs in the group divided by number of utilization days in 
the group, and similarly the average variable per diem adjustment 
factor equals the sum of PT/OT variable per diem adjustment factors for 
all utilization days in the group divided by the number of utilization 
days. More information on the variable per diem adjustment factor is 
discussed in section III.D.4 of this ANPRM. This method would help 
ensure that the share of payment for each case-mix group is equal to 
its share of total costs of the component. The full methodology used to 
develop CMIs is presented in section 3.12 of the SNF PMR Technical 
Report is available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.

                                  Table 7--PT/OT Case-Mix Classification Groups
----------------------------------------------------------------------------------------------------------------
                                                Moderate/severe cognitive
      Clinical category       Function score           impairment              Case-mix group     Case-mix index
----------------------------------------------------------------------------------------------------------------
Major Joint Replacement or             14-18  No..........................  TA                              1.82
 Spinal Surgery.
                                       14-18  Yes.........................  TB                              1.59
                                        8-13  No..........................  TC                              1.73
                                        8-13  Yes.........................  TD                              1.45
                                         0-7  No..........................  TE                              1.68
                                         0-7  Yes.........................  TF                              1.36
Other Orthopedic............           14-18  No..........................  TG                              1.70
                                       14-18  Yes.........................  TH                              1.55
                                        8-13  No..........................  TI                              1.58
                                        8-13  Yes.........................  TJ                              1.39
                                         0-7  No..........................  TK                              1.38
                                         0-7  Yes.........................  TL                              1.14

[[Page 20993]]

 
Acute Neurologic............           14-18  No..........................  TM                              1.61
                                       14-18  Yes.........................  TN                              1.48
                                        8-13  No..........................  TO                              1.52
                                        8-13  Yes.........................  TP                              1.36
                                         0-7  No..........................  TQ                              1.47
                                         0-7  Yes.........................  TR                              1.17
Non-Orthopedic Surgery......           14-18  No..........................  TS                              1.57
                                       14-18  Yes.........................  TT                              1.43
                                        8-13  No..........................  TU                              1.38
                                        8-13  Yes.........................  TV                              1.17
                                         0-7  No..........................  TW                              1.11
                                         0-7  Yes.........................  TX                              0.80
Medical Management..........           14-18  No..........................  T1                              1.55
                                       14-18  Yes.........................  T2                              1.39
                                        8-13  No..........................  T3                              1.36
                                        8-13  Yes.........................  T4                              1.17
                                         0-7  No..........................  T5                              1.10
                                         0-7  Yes.........................  T6                              0.82
----------------------------------------------------------------------------------------------------------------

    Under the RCS-I case-mix model, all residents would be classified 
into one, and only one, of these 30 PT/OT case-mix groups. As opposed 
to the RUG-IV system that determines therapy payments based only on the 
amount of therapy provided, these groups classify residents based on 
three resident characteristics shown to be predictive of PT/OT 
utilization. Thus, we believe that the PT/OT case-mix groups would 
provide a better measure of resource use and would provide for more 
appropriate payment under the SNF PPS. We invite comments on the series 
of ideas and the approach we are considering above associated with the 
PT/OT component of the RCS-I case-mix model.
c. Speech-Language Pathology Case-Mix Classification
    As discussed above, many of the resident characteristics which we 
found to be predictive of increased PT/OT costs were predictive of 
lower SLP costs. As a result of this inverse relationship, using the 
same set of predictors to case-mix adjust a single therapy component 
would obscure important differences in predicting relative differences 
in resident therapy costs and make any predictive model that attempts 
to predict total therapy cost inherently less accurate. Therefore, we 
believe it is appropriate to have a separately adjusted case-mix SLP 
component that is specifically designed to predict relative differences 
in SLP costs. As discussed in the prior section, costs derived from the 
charges on claims and CCRs on facility cost reports were used as the 
measure of resource use to develop an alternative payment system. Costs 
are reflective of therapy utilization as they are correlated to therapy 
minutes recorded for each therapy discipline.
    Following the same methodology we used to identify predictors of 
PT/OT costs, our project team conducted cost regressions with a host of 
variables from the MDS assessment, prior inpatient claims, and SNF 
claims that were identified as likely to be predictive of relative 
increases in SLP costs. The variables were selected with the goal of 
being as inclusive of the measures recorded on the MDS assessment as 
possible, and also included information from the prior inpatient stay. 
The selection also incorporated clinical input from TEP panelists, 
Acumen clinical staff, and CMS clinical staff. These initial costs 
regressions were exploratory and meant to identify a broad set of 
resident characteristics that are predictive of SLP resource 
utilization. The results were used to inform which variables should be 
investigated further and ultimately included in the payment system. A 
table of all of the variables considered in this analysis appears in 
the Appendix of the SNF PMR Technical Report. Based on these cost 
regressions, we identified a set of three categories of predictors 
relevant in predicting relative differences in SLP costs: Clinical 
reasons for the SNF stay, presence of a swallowing disorder or 
mechanically-altered diet, and the presence of an SLP-related 
comorbidity or cognitive impairment. A model using these predictors to 
predict SLP costs per day accounted for 14.5 percent of the variation 
in costs, while a very extensive model using 1,016 resident 
characteristics only predicted 19.3 percent of the variation. This 
shows that these predictors alone explain a large share of the 
variation in SLP costs per day that can be explained with resident 
characteristics. More information on this analysis can be found in 
section 3.5.1 of the SNF PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.
    As with the PT/OT component, we began with the set of clinical 
categories identified in Table 3 (meant to capture general differences 
in resident resource utilization) and ran cost regressions to determine 
which categories may be predictive of generally higher relative SLP 
costs. Through this analysis, we found that one clinical group was 
particularly predictive of increased SLP cost, which was the Acute 
Neurologic group. More detail on this investigation can be found in 
section 3.5.2 of the SNF PMR Technical Report. Therefore, to determine 
the initial resident classification into an SLP group under the RCS-I, 
residents would first be categorized, using the clinical reasons for 
the resident's SNF stay recorded on the first line of Item I8000 on the 
MDS assessment, into one of two groups, either the ``Acute Neurologic'' 
clinical category, or into a Non-Neurologic group that includes the 
remaining clinical categories found in Table 3: Major Joint Replacement 
or Spinal Surgery; Non-Surgical Orthopedic/Musculoskeletal; Orthopedic 
Surgery (Except Major Joint); Acute Infections, Cancer, Pulmonary; Non-
Orthopedic Surgery; Cardiovascular and Coagulations; and Medical 
Management.

[[Page 20994]]

    In addition to the clinical reason for the SNF stay, cost 
regressions and TEP members also identified the presence of a 
swallowing disorder or a mechanically-altered diet (which refers to 
food that has been altered to make it easier for the resident to chew 
and swallow to address a specific resident need), as a predictor of 
relative increases in SLP costs. First, residents who exhibited the 
signs and symptoms of a swallowing disorder, as identified using K0100Z 
on the MDS 3.0, demonstrated significantly higher SLP costs than those 
who did not exhibit such signs and symptoms. Therefore, we considered 
including the presence of a swallowing disorder as a component in 
predicting SLP costs. However, when this information was presented 
during the October 2016 TEP, stakeholders indicated that the signs and 
symptoms of a swallowing disorder may not be as readily observed when a 
resident is on a mechanically-altered diet, and requested that we also 
consider evaluating the presence of a mechanically-altered diet, as 
determined by item K0510C2 on the MDS 3.0, as an additional predictor 
of increased SLP costs. Our project team conducted this analysis and 
found that there was an associated increase in SLP costs when a 
mechanically-altered diet was present. Moreover, this analysis revealed 
that while SLP costs may increase when either a swallowing disorder or 
mechanically-altered diet is present, resident SLP costs increased even 
more when both of these items were present. More detail on this 
investigation and these analyses can be found in section 3.5.1 of the 
SNF PMR Technical Report. As a result, we agree with the stakeholders 
that including a mechanically-altered diet would be an important 
component of predicting relative increases in resident SLP costs, and 
thus, in addition to the clinical categorization, we are considering 
classifying residents as having either a swallowing disorder, being on 
a mechanically altered diet, both, or neither for purposes of 
classifying the resident under the SLP component.
    As a final aspect of the SLP component case-mix adjustment, we 
found that the presence of a cognitive impairment or SLP-related 
comorbidity affected relative differences in SLP costs. More 
specifically, we found that the presence of certain SLP-related 
comorbidities or the presence of a mild to severe cognitive impairment 
(as defined by the CFS methodology described in Table 6 in section 
III.D.3.b. of this ANPRM) was correlated with relative increases in SLP 
costs. For each condition or service included as an SLP-related 
comorbidity, the presence of the condition or service was associated 
with at least a 43 percent increase in average SLP costs per day. The 
presence of a mild to severe cognitive impairment was associated with 
at least a 100 percent increase in average SLP costs per day. Similar 
to the analysis conducted in relation to the PT/OT component, the 
project team ran cost regressions on a broad list of possible 
conditions, with that list being available in section 3.5.1 of the SNF 
PMR Technical Report (available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html). Based on 
that analysis, and in consultation with stakeholders during our TEPs 
and clinicians, we have identified the conditions listed in Table 8 to 
be those SLP-related comorbidities which we believe would best serve to 
predict relative differences in SLP costs. Acumen used diagnosis codes 
on the most recent inpatient claim for each SNF stay and the SNF claim 
to identify these diagnoses and found that residents with these 
conditions had much higher SLP costs per day. More detail on these 
analyses can be found in section 3.5.1 of the SNF PMR Technical Report 
available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.

                   Table 8--SLP-Related Comorbidities
------------------------------------------------------------------------
 
------------------------------------------------------------------------
Aphasia...................................  Laryngeal Cancer.
CVA, TIA, or Stroke.......................  Apraxia.
Hemiplegia or Hemiparesis.................  Dysphagia.
Traumatic Brain Injury....................  ALS.
Tracheostomy (while Resident).............  Oral Cancers.
Ventilator (while Resident)...............  Speech and Language
                                             Deficits.
------------------------------------------------------------------------

    Once each of these variables--clinical reasons for the SNF stay, 
presence of a swallowing disorder or mechanically-altered diet, and the 
presence of an SLP-related comorbidity or cognitive impairment--found 
to be useful in predicting resident SLP costs was identified, we then 
used the CART algorithm, as we discussed above in relation to the PT/OT 
component, to determine the most appropriate splits in resident SLP 
case-mix groups using these three variables. This methodology and the 
results of our analysis are more thoroughly explained in sections 3.4.2 
and 3.5.2 of the SNF PMR Technical Report. Based on the CART algorithm, 
we determined that 18 case-mix groups would be necessary to classify 
residents adequately in terms of their SLP costs, in a manner that 
captures sufficient variation in SLP costs without creating 
unnecessarily granular separations. The accuracy of this model was 
confirmed by comparing the ability of the CART model and various 
consistent split models to predict SLP costs per day. More information 
on this analysis can be found in section 3.5.2 of the SNF PMR technical 
report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. We provide the criteria 
for each of these groups, along with the CMI for each group, in Table 
9.
    To help ensure that payments reflect the average relative resource 
use at the per diem level, CMIs would be set to reflect case-mix 
related relative differences in costs across groups. CMIs for the SLP 
component would be calculated based on the average per diem costs of a 
case-mix group relative to the population average. Relative differences 
in costs due to different length of stay distribution across groups are 
removed from the calculation. In this calculation, average per diem 
costs equal total SLP costs in the group divided by number of 
utilization days in the group. This method would help ensure that the 
share of payment for each case-mix group is equal to its share of total 
costs of the component. The full methodology used to develop CMIs is 
presented in section 3.12 of the SNF PMR Technical Report.

                                                       Table 9--SLP Case-Mix Classification Groups
--------------------------------------------------------------------------------------------------------------------------------------------------------
                                      Presence of swallowing disorder or  SLP-related comorbidity or mild to
         Clinical category                mechanically- altered diet          severe cognitive impairment           Case-mix group        Case-mix index
--------------------------------------------------------------------------------------------------------------------------------------------------------
Acute Neurologic...................  Both...............................  Both..............................  SA                                    4.19
                                     Both...............................  Either............................  SB                                    3.71
                                     Both...............................  Neither...........................  SC                                    3.37

[[Page 20995]]

 
                                     Either.............................  Both..............................  SD                                    3.67
                                     Either.............................  Either............................  SE                                    3.12
                                     Either.............................  Neither...........................  SF                                    2.54
                                     Neither............................  Both..............................  SG                                    2.97
                                     Neither............................  Either............................  SH                                    2.06
                                     Neither............................  Neither...........................  SI                                    1.28
Non-Neurologic.....................  Both...............................  Both..............................  SJ                                    3.21
                                     Both...............................  Either............................  SK                                    2.96
                                     Both...............................  Neither...........................  SL                                    2.63
                                     Either.............................  Both..............................  SM                                    2.62
                                     Either.............................  Either............................  SN                                    2.22
                                     Either.............................  Neither...........................  SO                                    1.70
                                     Neither............................  Both..............................  SP                                    1.91
                                     Neither............................  Either............................  SQ                                    1.38
                                     Neither............................  Neither...........................  SR                                    0.61
--------------------------------------------------------------------------------------------------------------------------------------------------------

    As with the PT/OT component, under the RCS-I case-mix model, all 
residents would be classified into one, and only one, of these 18 SLP 
case-mix groups. As opposed to the RUG-IV system that determines 
therapy payments based only on the amount of therapy provided, under 
the RCS-I case-mix model, residents are classified into SLP case-mix 
groups based on resident characteristics shown to be predictive of SLP 
utilization. Thus, we believe that the SLP case-mix groups would 
provide a better measure of resource use and would provide for more 
appropriate payment under the SNF PPS. We invite comments on the series 
of ideas and the approach we are considering above associated with the 
SLP component of the RCS-I case-mix model.
d. Nursing Case-Mix Classification
    The RUG-IV classification system first divides residents into 
``rehabilitation residents'' and ``non-rehabilitation residents'' based 
on the amount of therapy a resident receives and other aspects of a 
resident's care. For rehabilitation residents, where the primary driver 
of payment classification is the intensity of therapy services that a 
resident receives, differences in nursing needs can be obscured. For 
example, for two residents classified into the RUB RUG-IV category, 
which would occur on the basis of therapy intensity and ADL score 
alone, the nursing component for each of these residents would be 
multiplied by a CMI of 1.56. This reflects that residents in that group 
were found, during our previous STM work, to have nursing costs 56 
percent higher than residents with a 1.00 index. We would note that 
while this CMI also includes adjustments made in FY 2010 and FY 2012 
for budget-neutrality purposes, what is clear is that two residents, 
who may have significantly different nursing needs, are nevertheless 
deemed to have the very same nursing costs, and SNFs would receive the 
same nursing payment for each. Given the discussion above, which noted 
that approximately 60 percent of resident days are billed using one of 
three Ultra-High Rehabilitation RUGs (two of which have the same 
nursing index), the current case-mix model effectively classifies a 
significant portion of SNF therapy residents as having exactly the same 
degree of nursing needs and requiring exactly the same amount of 
nursing resources. As such, we believe that further refinement of the 
case-mix model would be appropriate to better differentiate among 
patients with different nursing needs.
    An additional concern in the RUG-IV system is the use of therapy 
minutes to determine not only therapy payments, but also nursing 
payments. For example, residents classified into the RUB RUG fall in 
the same ADL score range as residents classified into the RVB RUG. The 
only difference between those residents is the number of therapy 
minutes that they received. However, the difference in payment that 
results from this difference in therapy minutes impacts not only the 
RUG-IV therapy component, but also the nursing component: Nursing 
payments for RUB residents are 40 percent higher than nursing payments 
for RVB residents. As a result of this feature of the RUG-IV system, 
the amount of therapy minutes provided to a resident is one of the main 
sources of variation in nursing payments, at the expense of other 
resident characteristics that may better reflect nursing needs.
    We believe that the more nuanced and resident-centered 
classifications in current RUG-IV non-rehabilitation categories are 
obscured under the current payment system, which utilizes only a single 
RUG-IV category for payment purposes and which has over 90 percent of 
resident days billed using a rehabilitation RUG. The RUG-IV non-
rehabilitation groups classify residents based on their ADL score, the 
use of extensive services, the presence of specific clinical conditions 
such as depression, pneumonia or septicemia, and the use of restorative 
nursing services, among other characteristics. These characteristics 
are associated with nursing utilization, and the STRIVE study accounted 
for relative differences in nursing staff time across groups. 
Therefore, we are considering continuing to use the existing non-
rehabilitation RUGs for the purposes of resident classification under 
RCS-I, but also modify nursing payment so that a resident's non-
rehabilitation RUG classification is always a factor in a resident's 
payment calculation.
    For example, consider two residents. The first classifies into the 
RUB rehabilitation RUG (on the basis of the resident's therapy minutes) 
and into the CC1 non-rehabilitation RUG (on the basis of having 
Pneumonia), while the second classifies into the RUB rehabilitation RUG 
(on the basis of the resident's therapy minutes) and the HC1 non-
rehabilitation RUG (on the basis of the resident being a Quadriplegic 
with a high ADL score). Under the current RUG-IV based payment model, 
the billing for both residents would utilize only the RUB 
rehabilitation RUG, despite clear differences in their associated 
nursing needs and resident characteristics. We are considering an

[[Page 20996]]

approach where, under the RCS-I payment model, for purposes of 
determining payment under the nursing component, the first resident 
would be classified into CC1, while the second would be classified into 
HC1. We believe that classifying the residents in this manner for 
payment purposes would capture variation in nursing costs in a more 
accurate and granular way than relying on the rehabilitation RUG's 
nursing CMI.
    In addition to considering the use of the resident's non-
rehabilitation RUG-IV classification for purposes of RCS-I payments, we 
also are considering the possibility of revising the existing nursing 
CMIs and updating these indexes through use of the STRIVE STM data 
which were originally used to create these indexes. Under the current 
payment system, non-rehabilitation nursing indexes were calculated to 
capture variation in nursing utilization by using only the staff time 
collected for the non-rehabilitation population. We believe that, to 
provide a more accurate sense of the relative nursing resource needs of 
the SNF population, the nursing indexes should reflect nursing 
utilization for all residents. To accomplish this, Acumen first 
replicated the methodology described in the FY 2010 SNF PPS rule (74 FR 
22236 through 22238), but classified the full STRIVE study population 
under non-rehabilitation RUGs using updated wage data. That methodology 
proceeded according to the following steps:
    (1) Calculate average wage-weighted staff time (WWST) for each 
STRIVE study resident using FY 2015 SNF wages.
    (2) Assign the full STRIVE population to the appropriate non-
rehabilitation RUG.
    (3) Apply sample weights to WWST estimates to allow for unbiased 
population estimates. The reason for this weighting is that the STRIVE 
study was not a random sample of residents. Certain key subpopulations, 
such as residents with HIV/AIDS, were over-sampled to ensure that there 
were enough residents to draw conclusions on the subpopulations' 
resource use. As a result, STRIVE researchers also developed sample 
weights, equal to the inverse of each resident's probability of 
selection, to permit calculation of unbiased population estimates. 
Applying the sample weights to a summary statistic results in an 
estimate that is representative of the actual population. The sample 
weight method is explained in Phase I of the STRIVE study. A link to 
the STRIVE study is available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/TimeStudy.html.
    (4) Smooth WWST estimates that do not match RUG hierarchy, as was 
done during the STRIVE study. RUG-IV, from which the nursing RUGs are 
derived, is a hierarchical classification in which payment should track 
clinical acuity. It is intended that residents who are more clinically 
complex or who have other indicators of acuity, including a higher ADL 
score, depression, or restorative nursing services, would receive 
higher payment. When STRIVE researchers estimated WWST for each RUG, 
several inversions occurred because of imprecision in the means. These 
are defined as WWST estimates that are not in line with clinical 
expectations. The methodology used to smooth WWST estimates is 
explained in Phase II of the STRIVE study. A link to the STRIVE study 
is available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/TimeStudy.html.
    (5) Calculate nursing indexes, which reflect the average WWST for 
each non-rehabilitation RUG divided by the average WWST for the study 
population used throughout our research. This analysis is presented in 
section 3.6.6 of the SNF PMR Technical Report.
    Through this refinement, we believe the nursing indexes under the 
RCS-I classification model would better reflect the varied nursing 
resource needs of the full SNF population. In Table 10, we provide the 
nursing indexes under the RCS-I classification model.
    To help ensure that payment reflects the average relative resource 
use at per diem level, nursing CMIs would be set to reflect case-mix 
related relative differences in WWST across groups. Nursing CMIs would 
be calculated based on the average per diem nursing WWST of a case-mix 
group relative to the population average. In this calculation, average 
per diem WWST equals total WWST in the group divided by number of 
utilization days in the group. The full methodology used to develop 
CMIs is presented in section 3.12 of the SNF PMR Technical Report.

       Table 10--Nursing Indexes Under RCS-I Classification Model
------------------------------------------------------------------------
                                                     Current
                                                     nursing    Nursing
                  RUG-IV category                    case-mix   case-mix
                                                      index      index
------------------------------------------------------------------------
ES3...............................................       3.58       3.84
ES2...............................................       2.67       2.90
ES1...............................................       2.32       2.77
HE2...............................................       2.22       2.27
HE1...............................................       1.74       2.02
HD2...............................................       2.04       2.08
HD1...............................................       1.60       1.86
HC2...............................................       1.89       2.06
HC1...............................................       1.48       1.84
HB2...............................................       1.86       1.88
HB1...............................................       1.46       1.67
LE2...............................................       1.96       1.88
LE1...............................................       1.54       1.68
LD2...............................................       1.86       1.84
LD1...............................................       1.46       1.64
LC2...............................................       1.56       1.55
LC1...............................................       1.22       1.39
LB2...............................................       1.45       1.48
LB1...............................................       1.14       1.32
CE2...............................................       1.68       1.84
CE1...............................................       1.50       1.60
CD2...............................................       1.56       1.74
CD1...............................................       1.38       1.51
CC2...............................................       1.29       1.49
CC1...............................................       1.15       1.30
CB2...............................................       1.15       1.37
CB1...............................................       1.02       1.19
CA2...............................................       0.88       1.03
CA1...............................................       0.78       0.89
BB2...............................................       0.97       1.05
BB1...............................................       0.90       0.97
BA2...............................................       0.70       0.74
BA1...............................................       0.64       0.68
PE2...............................................       1.50       1.60
PE1...............................................       1.40       1.47
PD2...............................................       1.38       1.48
PD1...............................................       1.28       1.36
PC2...............................................       1.10       1.23
PC1...............................................       1.02       1.13
PB2...............................................       0.84       0.98
PB1...............................................       0.78       0.90
PA2...............................................       0.59       0.68
PA1...............................................       0.54       0.63
------------------------------------------------------------------------

As with the previously discussed components, under the RCS-I case-mix 
model, all residents would be classified into one, and only one, of 
these 43 nursing case-mix groups.
    We also used the STRIVE data to quantify the effects of HIV/AIDS 
diagnosis on nursing resource use. Acumen controlled for case mix by 
including the RCS-I resident groups (in this case, the nursing RUGs) as 
independent variables. The results show that even after controlling for 
nursing RUG, HIV/AIDS status is associated with a positive and 
significant increase in nursing utilization. Based on the results of 
regression analyses, we found that wage-weighted nursing staff time is 
19 percent higher for residents with HIV/AIDS. (The weighting adjusted 
this estimate to account for the deliberate over-sampling of certain 
sub-populations in the STRIVE study, as described above.) Based on 
these findings, we concluded that the RCS-I nursing groups may not 
completely

[[Page 20997]]

capture the additional nursing costs associated with HIV/AIDS 
residents. More information on this analysis can be found in section 
3.8.2 of the SNF PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. 
Thus, as part of the case-mix adjustment of the nursing component, we 
are considering a 19 percent increase in payment for the nursing 
component for residents with HIV/AIDS. This adjustment would be applied 
based on the presence of ICD-10-CM code B20 on the SNF claim.
    We invite comments on the series of ideas and the approach we are 
considering above associated with the nursing component of the RCS-I 
case-mix model.
e. Non-Therapy Ancillary Case-Mix Classification
    Currently under the SNF PPS, payments for NTA costs incurred by 
SNFs are incorporated into the nursing component, which means that the 
CMIs used to adjust the nursing component of the SNF PPS are intended 
to reflect not only differences in nursing resource use, but also NTA 
costs. However, there have been concerns that the current nursing CMIs 
do not accurately reflect the basis for or the magnitude of relative 
differences in resident NTA costs. In its March 2016 Report to 
Congress, MedPAC wrote that ``Almost since its inception, the SNF PPS 
has been criticized for encouraging the provision of unnecessary 
rehabilitation therapy services and not accurately targeting payments 
for nontherapy ancillary (NTA) services such as drugs (Government 
Accountability Office 2002, Government Accountability Office 1999, 
White et al. 2002).'' (available at http://medpac.gov/docs/default-source/reports/chapter-7-skilled-nursing-facility-services-march-2016-report-.pdf). While the PT/OT and SLP components were designed to 
address the first criticism raised by MedPAC above, the NTA component 
discussed in this section was designed to address the second 
criticism--specifically, that the current manner of case-mix adjusting 
for NTAs under the RUG-IV case-mix system is inadequate in adjusting, 
in a targeted manner, for relative differences in resident NTA costs. 
As noted in the quotation from MedPAC above, MedPAC is not the only 
group to offer this critique of the SNF PPS. Just as the aforementioned 
criticisms that MedPAC cited have existed almost since the inception of 
the SNF PPS itself, ideas for addressing this concern have a similarly 
long history.
    In response to comments on the 1998 interim final rule which served 
to establish the SNF PPS, we published a final rule on July 30, 1999 
(64 FR 41644). In this final rule, we acknowledged the commenters' 
concerns about the new system's ability to account accurately for NTA 
costs, such as the following:

    There were a number of comments expressing concern with the 
adequacy of the PPS rates to cover the costs of ancillary services 
other than occupational, physical, and speech therapy (non-therapy 
ancillaries), including such things as drugs, laboratory services, 
respiratory therapy, and medical supplies. Prescription drugs or 
medication therapy were frequently noted areas of concern due to 
their potentially high cost for particular residents. Some 
commenters suggested that the RUG-III case-mix classification 
methodology does not adequately provide for payments that account 
for the variation in, or the real costs of, these services provided 
to their residents. (64 FR 41647)

    In response to those comments, we stated that ``we are funding 
substantial research to examine the potential for refinements to the 
case-mix methodology, including an examination of medication therapy, 
medically complex patients, and other nontherapy ancillary services.'' 
(64 FR 41648). Since that time, we have discussed various research 
initiatives engaged in identifying a more appropriate means to case-mix 
adjust SNF PPS payments to reflect relative differences in resident NTA 
costs. In this ANPRM, we are considering such a methodology, which we 
believe would case-mix adjust SNF PPS payments more appropriately to 
reflect differences in NTA costs.
    Following the same methodology we used for the PT/OT and SLP 
components, the project team ran cost regression models to determine 
which resident characteristics may be predictive of relative increases 
in NTA costs. The three cost-related resident characteristics 
identified through this analysis were resident comorbidities, the use 
of extensive services (services provided to residents that are 
particularly expensive and/or invasive), and resident age. A simple 
resident classification generated by CART using these three 
characteristics alone explained 11.7 percent of the variation in NTA 
costs per day. We would note that while we did find a correlation 
between relative differences in NTA costs and resident age, we also 
found that the correlation between NTA costs and resident comorbidities 
and extensive services was much stronger and heard concerns from TEP 
panelists during the June 2016 TEP, which led us to remove age from 
further consideration as part of the NTA component. Particularly, some 
panelists expressed concern that including age as a determinant of NTA 
payment could create access issues for the older population.
    With regard to capturing comorbidity information, the project team 
first mapped ICD-10 diagnosis codes from the prior inpatient claim, SNF 
claim, and Section I of the 5-day MDS assessment to condition 
categories (CCs), which provide a broader sense of the impact of 
similar conditions on NTA costs. The full list of conditions and 
extensive services considered for inclusion in the NTA component 
appears in the Appendix of the SNF PMR Technical Report available at 
https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. This list was meant to encompass as many 
conditions and extensive services as possible from the MDS assessment 
and the CCs. We found, using cost regressions, that certain comorbidity 
conditions and extensive services were highly predictive of relative 
differences in resident NTA costs. These conditions and services are 
identified in Table 11. More information on this analysis can be found 
in section 3.7.1 of the SNF PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. We would note that, based on our analysis and 
feedback from stakeholders at the June 2016 TEP, certain services which 
showed increased NTA costs were eliminated from consideration based on 
potential adverse incentives which may be created by linking these 
services to payment. Oxygen therapy and BiPAP/CPAP were excluded from 
consideration. Clinicians associated with the project team noted that 
these services are easily delivered and prone to overutilization. 
Additionally, the costs for these treatments for respiratory conditions 
are likely captured by the increase in costs associated with MDS item 
I6200 (asthma, COPD, or chronic lung disease). Finally, three CCs are 
excluded due to concerns about coding reliability: 33 (inflammatory 
bowel disease), 57 (personality disorders), and 66 (attention deficit 
disorder).
    Having identified the list of relevant conditions and services for 
adjusting NTA payments, we considered different options for how to 
capture the variation in NTA costs explained by these identified 
conditions and services. One such method would be merely to count the 
number of comorbidities and services a resident receives and assign a 
score to that resident based on this

[[Page 20998]]

simple count. We found that this option did account for the additive 
effect of having multiple comorbidities and extensive services, but did 
not adequately reflect the relative differences in the impact of 
certain higher-cost conditions and services. We also considered a tier 
system similar to the one used in the IRF PPS, where SNF residents 
would be placed into payment tiers based on the costliest comorbidity 
or extensive service. However, we found that this option did not 
account for the additive effect noted above. To address both of these 
issues, we are considering the possibility of basing a resident's NTA 
score (which would be used to classify the resident into an NTA case-
mix classification group) on a weighted-count methodology. 
Specifically, as shown in Table 11, each of the comorbidities and 
services which factor into a resident's NTA classification is assigned 
a certain number of points based on its relative impact on a resident's 
NTA costs. Those conditions and services with a greater impact on NTA 
costs are assigned more points, while those with less of an impact are 
assigned fewer points. Points are assigned by grouping together 
conditions and extensive services with similar ordinary least squares 
(OLS) regression estimates. The regression used the selected conditions 
and extensive services to predict NTA costs per day. More information 
on this methodology and analysis can be found in section 3.7.1 of the 
SNF PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. The 
effect of this methodology is that the NTA component would adequately 
reflect relative differences in NTA costs of each condition or service, 
as well as the additive effect of having multiple comorbidities.
    A resident's total comorbidity/extensive services score, which 
would be the sum of the points associated with all of a resident's 
comorbidities and services, would be used to classify the resident into 
an NTA case-mix group. For conditions and services where the source is 
indicated as MDS item I8000, we would consider providing a crosswalk 
between the listed condition and the ICD-10-CM codes which may be coded 
to qualify that condition to serve as part of the resident's NTA 
classification. MDS item I8000 is an open-ended item in the MDS 
assessment where the assessment provider can fill in additional active 
diagnoses (in the form of ICD-10 codes) for the resident that are not 
explicitly on the MDS. In the case of Parenteral/IV Feeding, we are 
considering the possibility of separating this item into a high 
intensity item and a low intensity item, similar to how it is defined 
in the RUG-IV system. For a resident to qualify for the high intensity 
category, the percent of calories taken in by the resident by 
parenteral or tube feeding, as reported in item K0710A2 on the MDS 3.0, 
must be greater than 50 percent. To qualify for the low intensity 
category, the percent of calories taken in by the resident by 
parenteral or tube feeding, as reported in item K0710A2 on the MDS 3.0, 
must be greater than 25 percent but less than or equal to 50 percent, 
and the resident must receive an average fluid intake by IV or tube 
feeding of at least 501cc per day, as reported in item K0710B2 of the 
MDS 3.0. The criteria used to distinguish between high and low 
intensity parenteral or tube feeding is the same as is used to classify 
residents using this variable in the RUG-IV classification. We also 
want to note that the source of the HIV/AIDS score is listed as coming 
from the SNF claim. This is because certain states, comprising 16 in 
all, have state laws which prevent the reporting of HIV/AIDS diagnosis 
information to us through the current assessment system and/or prevent 
us from seeing such diagnosis information within that system, should 
that information be mistakenly reported. The states are Alabama, 
Alaska, California, Colorado, Connecticut, Idaho, Illinois, 
Massachusetts, Nevada, New Hampshire, New Jersey, New Mexico, South 
Carolina, Texas, Washington, and West Virginia.
    Given this restriction, it would not be possible to have SNFs 
utilize the MDS 3.0 as the vehicle to report HIV/AIDS diagnosis 
information for purposes of determining a resident's NTA 
classification. We note that, currently, we use a claims reporting 
mechanism as the basis for the temporary AIDS add-on payment which 
exists under the current SNF PPS. To address the issue discussed above 
with respect to reporting of HIV/AIDS diagnosis information under the 
RCS-I model, we are considering utilizing this existing claims 
reporting mechanism to determine a resident's HIV/AIDS score for 
purposes of NTA classification. More specifically, HIV/AIDS diagnosis 
information reported on the MDS would be ignored by the GROUPER 
software used to classify a resident into an NTA case-mix group. 
Instead, providers would be instructed to report to us on the 
associated SNF claims the HIPPS code provided to the SNF on the 
validation report associated with that assessment. The provider would 
then, following current protocol, enter ICD-10-CM code B20 on the 
associated SNF claim, as if it were being coded to receive payment 
through the current AIDS add-on payment. The PRICER software, which we 
use to determine the appropriate per diem payment for a provider based 
on their wage index and other factors, would make the adjustment to the 
resident's NTA case-mix group, based on the presence of the B20 code on 
the claim, and adjust the associated per diem payment based on the 
adjusted resident HIPPS code. Again, we would note that this 
methodology follows the same logic as the SNF PPS currently uses to pay 
the temporary AIDS add-on adjustment, but merely changes the target and 
type of adjustment from the SNF PPS per diem to the NTA component of 
the RCS-I case-mix model. The difference is that while under the 
current system, the presence of the B20 code would lead to a 128 
percent increase in the per diem rate, under RCS-I, the presence of the 
B20 code would mean the addition of 8 points (as determined by the OLS 
regression described above) to the resident's NTA score and categorize 
the resident into the appropriate NTA group, as well as an adjustment 
to the nursing component, as described in section III.D.3.d. of this 
ANPRM.
    Table 11 provides the list of conditions and extensive services 
that would be used for NTA classification, the source of that 
information, the tier into which each item falls, and the associated 
number of points for that condition. The tier for each comorbidity 
condition and extensive service is determined based on the number of 
points assigned to that condition. For example, all comorbidities 
assigned 2 points are in the ``medium'' tier. The tiers are only used 
as a mechanism to simplify understanding of the points for each 
condition or extensive service. Only the points are factored into the 
determination of the comorbidity score and ultimately the NTA resident 
group classification.

[[Page 20999]]



                     Table 11--Conditions and Extensive Services Used for NTA Classification
----------------------------------------------------------------------------------------------------------------
    Condition/extensive service               Source                        NTA tier                  Points
----------------------------------------------------------------------------------------------------------------
HIV/AIDS..........................  SNF Claim................  Ultra-High.......................              +8
Parenteral/IV Feeding--High         MDS Item K0510A2.........  Very-High........................              +7
 Intensity.
IV Medication.....................  MDS Item O0100H2.........  High.............................              +5
Parenteral/IV Feeding--Low          MDS Item K0710A2, K0710B2  High.............................              +5
 Intensity.
Ventilator/Respirator.............  MDS Item O0100F2.........  High.............................              +5
Transfusion.......................  MDS Item O0100I2.........  Medium...........................              +2
Kidney Transplant Status..........  MDS Item I8000...........  Medium...........................              +2
Opportunistic Infections..........  MDS Item I8000...........  Medium...........................              +2
Infection with multi-resistant      MDS Item I1700...........  Medium...........................              +2
 organisms.
Cystic Fibrosis...................  MDS Item I8000...........  Medium...........................              +2
Multiple Sclerosis (MS)...........  MDS Item I5200...........  Medium...........................              +2
Major Organ Transplant Status.....  MDS Item I8000...........  Medium...........................              +2
Tracheostomy......................  MDS Item O0100E2.........  Medium...........................              +2
Asthma, COPD, or Chronic Lung       MDS Item I6200...........  Medium...........................              +2
 Disease.
Chemotherapy......................  MDS Item O0100A2.........  Medium...........................              +2
Diabetes Mellitus (DM)............  MDS Item I2900...........  Medium...........................              +2
End-Stage Liver Disease...........  MDS Item I8000...........  Low..............................              +1
Wound Infection (other than foot).  MDS Item I2500...........  Low..............................              +1
Transplant........................  MDS Item I8000...........  Low..............................              +1
Infection Isolation...............  MDS Item O0100M2.........  Low..............................              +1
MRSA..............................  MDS Item I8000...........  Low..............................              +1
Radiation.........................  MDS Item O0100B2.........  Low..............................              +1
Diabetic Foot Ulcer...............  MDS Item M1040B..........  Low..............................              +1
Bone/Joint/Muscle Infections/       MDS Item I8000...........  Low..............................              +1
 Necrosis.
Highest Ulcer Stage is Stage 4....  MDS Item M300D1..........  Low..............................              +1
Osteomyelitis and Endocarditis....  MDS Item I8000...........  Low..............................              +1
Suctioning........................  MDS Item O0100D2.........  Low..............................              +1
DVT/Pulmonary Embolism............  MDS Item I8000...........  Low..............................              +1
----------------------------------------------------------------------------------------------------------------

    Given the NTA scoring methodology described above, and following 
the same methodology used for the PT/OT and SLP components, we then 
used the CART algorithm to determine the most appropriate splits in 
resident NTA case-mix groups. This methodology is more thoroughly 
explained in section 3.4.2 of the SNF PMR Technical Report available at 
https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. Based on the CART algorithm, we determined that 6 
case-mix groups would be necessary to classify residents adequately in 
terms of their NTA costs in a manner that captures sufficient variation 
in NTA costs without creating unnecessarily granular separations. More 
information on this analysis can be found in section 3.7.2 of the SNF 
PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. We 
provide the criteria for each of these groups, along with the CMI for 
each group, in Table 12.
    To help ensure that payment reflects the relative resource use at 
the per diem level, CMIs would be set to reflect case-mix related 
relative differences in costs across groups. CMIs for the NTA component 
would be calculated based on two factors. One factor is the average per 
diem costs of a case-mix group relative to the population average. 
Relative differences in costs due to different length of stay 
distribution across groups are removed from this calculation. The other 
factor is the average variable per diem adjustment factor of the group 
relative to the population average. In this calculation, average per 
diem costs equal total NTA costs in the group divided by number of 
utilization days in the group, and similarly the average variable per 
diem adjustment factor equals the sum of NTA variable per diem 
adjustment factors for all utilization days in the group divided by the 
number of utilization days. More information on the variable per diem 
adjustments factor is discussed in section III.D.4 of this ANPRM. This 
method would help ensure that the share of payment for each case-mix 
group is equal to its share of total costs of the component, which is 
consistent with the notion that per diem payments reflect differences 
in average per diem relative resource use. The full methodology used to 
develop CMIs is presented in section 3.12 of the SNF PMR Technical 
Report.

              Table 12--NTA Case-Mix Classification Groups
------------------------------------------------------------------------
                                                               NTA case-
             NTA score range                   NTA group       mix index
------------------------------------------------------------------------
11+.....................................  NA                        3.33
8-10....................................  NB                        2.59
6-7.....................................  NC                        2.02
3-5.....................................  ND                        1.52
1-2.....................................  NE                        1.16
0.......................................  NF                        0.83
------------------------------------------------------------------------

As with the previously discussed components, under the RCS-I case-mix 
model, all residents would be classified into one, and only one, of 
these 6 NTA case-mix groups. The RCS-I case-mix model creates a 
separate payment component for NTA services, as opposed to combining 
NTA and nursing into one component as in the RUG-IV system. This 
separation allows payment for NTA services to be based on resident 
characteristics that predict NTA resource utilization, rather than 
nursing staff time. Thus, we believe that the NTA case-mix groups would 
provide a better measure of resource utilization and would lead to more 
accurate payments under the SNF PPS.
    We invite comments on the series of ideas and the approach we are 
considering above associated with the NTA component of the RCS-I case-
mix model.

[[Page 21000]]

f. Payment Classifications Under RCS-I
    The current SNF PPS case-mix classification system, RUG-IV, 
classifies each resident into a single RUG, with a single payment for 
all services. By contrast, the RCS-I case-mix classification system 
would classify each resident into four components (PT/OT; SLP; NTA; and 
nursing) and provide a single payment based on these classifications. 
The payment for each component would be calculated by multiplying the 
CMI for the resident's group by the component federal base payment 
rate, and then by the specific day in the variable per diem adjustment 
schedule (as discussed in section III.B.4. of this ANPRM). 
Additionally, for residents with HIV/AIDS indicated on their claim, the 
nursing portion of payment would be multiplied by 1.19 (as discussed in 
section III.B.3.d of this ANPRM). These payments would then be added 
together, along with the non-case-mix component payment rate, to create 
a resident's total SNF PPS per diem rate under RCS-I. This section 
describes how two hypothetical residents would be classified into 
payment groups under the current payment system and the RCS-I model we 
are considering. To begin, consider two residents, Resident A and 
Resident B, with the resident characteristics identified in Table 13.

                                 Table 13--Hypothetical Resident Characteristics
----------------------------------------------------------------------------------------------------------------
      Resident characteristics                Resident A                            Resident B
----------------------------------------------------------------------------------------------------------------
Rehabilitation Received?...........  Yes........................  Yes.
Therapy Minutes....................  730........................  730.
Extensive Services.................  No.........................  No.
ADL Score..........................  9..........................  9.
Clinical Category..................  Acute Neurologic...........  Major Joint Replacement.
Functional Score...................  15.........................  15.
Cognitive Impairment...............  Moderate...................  Intact.
Swallowing Disorder?...............  No.........................  No.
Mechanically Altered Diet?.........  Yes........................  No.
SLP Comorbidity?...................  No.........................  No.
Comorbidity Score..................  7 (IV Medication and DM)...  1 (DVT).
Other Conditions...................  Dialysis...................  Septicemia.
Depression?........................  No.........................  Yes.
----------------------------------------------------------------------------------------------------------------

    Currently under the SNF PPS, Resident A and Resident B would be 
classified into the same RUG-IV group. They both received 
rehabilitation, did not receive extensive services, received 730 
minutes of therapy, and have an ADL score of 9. This places the two 
residents into the ``RUB'' RUG-IV group and SNFs would be paid at the 
same rate, despite the many differences between these two residents in 
terms of their characteristics, expected care needs, and predicted 
costs of care.
    Under the RCS-I case-mix model, however, these two residents would 
be classified very differently. With regard to the PT/OT component, 
Resident A would fall into group TN, as a result of his categorization 
in the Acute Neurologic group, functional score within the 14 to 18 
range, and the presence of a moderate to severe cognitive impairment. 
Resident B, however, would fall into group TA for the PT/OT component, 
as a result of his categorization in the Major Joint Replacement group, 
a functional score within the 14 to 18 range, and the absence of any 
moderate or severe cognitive impairment. For the SLP component, 
Resident A would be classified into group SE., based on his 
categorization in the Acute Neurologic group, the presence of 
Mechanically-Altered Diet and presence of moderate cognitive 
impairment, while Resident B would be classified into group SR, based 
on his categorization in the Non-Neurologic group, the lack of any 
swallowing disorder or mechanically-altered diet, and absence of any 
SLP-related comorbidity or cognitive impairment. For the Nursing 
component, following the existing nursing case-mix methodology, 
Resident A would fall into group LC1, based on his use of dialysis 
services and an ADL score of 9, while Resident B would fall into group 
HC2, due to the diagnosis of septicemia, presence of depression, and 
ADL score of 9. Finally, with regard to NTA classification, Resident A 
would be classified in group NC, with an NTA score of 7, while Resident 
B would be classified in group NE., with an NTA score of 1. This 
demonstrates that, under the RCS-I case-mix model, more aspects of a 
resident's unique characteristics and needs factor into determining the 
resident's payment classification, which makes for a more resident-
centered case-mix model while also eliminating, or greatly reducing, 
the number of service-based factors which are used to determine the 
resident's payment classification. Because the RCS-I system would be 
based on specific resident characteristics predictive of resource 
utilization for each component, we expect that payments would be better 
aligned with resident need.
4. Variable Per Diem Adjustment Factors and Payment Schedule
    Section 1888(e)(4)(G)(i) of the Act provides that payments must be 
adjusted for case mix, based on a resident classification system which 
accounts for the relative resource utilization of different types of 
residents. Additionally, section 1888(e)(1)(B) of the Act specifies 
that payments to SNFs through the SNF PPS must be made on a per-diem 
basis. Currently under the SNF PPS, each RUG is paid at a constant per 
diem rate, regardless of how many days a resident is classified in that 
particular RUG. However, during the course of the SNF PMR project, 
analyses on cost over the stay for each of the case-mix adjusted 
components revealed different trends in resource utilization over the 
course of the SNF stay. These analyses utilized costs derived from 
claim charges as a measure of resource utilization. Costs were derived 
by multiplying charges from claims by the CCRs on facility-level costs 
reports. As described in section III.B.3.b of this ANPRM, costs better 
reflect differences in the relative resource use of residents as 
opposed to charges, which partly reflect decisions made by providers 
about how much to charge payers for certain services. In examining 
costs over a stay, we found that for certain categories of SNF 
services, notably therapy and NTA services, costs declined over the 
course

[[Page 21001]]

of a stay. Based on the claim submission schedule and variation in the 
point during the month when a stay began, we were able to estimate 
resource use for a specific day in a stay. Facilities are required to 
submit monthly claims. Each claim covers the period from the first day 
during the month a resident is in the facility to the end of the month. 
If a resident was admitted on the first day of the month and remains in 
the facility (and continues to have Part A SNF coverage) until the end 
of the month, the claim for that month will include all days in the 
month. However, if a resident is admitted after the first day of the 
month, the first claim associated with the resident's stay will be 
shorter than a month. To estimate resource utilization for each day in 
the stay, we used the marginal estimated cost from claims of varying 
length based on random variation in the day of a month when a stay 
began. To supplement this analysis, we also looked at changes in the 
number of therapy minutes reported in different assessments throughout 
the stay. Because therapy minutes are recorded on the MDS, the presence 
of multiple assessments throughout the stay provided information on 
changes in resource use. For example, it was clear whether the number 
of therapy minutes a resident received changed from the 5-day 
assessment to the 14-day assessment. The results from this analysis 
were consistent with the cost from claims analysis, and showed that on 
average, the number of therapy minutes is lower for assessments 
conducted later in the stay. This finding is consistent across 
different lengths of stay. More information on these analyses can be 
found in section 3.9.1 of the SNF PMR technical report is available at 
https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.
    Analyses of the SLP component revealed that the per diem costs 
remain relatively constant over time, while the PT/OT and NTA component 
cost analyses indicate that the per diem cost for these two components 
decline over the course of the stay. More specifically, in the case of 
the PT/OT component, costs start higher in the beginning of the stay 
and decline slowly over the course of the stay. The NTA component cost 
analyses indicate significantly increased NTA costs at the beginning of 
a stay, consistent with how most SNF drug costs are typically incurred 
at the outset of a SNF stay, and then drop to a much lower level that 
holds relatively constant over the remainder of the SNF stay. This 
indicates that resource utilization for PT/OT and NTA services change 
over the course of the stay. More information on these analyses can be 
found in section 3.9.1 of the SNF PMR technical report available at 
https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. We were unable to assess potential changes in the 
level of nursing costs over a resident's stay, in particular because 
nursing charges are not separately identifiable in SNF claims, and 
nursing minutes are not reported on the MDS assessments. However, 
stakeholders (industry representatives and clinicians) at multiple TEPs 
indicated that nursing costs tend to remain relatively constant over 
the course of a resident's stay.
    Constant per diem rates, by definition, do not track variations in 
resource use throughout a SNF stay, and we believe may allocate too few 
resources for SNF providers at the beginning of a stay. Given the 
trends in resource utilization discussed above, and that section 
1888(e)(4)(G)(i) of the Act requires the case-mix classification system 
to account for relative resource use, we are considering adjustments to 
the PT/OT and NTA components in the RCS-I model under consideration to 
account for the effect of length of stay on per diem costs (the 
variable per diem adjustments). We are not considering such adjustments 
to the SLP and nursing components based on findings and stakeholder 
feedback, as discussed above, that resource use tends to remain 
relatively constant over the course of a SNF stay.
    As noted above and as discussed more thoroughly in section 3.9.4 of 
the SNF PMR Technical Report (available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html), 
PT/OT costs decline at a slower rate relative to the decline in NTA 
costs. Therefore, in addition to considering a variable per diem 
adjustment, we further are considering to have separate adjustment 
schedules and indexes for the PT/OT component and the NTA component to 
more closely reflect the rate of decline in resource utilization for 
each component. Table 14 provides the adjustment factors and schedule 
we are considering for the PT/OT component, while Table 15 provides the 
adjustment factors and schedule we are considering for the NTA 
component.
    In Table 14, the adjustment factor is 1.00 for days 1 to 14. This 
is because the analyses described above indicated that PT/OT costs 
remain relatively high for the first 14 days and then decline. The 
estimated daily rate of decline for PT/OT costs relative to the initial 
fourteen days is 0.34 percent. Therefore, we believe a convenient and 
appropriate way to reflect this in the adjustment factors would be to 
have a decline of 1 percent every 3 days after day 14. The 0.34 percent 
rate of decline is derived from a regression model that estimates the 
level of resource use for each day in the stay relative to the 
beginning of the stay. The regression methodology and results are 
presented in section 3.9.3 of the SNF PMR Technical Report.
    NTA resource utilization, as described above, exhibits a somewhat 
different pattern. NTA costs are very high at the beginning of the 
stay, drop rapidly after the first three days, and remain relatively 
stable from the fourth day of the stay. Starting on day 4 of a stay, 
the per diem costs drop to roughly one-third of the per diem costs in 
the initial 3 days. This suggests that many NTA services are provided 
in the first few days of a SNF stay. Therefore, we are considering 
setting the NTA adjustment factor for days 1 to 3 at 3.00 to reflect 
the extremely high initial costs, and then setting it at 1.00 (two-
thirds lower than the initial level) for subsequent days. The 
adjustment factor was set at 3.00 for the first 3 days and 1.00 after 
(rather than, for example, 1.00 and 0.33, respectively) for simplicity.
    Case-mix adjusted federal per diem payment for a given component 
and a given day would be equal to the base rate for the relevant 
component (either urban or rural), multiplied by the CMI for that 
resident, multiplied by the variable per diem adjustment factor for 
that specific day, as applicable. Additionally, as described in further 
detail in section III.B.3.d of this ANPRM, an additional 19 percent 
would be added to the nursing per-diem payment to account for the 
additional nursing costs associated with residents who have HIV/AIDS. 
These payments would then be added together, along with the non-case-
mix component payment rate, to create a resident's total SNF PPS per 
diem rate under the RCS-I model under consideration.
    We invite comments on the ideas and the approach we are 
considering, as discussed above.

   Table 14--Variable Per-Diem Adjustment Factors and Schedule--PT/OT
------------------------------------------------------------------------
                                                              Adjustment
                   Medicare payment days                        factor
------------------------------------------------------------------------
1-14.......................................................         1.00
15-17......................................................         0.99
18-20......................................................         0.98
21-23......................................................         0.97

[[Page 21002]]

 
24-26......................................................         0.96
27-29......................................................         0.95
30-32......................................................         0.94
33-35......................................................         0.93
36-38......................................................         0.92
39-41......................................................         0.91
42-44......................................................         0.90
45-47......................................................         0.89
48-50......................................................         0.88
51-53......................................................         0.87
54-56......................................................         0.86
57-59......................................................         0.85
60-62......................................................         0.84
63-65......................................................         0.83
66-68......................................................         0.82
69-71......................................................         0.81
72-74......................................................         0.80
75-77......................................................         0.79
78-80......................................................         0.78
81-83......................................................         0.77
84-86......................................................         0.76
87-89......................................................         0.75
90-92......................................................         0.74
93-95......................................................         0.73
96-98......................................................         0.72
99-100.....................................................         0.71
------------------------------------------------------------------------


    Table 15--Variable Per-Diem Adjustment Factors and Schedule--NTA
------------------------------------------------------------------------
                                                              Adjustment
                   Medicare payment days                        factor
------------------------------------------------------------------------
1-3........................................................          3.0
4-100......................................................          1.0
------------------------------------------------------------------------

C. Use of the Resident Assessment Instrument--Minimum Data Set, Version 
3

1. Potential Revisions to Minimum Data Set (MDS) Completion Schedule
    Consistent with section 1888(e)(6)(B) of the Act, to classify 
residents under the SNF PPS, we use the MDS 3.0 Resident Assessment 
Instrument. Within the SNF PPS, there are two categories of 
assessments, scheduled and unscheduled. In terms of scheduled 
assessments, SNFs are required to complete assessments on or around 
Days 5, 14, 30, 60, and 90 of a resident's Part A SNF stay, including 
certain grace days. Payments based on these assessments depend upon 
standard Medicare payment windows associated with each scheduled 
assessment. More specifically, each of the Medicare-required scheduled 
assessments has defined days within which the Assessment Reference Date 
(ARD) must be set. The ARD is the last day of the observation (or 
``look-back'') period that the assessment covers for the resident. The 
facility is required to set the ARD on the MDS form itself or in the 
facility software within the appropriate timeframe of the assessment 
type being completed. The clinical data collected from the look-back 
period is used to determine the payment associated with each 
assessment. For example, the ARD for the 5-day PPS Assessment is any 
day between Days 1 to 8 (including Grace Days). The clinical data 
collected during the look-back period for that assessment is used to 
determine the SNF payment for Days 1 to 14. Section 413.343(b), MDS 3.0 
RAI Manual Chapter 2.5, 2.8. Unscheduled assessments, such as the Start 
of Therapy (SOT) Other Medicare Required Assessment (OMRA), the End of 
Therapy OMRA (EOT OMRA), the Change of Therapy (COT) OMRA, and the 
Significant Change in Status Assessment (SCSA or Significant Change), 
may be required during the resident's Part A SNF stay when triggered by 
certain defined events. For example, if a resident is being discharged 
from therapy services, but remaining within the facility to continue 
the Part A stay, then the facility may be required to complete an EOT 
OMRA. Each of the unscheduled assessments affects payment in different 
and defined manners. A description of the SNF PPS scheduled and 
unscheduled assessments, including the criteria for using each 
assessment, the assessment schedule, payment days covered by each 
assessment, and other related policies, are set forth in the MDS 3.0 
RAI manual on the CMS Web site (available at https://downloads.cms.gov/files/MDS-30-RAI-Manual-V114-October-2016.pdf). Table 16 outlines when 
each SNF PPS assessment is required to be completed and its effect on 
SNF PPS payment.

                                    Table 16--Current PPS Assessment Schedule
----------------------------------------------------------------------------------------------------------------
                                            Scheduled PPS assessments
-----------------------------------------------------------------------------------------------------------------
                                                               Assessment
Medicare MDS assessment schedule    Assessment reference     reference date      Applicable standard Medicare
              type                          date               grace days                payment days
----------------------------------------------------------------------------------------------------------------
5-day...........................  Days 1-5................               6-8  1 through 14.
14-day..........................  Days 13-14..............             15-18  15 through 30.
30-day..........................  Days 27-29..............             30-33  31 through 60.
60-day..........................  Days 57-59..............             60-63  61 through 90.
90-day..........................  Days 87-89..............             90-93  91 through 100.
----------------------------------------------------------------------------------------------------------------
                                           Unscheduled PPS assessments
----------------------------------------------------------------------------------------------------------------
Start of Therapy OMRA...........  5-7 days after the start of therapy         Date of the first day of therapy
                                                                               through the end of the standard
                                                                               payment period.
End of Therapy OMRA.............  1-3 days after all therapy has ended        First non-therapy day through the
                                                                               end of the standard payment
                                                                               period.
Change of Therapy OMRA..........  Day 7 (last day) of the COT observation     The first day of the COT
                                   period                                      observation period until End of
                                                                               standard payment period, or until
                                                                               interrupted by the next COT-OMRA
                                                                               assessment or scheduled or
                                                                               unscheduled PPS Assessment.
Significant Change in Status      No later than 14 days after significant     ARD of Assessment through the end
 Assessment.                       change identified                           of the standard payment period.
----------------------------------------------------------------------------------------------------------------

    An issue which has been raised in the past with regard to the 
existing SNF PPS assessment schedule is that the sheer number of 
assessments, as well as the complex interplay of the assessment rules, 
significantly increases the

[[Page 21003]]

administrative burden associated with the SNF PPS. Case-mix 
classification under the RCS-I model under consideration relies to a 
much lesser extent on characteristics that may change very frequently 
over the course of a resident's stay (for example, therapy minutes may 
change due to resident refusal or unexpected changes in resident 
status), but instead relies on more stable predictors of resource 
utilization by tying case-mix classification, to a much greater extent, 
to resident characteristics such as diagnosis information. In view of 
the greater reliance of the RCS-I case-mix classification system under 
consideration (as compared to the RUG-IV model) on resident 
characteristics that are relatively stable over a stay and our general 
focus on reducing administrative burden for providers across the 
Medicare program, if we were to implement the RCS-I model, we are 
considering the possibility of reducing the administrative burden on 
providers by concurrently revising the assessments that would be 
required under the RCS-I model. Specifically, we are considering the 
possibility of using the 5-day SNF PPS scheduled assessment to classify 
a resident under the RCS-I model under consideration for payment 
purposes for the entirety of his or her Part A SNF stay, except as 
described below. If we were to finalize this policy, we would revise 
the regulations at Sec.  413.343(b) so that such regulations would no 
longer reflect the RUG-IV assessment schedule.
    We understand that Medicare beneficiaries are each unique and can 
experience clinical changes which may require a SNF to reassess the 
resident to capture significant changes in the resident's condition. 
Therefore, to allow SNFs to capture these types of significant changes, 
under the RCS-I model we are considering, we would permit providers to 
reclassify residents from the initial 5-day classification using the 
Significant Change in Status Assessment (SCSA), which is a 
Comprehensive assessment (that is, an MDS assessment which includes 
both the completion of the MDS, as well as completion of the Care Area 
Assessment (CAA) process and care planning), but only in cases where 
the criteria for a significant change are met. A ``significant 
change,'' according to the MDS manual, is a major decline or 
improvement in a resident's status that: (1) Will not normally resolve 
itself without intervention by staff or by implementing standard 
disease-related clinical interventions, and is not ``self-limiting'' 
(for declines only); (2) Affects more than one area of the resident's 
health status; and (3) Requires interdisciplinary review and/or 
revision of the care plan. See the regulations at 42 CFR 
483.20(b)(2)(ii), and the MDS 3.0 RAI Manual, Chapter 2.6.
    In addition to providing for the completion of the SCSA, as 
described above, we have also considered the implications of a SNF 
completing an SCSA on the variable per diem adjustment schedule 
described in section III.B.4. of this ANPRM. More specifically, we have 
considered whether an SNF completing an SCSA should cause a reset in 
the variable per diem adjustment schedule for the associated resident. 
While we do believe that a significant change may be sufficient to 
cause a change in the resident's RCS-I classification, we do not 
believe that, in most instances, such a change would require a SNF to 
expend all of the resources that would be necessary to treat an 
individual who initially presented with that condition at admission. 
Furthermore, we are concerned that by providing for the variable per 
diem adjustment schedule to be reset after an SCSA is completed, 
providers may be incentivized to conduct multiple SCSAs during the 
course of a resident's stay to reset the variable per diem adjustment 
schedule each time the adjustment is reduced. Therefore, in cases where 
an SCSA is completed, we are considering an approach in which this 
assessment could reclassify the resident for payment purposes as 
outlined in Table 17, but the resident's variable per diem adjustment 
schedule would continue rather than being reset on the basis of 
completing the SCSA.
    Finally, under the RCS-I model we are considering, SNFs would 
continue to be required to complete a PPS Discharge Assessment. In 
addition, we are considering the possibility of adding certain items to 
this PPS Discharge Assessment that would allow CMS to track therapy 
minutes over the course of a resident's Part A stay. We believe that 
the combination of the 5-day Scheduled PPS Assessment, the Significant 
Change in Status Assessment, and the PPS Discharge Assessment would 
provide flexibility for providers to capture and report accurately the 
resident's condition, as well as accurately reflect resource 
utilization associated with that resident, while minimizing the 
administrative burden on providers under the RCS-I model being 
considered.
    Table 17 sets forth the PPS assessment schedule that we are 
considering, incorporating our ideas above.

                    Table 17--PPS Assessment Schedule
------------------------------------------------------------------------
    Medicare MDS assessment         Assessment      Applicable standard
         schedule type            reference date   medicare payment days
------------------------------------------------------------------------
5-day Scheduled PPS Assessment  Days 1-8.........  All covered Part A
                                                    days until Part A
                                                    discharge (unless a
                                                    Significant Change
                                                    in Status assessment
                                                    is completed).
Significant Change In Status    No later than 14   ARD of the assessment
 Assessment (SCSA).              days after         through Part A
                                 significant        discharge (unless
                                 change is          another Significant
                                 identified.        Change in Status
                                                    assessment is
                                                    completed).
PPS Discharge Assessment......  Equal to the End   N/A.
                                 Date of the Most
                                 Recent Medicare
                                 Stay (A2400C).
------------------------------------------------------------------------

    We would note that, as in previous years, we intend to continue to 
work with providers and software developers in understanding changes we 
might consider to the MDS. We invite comments on our ideas for 
revisions to the SNF PPS assessment schedule and related policies as 
discussed above. We also solicit comment on the extent to which 
implementing these ideas would reduce provider burden.
2. Potential Revisions to Therapy Provision Policies Under the SNF PPS
    Currently, almost 90 percent of residents in a Medicare Part A SNF 
stay receive therapy services. Under the current RUG-IV model, therapy 
services are case mix-adjusted primarily based on the therapy minutes 
reported on the MDS. When the original SNF PPS model was developed, 
most therapy services were furnished on an individual basis, and the 
minutes reported on the MDS served as a proxy for the staff resource 
time needed to provide the therapy care. Over the years, we have 
monitored

[[Page 21004]]

provider behavior and have made policy changes as it became apparent 
that, absent safeguards like quality measurement to ensure that the 
amount of therapy provided did not exceed the resident's actual needs, 
there were certain inherent incentives for providers to furnish as much 
therapy as possible. Thus, for example, in the SNF PPS FY 2010 final 
rule (74 FR 40315 through 40319), we decided to allocate concurrent 
therapy minutes for purposes of establishing the RUG-IV group to which 
the patient belongs, and to limit concurrent therapy to two patients at 
a time who were performing different activities.
    Following the decision to allocate concurrent therapy, using STRIVE 
data as a baseline, we found two significant provider behavior changes 
with regard to therapy provision under the RUG-IV payment system. 
First, there was a significant decrease in the amount of concurrent 
therapy that was provided in SNFs. Simultaneously, we observed a 
significant increase in the provision of group therapy, which was not 
subject to allocation at that time. We concluded that the manner in 
which group therapy minutes were counted in determining a patient's 
RUG-IV group created a payment incentive to provide group therapy 
rather than individual therapy or concurrent therapy, even in cases 
where individual therapy (or concurrent therapy) was more appropriate 
for the resident. Thus, we made two policy changes regarding group 
therapy in the FY 2012 SNF PPS final rule (76 FR 48511 through 48517). 
We defined group therapy as exactly four residents who are performing 
the same or similar therapy activities simultaneously. Additionally, we 
allocated group therapy among the four patients participating in group 
therapy--meaning that the total amount of time that a therapist spent 
with a group would be divided by 4 (the number of patients that 
comprise a group) to establish the RUG-IV group to which the patient 
belongs.
    Since we began allocating group therapy and concurrent therapy, 
these modes of therapy (group and concurrent) represent less than one 
percent of total therapy provided to SNF residents. Based on prior 
experience with the provision of concurrent and group therapy in SNFs, 
we again are concerned that if we were to implement the RCS-I model we 
are considering, providers may base decisions regarding the particular 
mode of therapy to use for a given resident on financial considerations 
rather than on the clinical needs of SNF residents. Because the RCS-I 
case-mix model would not use the minutes of therapy provided to a 
resident to classify the resident for payment purposes, we are 
concerned that SNFs may once again become incentivized to emphasize 
group and concurrent therapy, over the kind of individualized therapy 
which is tailored to address each beneficiary's specific care needs 
which we believe is generally the most appropriate mode of therapy for 
SNF residents.
    Since the inception of the SNF PPS, we have limited the amount of 
group therapy provided to each SNF Part A resident to 25 percent of the 
therapy provided to them. As stated in the FY 2000 final rule (64 FR 
41662):

    Although we recognize that receiving PT, OT, or ST as part of a 
group has clinical merit in select situations, we do not believe 
that services received within a group setting should account for 
more than 25 percent of the Medicare resident's therapy regimen 
during the SNF stay. For this reason, no more than 25 percent of the 
minutes reported in the MDS may be provided within a group setting. 
This limit is to be applied for each therapy discipline; that is, 
only 25 percent of the PT minutes reported in the MDS may be minutes 
received in a group setting and, similarly, only 25 percent of the 
OT, or the ST minutes reported may be minutes received in a group 
setting.

    Although we recognize that group and concurrent therapy may have 
clinical merit in specific situations, we also continue to believe that 
individual therapy is generally the best way of providing therapy to a 
resident because it is most tailored to that specific resident's care 
needs. As such, we believe that individual therapy should represent at 
least the majority of the therapy services received by SNF residents. 
To ensure that SNF residents would receive the majority of therapy 
services on an individual basis, if we were to implement the RCS-I 
model, we believe concurrent therapy should be limited to no more than 
25 percent of a SNF resident's therapy minutes, consistent with the 
existing 25 percent limit on group therapy. In combination, these two 
limits would ensure that at least 50 percent of a resident's therapy 
minutes are provided on an individual basis. For this reason, and 
because of the change in how therapy services would be used to classify 
residents under the RCS-I, and the concern that providers may begin to 
utilize more group and concurrent therapy due to financial 
considerations, we are considering setting a 25 percent limit on 
concurrent therapy, in addition to the 25 percent limit on group 
therapy that was established at the inception of the SNF PPS. Further, 
as with current policy as it relates to the group therapy cap, we are 
considering making the concurrent therapy limit discipline-specific. 
For example, if a resident received 800 minutes of physical therapy, no 
more than 200 minutes of this therapy could be provided on a concurrent 
basis and no more than 200 minutes of this therapy could be provided on 
a group basis.
    With a 25 percent limit on group therapy and a 25 percent limit on 
concurrent therapy, providers would be permitted to provide a total of 
50 percent of the total therapy furnished to each resident in a mode 
other than individual therapy. We believe that individual therapy is 
usually the best mode of therapy provision as it permits the greatest 
degree of interaction between the resident and therapist, and should 
therefore represent, at a minimum, the majority of therapy provided to 
an SNF resident. However, we recognize that, in very specific clinical 
situations, group or concurrent therapy may be the more appropriate 
mode of therapy provision, and therefore, we would want to allow 
providers the flexibility to be able to utilize these modes. We 
continue to stress that group and concurrent therapy should not be 
utilized to satisfy therapist or resident schedules, and that all group 
and concurrent therapy should be well documented in a specific way to 
demonstrate why they are the most appropriate mode for the resident and 
reasonable and necessary for his or her individual condition. We have 
also considered a combined limit on both concurrent and group therapy 
of 25 percent, but believe that this may not afford sufficient 
flexibility to SNFs to provide services as appropriate given the needs 
of the resident. We invite comments on the ideas discussed here and 
other ways in which these limits may be applied.
3. Interrupted Stay Policy
    Under section 1812(a)(2)(A) of the Act, Medicare Part A covers a 
maximum of 100 days of SNF services per spell of illness, or ``benefit 
period''. A benefit period starts on the day the beneficiary begins 
receiving inpatient hospital or SNF benefits under Medicare Part A. 
(See section 1861(a) of the Act; Sec.  409.60). SNF coverage also 
requires a prior qualifying, inpatient hospital stay of at least 3 
consecutive days' duration (counting the day of inpatient admission but 
not the day of discharge). (See section 1861(i) of the Act; Sec.  
409.30(a)(1)). Once the 100 available days of SNF benefits are used, 
the current benefit period must end before a beneficiary can renew SNF 
benefits under a new benefit period. For the

[[Page 21005]]

current benefit period to end so a new benefit period can begin, a 
period of 60 consecutive days must elapse throughout which the 
beneficiary is neither an inpatient of a hospital nor receiving skilled 
care in a SNF. (See section 1861(a) of the Act; Sec.  409.60). Once a 
benefit period ends, the beneficiary must have another qualifying 3-day 
inpatient hospital stay and meet the other applicable requirements 
before Medicare Part A coverage of SNF care can resume. (See section 
1861(i); Sec.  409.30)
    While the majority of SNF benefit periods, approximately 77 
percent, involve a single SNF stay, it is possible for a beneficiary to 
be readmitted multiple times to a SNF within a single benefit period, 
and such cases represent the remaining 23 percent of SNF benefit 
periods. For instance, a resident can be readmitted to a SNF within 30 
days after a SNF discharge without requiring a new qualifying 3-day 
inpatient hospital stay or beginning a new benefit period. SNF 
admissions that occur between 31 and 60 days after a SNF discharge 
require a new qualifying 3-day inpatient hospital stay, but fall within 
the same benefit period. (See sections 1861(a) and (i) of the Act; 
Sec. Sec.  409.30, 409.60)
    Other Medicare post-acute care (PAC) benefits have ``interrupted 
stay'' policies that provide for a payment adjustment when the 
beneficiary temporarily goes to another setting, such as an acute care 
hospital, and then returns within a specific timeframe. In the 
inpatient rehabilitation facility (IRF) and inpatient psychiatric 
facility (IPF) settings, for instance, an interrupted stay occurs when 
a patient returns to the same facility within 3 days of discharge. The 
interrupted stay policy for long-term care hospitals (LTCHs) is more 
complex, consisting of several policies depending on the length of the 
interruption and, at times, the discharge destination: An interruption 
of 3 or fewer days is always treated as an interrupted stay, which is 
similar to the IRF PPS and IPF PPS policies; if there is an 
interruption of more than 3 days, the length of the gap required to 
trigger a new stay varies depending on the discharge setting. In these 
three settings, when a beneficiary is discharged and returns to the 
facility within the interrupted stay window, Medicare treats the two 
segments as a single stay.
    While other PAC benefits have interrupted stay policies, the SNF 
benefit under the RUG-IV case-mix model has had no need for such a 
policy because given a resident's case-mix group, payment does not 
change over the course of a stay. In other words, assuming no change in 
a patient's condition or treatment, the payment rate is the same on Day 
1 of a covered SNF stay as it is at Day 7. Accordingly, a beneficiary's 
readmission to the SNF--even if only a few days may have elapsed since 
a previous discharge--could essentially be treated as a new and 
different stay without affecting the payment rates.
    However, as discussed in section III.B.4 of this ANPRM, under the 
RCS-I case-mix model, we are considering adjusting the PT/OT and NTA 
components of the per diem rate across the length of a stay (the 
variable per diem adjustment) to better reflect how and when costs are 
incurred and resources used over the course of the stay, such that 
earlier days in a given stay receive higher payments, with payments 
trending lower as the stay continues. In other words, the adjusted 
payment rate on Day 1 and Day 7 of a SNF stay would not be the same. 
Although we believe this variable per diem adjustment schedule more 
accurately reflects the increased resource utilization in the early 
portion of a stay for single-stay benefit periods (which represent the 
majority of cases), we have considered whether and how such an 
adjustment should be applied to payment rates for cases involving 
multiple stays per benefit period. In other words, if a resident has a 
Part A stay in a SNF, leaves the facility for some reason, and then is 
readmitted to the same SNF or a different SNF, we have considered how 
this readmission should be viewed in terms of both resident 
classification and the variable per diem adjustment schedule under the 
RCS-I model under consideration. Application of the variable per diem 
adjustment is of particular concern because providers may consider 
discharging a resident and then readmitting the resident shortly 
thereafter to reset the resident's variable per diem adjustment 
schedule and maximize the payment rates for that resident.
    Given the potential harm which may be caused to the resident if 
discharged inappropriately, and other concerns outlined above, we are 
considering the possibility of adopting an interrupted stay policy 
under the SNF PPS, in conjunction with the implementation of the RCS-I 
case-mix model. Specifically, as further explained below, in cases 
where a resident is discharged from a SNF and returns to the same SNF 
within 3 calendar days after having been discharged, we are considering 
the possibility of treating the resident's stay as a continuation of 
the previous stay for purposes of both resident classification and the 
variable per diem adjustment schedule. In cases where the resident is 
readmitted to the same SNF more than 3 calendar days after having been 
discharged, or in any case where the resident is readmitted to a 
different SNF, we are considering the possibility of treating the 
readmission as a new stay, in which the resident would receive a new 5-
day assessment upon admission and the variable per diem adjustment 
schedule for that resident would reset to Day 1. For the purposes of 
the interrupted stay policy, the source of the readmission would not be 
relevant. That is, the beneficiary may be readmitted from the 
community, from an intervening hospital stay, or from a different kind 
of facility and the interrupted stay policy would operate in the same 
manner. The only relevant factors in determining if the interrupted 
stay policy would apply are the number of days between the resident's 
discharge from a SNF and subsequent readmission to a SNF, and whether 
the resident is re-admitted to the same or a different SNF.
    Consider the following examples, which we believe aid in clarifying 
how this policy would be implemented:
    Example A: A beneficiary is discharged from a SNF stay on Day 3 of 
admission. Four days after the date of discharge, the beneficiary is 
then readmitted (as explained above, this readmission would be in the 
same benefit period). The SNF would conduct a new 5-day assessment at 
the start of the second admission and reclassify the beneficiary 
accordingly. In addition, for purposes of the variable per diem 
adjustment schedule, the payment schedule for the second admission 
would reset to Day 1 payment rates for the beneficiary's new case-mix 
classification.
    Example B: A beneficiary is discharged from a SNF stay on Day 7 and 
is readmitted to the same SNF before midnight of the date 3 calendar 
days from the day of discharge. For the purposes of classification and 
payment, this would be considered a continuation of the previous stay 
(an interrupted stay). The SNF would not conduct a new assessment to 
reclassify the patient and for purposes of the variable per diem 
adjustment schedule, the payment schedule would continue where it left 
off; in this case, the first day of the second stay would be paid at 
the Day 8 per diem rates under that schedule.
    We have also considered alternatives ways of structuring the 
interrupted stay policy. For example, we have considered possible 
ranges for the interrupted stay window other than the three calendar 
day window discussed in this ANPRM. For example, we considered windows 
of fewer than 3

[[Page 21006]]

days (for example, 1 or 2 day windows for readmission) as well as 
windows of more than 3 days (for example, 4 or 5 day windows for 
readmission). However, we believe that 3 days represents a reasonable 
window after which it is more likely that a resident's condition and 
resource needs will have changed. We also believe that consistency with 
other payment systems, like that of IRF and IPF, is helpful in 
providing clarity and consistency to providers in understanding 
Medicare payment systems, as well as making progress toward 
standardization among PAC payment systems. We invite comments on the 
appropriate length of the window for an interrupted stay policy.
    In addition, to determine how best to operationalize an interrupted 
stay policy within the SNF setting, we have considered three broad 
categories of benefit periods consisting of multiple stays. The first 
type of scenario, SNF-to-SNF transfers, is one in which a resident is 
transferred directly from one SNF to a different SNF. The second case 
we have considered, and the most common of all three multiple-stay 
benefit period scenarios, is a benefit period that includes a 
readmission following a new hospitalization between the two stays--for 
instance, a resident who was discharged from a SNF back to the 
community, re-hospitalized at a later date, and readmitted to a SNF 
(the same SNF or a different SNF) following the new hospital stay. The 
last case we have considered was a readmission to the same SNF or a 
different SNF following a discharge to the community, with no 
intervening re-hospitalization. Since benefit periods with exactly two 
stays account for a large majority of all benefit periods with multiple 
stays, we primarily examined benefit periods with two stays. Of these 
cases, over three quarters (76.4 percent) consist of re-hospitalization 
and readmission (to the same SNF or a different SNF). Community 
discharge and readmission without re-hospitalization cases represent 
approximately 14 percent of cases, while direct SNF-to-SNF transfers 
represent approximately 10 percent.
    For each of these case types, in which a resident was readmitted to 
a SNF no more than 3 days after discharge, we examined whether (1) the 
variable per diem adjustment schedule should be ``reset'' back to the 
Day 1 rates at the outset of the second stay versus ``continuing'' the 
variable per diem adjustment schedule at the point at which the 
previous stay ended, and (2) a new 5-day assessment and resident 
classification should be required at the start of the second, or other 
subsequent, SNF stay.
    With regard to the first question above, specifically whether or 
not a re-admission to a SNF no more than three calendar days after 
discharge from that SNF would reset the resident's variable per diem 
adjustment schedule, in each of the cases described above, we were 
concerned generally that an interrupted stay policy that ``restarts'' 
the variable per diem adjustment schedule to Day 1 after readmissions 
could incentivize unnecessary discharges with quick readmissions. This 
concern is particularly notable in the second and third cases described 
above, as the beneficiary may return to the same facility. Regression 
analyses showed that the second stay following a direct SNF-to-SNF 
transfer had similar costs to the first stay in a benefit period. As a 
result, the first case described above was excluded from the 
interrupted stay policy, which is restricted to readmissions to the 
same SNF. These types of transfers were also excluded from the 
interrupted stay policy because including such stays could potentially 
incentivize frequent discharge and readmission issues among facilities 
that share common ownership. In the second and third cases, the second 
stay tended to have lower costs than the first stay, suggesting that it 
is reasonable not to reset the resident's variable per diem adjustment 
schedule to address the incentive concerns described above.
    With regard to the first question above, we examined changes in 
costs from the first to second admission for the three scenarios 
described above (SNF-to-SNF direct transfers, readmissions following 
re-hospitalization, and readmissions following community discharge). 
Regression analyses showed that costs from the first to second 
admission were similar for SNF-to-SNF transfers and slightly lower for 
readmissions following re-hospitalizations. For readmissions following 
community discharges, costs were notably lower when residents returned 
to the same provider but similar when residents were admitted to a 
different facility. Because these results showed that an admission to a 
different SNF, regardless of the length of the gap between discharge 
and readmission, resulted in similar costs to the first admission, we 
are considering the possibility of always resetting the variable per 
diem adjustment schedule to Day 1 whenever residents are discharged and 
readmitted to a different SNF. We acknowledge that this could lead to 
patterns of inappropriate readmission that could be inconsistent with 
the intent of this policy; for example, we would be concerned about 
patients in SNF A consistently being admitted to SNF B to the exclusion 
of other SNFs in the area. However, because of the concern that a SNF 
provider could discharge and promptly readmit a resident to reset the 
variable per diem adjustment schedule to Day 1, in cases where a 
resident returns to the same provider we are considering allowing the 
payment schedule to reset only when the resident has been out of the 
facility for at least 3 days. More information on these analyses can be 
found in section 3.10.3 of the SNF PMR technical report available at 
https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.
    With regard to the question of whether or not SNFs would be 
required to complete a new 5-day assessment and reclassify the resident 
after returning to the SNF no more than 3 calendar days after discharge 
from the SNF, we investigated changes in resident characteristics from 
the first to the second stay within a benefit period. First, we looked 
at changes in clinical categories from the first to second stay for 
residents with an intervening re-hospitalization. This analysis could 
only be conducted for residents with a re-hospitalization because, as 
described in section 3.10.2 of the SNF PMR technical report, for 
research purposes classification into clinical categories was based on 
the diagnosis from the prior inpatient stay. Both SNF-to-SNF direct 
transfers and residents readmitted after a community discharge lacked a 
new hospitalization that would allow them to change clinical 
categories. (As described in section III.B.3.b of the ANPRM, 
classification into clinical categories would be operationalized under 
the RCS-I model under consideration using the primary diagnosis from 
item I8000 on the MDS 3.0. This information is not currently available; 
therefore, we used the prior inpatient diagnosis for research 
purposes.) For those residents who had a re-hospitalization and 
therefore could be reclassified into a new clinical category, we found 
that the vast majority fell into either the same category as in their 
first stay or the lowest-payment clinical category (medical 
management). For residents without a re-hospitalization between 
discharge and readmission, we examined changes in functional status 
from the first to second stay. Specifically, we looked at whether the 
RCS-I PT/OT group into which they were classified based on the 5-day

[[Page 21007]]

assessment of the second stay was associated with higher or lower 
functional status relative to the PT/OT group they were placed in based 
on the 5-day assessment of the first stay. We found that a large 
majority of these residents were classified into PT/OT groups 
associated with the same functional status across the first and second 
stays. More information on these analyses can be found in section 
3.10.2 of the SNF PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. Additionally, we note that under the approach 
discussed in section III.C.1 of this ANPRM, providers would be afforded 
the flexibility to use the SCSA, which would allow for reclassification 
in cases where a SCSA is warranted. Thus, we believe it would be 
appropriate to maintain the classification from the first stay for 
those residents returning to the SNF no more than 3 calendar days after 
discharge from the same facility.
    We invite comments on our ideas above.

D. Relationship of RCS-I to Existing Skilled Nursing Facility Level of 
Care Criteria

    Since the case-mix adjustment aspect of the SNF PPS has been based, 
in part, on the beneficiary's need for skilled nursing care and 
therapy, we have coordinated claims review procedures with the existing 
resident assessment process and case-mix classification system. This 
approach includes an administrative presumption that utilizes a 
beneficiary's initial classification in one of the upper 52 RUGs of the 
existing 66-group RUG-IV system to assist in making certain SNF level 
of care determinations.
    We are considering the possibility of adopting a similar approach 
under the RCS-I case-mix classification model, by retaining an 
administrative presumption mechanism that would utilize a beneficiary's 
initial classification into one of the designated upper groups to 
assist in making certain SNF level of care determinations. This 
designation would reflect an administrative presumption under the RCS-I 
model that beneficiaries who are correctly assigned to one of the 
designated groups on the initial 5-day, Medicare-required assessment 
are automatically classified as meeting the SNF level of care 
definition up to and including the assessment reference date on the 5-
day Medicare required assessment.
    As under the existing administrative presumption, a beneficiary who 
is not assigned to one of the designated groups would not automatically 
be classified as either meeting or not meeting the definition, but 
instead would receive an individual level of care determination using 
the existing administrative criteria. This presumption would recognize 
the strong likelihood that beneficiaries assigned to one of the 
designated upper groups during the immediate post-hospital period 
require a covered level of care, which would be less likely for those 
beneficiaries assigned to one of the lower groups.
    We note that the most direct crosswalk between the existing RUG-IV 
model and the RCS-I model under consideration would involve nursing 
services, for which each resident would be classified into one of the 
43 existing non-rehabilitation RUG-IV groups. Under the approach being 
considered, effective in conjunction with the implementation of the 
RCS-I model, the administrative presumption would continue to apply to 
those of the 43 groups that currently comprise the designated nursing 
categories under the existing RUG-IV model:
     Extensive Services;
     Special Care High;
     Special Care Low; and,
     Clinically Complex.
    In addition, along with the continued use of the remaining, nursing 
portion of the RUG-IV model, we also are considering the possibility of 
applying the administrative presumption using those other classifiers 
under the RCS-I model under consideration that we believe would relate 
the most directly to a given patient's acuity. As explained below, we 
would designate such classifiers for this purpose based on their 
ability to fulfill the administrative presumption's role as described 
in the FY 2000 SNF PPS final rule--that is, to identify those ``. . . 
situations that involve a high probability of the need for skilled care 
. . . when taken in combination with the characteristic tendency . . . 
for an SNF resident's condition to be at its most unstable and 
intensive state at the outset of the SNF stay'' (64 FR 41668 through 
41669, July 30, 1999).
    Specifically, we are considering the possibility of utilizing the 
PT/OT component's functional score, as well as the NTA component's 
comorbidity score for this purpose, which would be effective in 
conjunction with the implementation of the RCS-I model. Under this 
approach, those residents not classifying into one of the designated 
nursing RUG categories under the RCS-I model under consideration on the 
initial, 5-day Medicare-required assessment could nonetheless still 
qualify for the administrative presumption on that assessment, either 
by receiving the most intensive functional score (14 to 18) under the 
PT/OT component, or by receiving the uppermost comorbidity score (11+) 
under the NTA component. We believe that these particular clinical 
indicators would appropriately serve to fulfill the administrative 
presumption's role of identifying those cases with the highest 
probability of requiring an SNF level of care throughout the initial 
portion of the SNF stay. We note that to help improve the accuracy of 
these newly-designated groups in serving this function, we would 
continue to review the new designations going forward and could make 
further adjustments to the designations over time as we gain actual 
operating experience under the new classification model.
    We note that affording a streamlined and simplified administrative 
procedure for readily identifying such cases has been the basic purpose 
of the SNF PPS's level of care presumption ever since its inception. In 
this context, we wish to reiterate that an individual beneficiary's 
inability to qualify for the administrative presumption would not in 
itself serve to disqualify that resident from receiving SNF coverage. 
Instead, as we have noted repeatedly in previous rulemaking, while such 
residents are not automatically presumed to require a skilled level of 
care, neither are they automatically classified as requiring nonskilled 
care. Rather, any resident who does not qualify for the presumption 
would instead receive an individual level of care determination using 
the existing administrative criteria. As we explained in the FY 2016 
SNF PPS final rule, this approach serves ``. . . specifically to ensure 
that the presumption does not disadvantage such residents, by providing 
them with an individualized level of care determination that fully 
considers all pertinent factors'' (80 FR 46406, August 4, 2015).
    We invite comments on the ideas and the approach we are 
considering, as discussed above.

E. Effect of RCS-I on Temporary AIDS Add-on Payment

    Section 511(a) of the MMA amended section 1888(e)(12) of the Act to 
provide for a temporary increase of 128 percent in the PPS per diem 
payment for any SNF residents with Acquired Immune Deficiency Syndrome 
(AIDS), effective with services furnished on or after October 1, 2004. 
This special add-on for SNF residents with AIDS was intended to be of 
limited duration, as the MMA legislation specified that it was to 
remain in effect only until the Secretary

[[Page 21008]]

certifies that there is an appropriate adjustment in the case mix to 
compensate for the increased costs associated with such residents.
    The temporary add-on for SNF residents with AIDS is also discussed 
in Program Transmittal #160 (Change Request #3291), issued on April 30, 
2004, which is available online at www.cms.gov/transmittals/downloads/r160cp.pdf. In the SNF PPS final rule for FY 2010 (74 FR 40288, August 
11, 2009), we did not address this certification in that final rule's 
implementation of the case-mix refinements for RUG-IV, thus allowing 
the add-on payment required by section 511 of the MMA to remain in 
effect for the time being.
    In the House Ways and Means Committee Report that accompanied the 
MMA, the explanation of the MMA's temporary AIDS adjustment notes the 
following under Reason for Change: ``According to prior work by the 
Urban Institute, AIDS patients have much higher costs than other 
patients in the same resource utilization groups in skilled nursing 
facilities. The adjustment is based on that data analysis'' (H. Rep. 
No. 108-178, Part 2 at 221). The data analysis from that February 2001 
Urban Institute study (entitled ``Medicare Payments for Patients with 
HIV/AIDS in Skilled Nursing Facilities''), in turn, had been conducted 
under a Report to Congress mandated under a predecessor provision, 
section 105 of the BBRA. This earlier BBRA provision, which ultimately 
was superseded by the MMA's temporary AIDS add-on provision, had 
amended section 1888(e)(12) of the Act to provide for ``Special 
consideration for facilities serving specialized patient populations'' 
(that is, those who are ``immuno-compromised secondary to an infectious 
disease, with specific diagnoses as specified by the Secretary).
    We note that at this point, over 15 years have elapsed since the 
Urban Institute conducted its study on AIDS patients in SNFs, a period 
that has seen major advances in the state of medical practice in 
treating this condition. These advances have notably included the 
introduction of powerful new drugs and innovative prescription regimens 
that have dramatically improved the ability to manage the viral load 
(the amount of human immunodeficiency virus (HIV) in the blood). The 
decrease in viral load secondary to medications has contributed to a 
shift from intensive nursing services for AIDS-related illnesses to an 
increase in antiretroviral therapy. This phenomenon, in turn, is 
reflected in a recent analysis of differences in SNF resource 
utilization, which indicates that while the overall historical 
disparity in costs between AIDS and non-AIDS patients has not entirely 
disappeared, that disparity is now far greater with regard to drugs 
than it is for nursing. Specifically, NTA costs per day for residents 
with AIDS were 151 percent higher than those for other residents, while 
the difference in wage-weighted nursing staff time between the two 
groups was only 19 percent. More information on this analysis can be 
found in section 3.8.3 of the SNF PMR technical report available at 
https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html.
    As discussed previously in section III.B.3.e. of this ANPRM, the 
RCS-I model would include an NTA adjustment that we believe 
appropriately takes into account and compensates for those NTA costs, 
including drugs, which specifically relate to residents with AIDS. 
Regression analysis indicated that the case-mix adjustment for AIDS in 
the NTA component successfully accounts for the increased NTA resource 
utilization for residents with AIDS. Additionally, this analysis 
indicated that the case-mix adjustment of the NTA component accounts 
for most of the current disparity in payments between these and other 
residents, as suggested by a comparison of payments in RUG-IV and 
payments in RCS-I for residents with and without AIDS. More information 
on these analyses can be found in section 3.8.2 of the SNF PMR 
technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. Therefore, if we 
were to implement the RCS-I model we are considering, we believe it 
would be appropriate to issue the prescribed certification under 
section 511(a) of the MMA on the basis of the RCS-I model's NTA 
adjustment alone, as effectively representing the required appropriate 
adjustment in the case mix to compensate for the increased costs 
associated with such residents. However, to further ensure that the 
RCS-I model under consideration would account as fully as possible for 
any remaining disparity with regard to nursing costs, as discussed in 
section III.B.3.d., we are additionally considering the possibility of 
including a specific AIDS adjustment as part of the case-mix adjustment 
of the nursing component. As discussed in section III.B.3.d. of this 
ANPRM, we used the STRIVE data to quantify the effects of HIV/AIDS 
diagnosis on nursing resource use. Regression analyses found that wage-
weighted nursing staff time is 19 percent higher for residents with 
HIV/AIDS, controlling for the non-rehabilitation RUG of the resident. 
More information on this analysis can be found in section 3.8.2 of the 
SNF PMR technical report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. Thus, we 
are considering a 19 percent increase in payment for the nursing 
component for residents with HIV/AIDS under the RCS-I model under 
consideration to account for the increased nursing costs for such 
residents. Similar to the NTA adjustment for residents with HIV/AIDS 
discussed in section III.B.3.e. of this ANPRM, this adjustment would be 
identified by ICD-10-CM code B20 on the SNF claim and would be 
processed through the PRICER software used by CMS to set the 
appropriate payment rate for a resident's SNF stay. The 19 percent 
adjustment would be applied to the unadjusted base rate for the nursing 
component, and then this amount would be further case-mix adjusted per 
the resident's RCS-I classification.
    We believe that when taken collectively, these adjustments under 
the RCS-I case mix model that we discuss here would appropriately serve 
to justify issuing the certification prescribed under section 511(a) of 
the MMA effective with the conversion to the RCS-I model, which would 
permit the MMA's existing, temporary AIDS add-on to be replaced by a 
permanent adjustment in the case mix (under the RCS-I case mix model) 
that appropriately compensates for the increased costs associated with 
these residents. We invite comments on the ideas and the approach we 
are considering, as discussed above.

F. Potential Impacts of Implementing RCS-I

    To assess the potential effect of implementing the RCS-I case mix 
model, this section outlines the projected impacts of implementing this 
new case-mix classification model under the SNF PPS. The impacts 
presented here assume implementation of the RCS-I case-mix model and 
associated policy ideas discussed throughout section III. of this 
ANPRM.
    The impact analysis presented here makes a series of other 
assumptions as well, on all of which we solicit comment regarding their 
appropriateness. First, the impacts presented here assume consistent 
provider behavior in terms of how care is provided under RUG-IV and how 
care might be provided under RCS-I, as

[[Page 21009]]

we do not make any attempt to anticipate or predict provider reactions 
to the implementation of RCS-I. That being said, we acknowledge the 
possibility that implementing the RCS-I model could substantially 
affect resident care. Most notably, based on the concerns raised during 
a number of TEPs, we acknowledge the possibility that, as therapy 
payments under RCS-I would not have the same connection to service 
provision as they do under RUG-IV, it is possible that some providers 
may choose to reduce their provision of therapy services to increase 
margins under RCS-I. Additionally, we acknowledge that a number of 
states utilize some form of the RUG-IV case-mix classification system 
as part of their Medicaid programs and that any change in Medicare 
policy can have an impact on state programs. We solicit comments on 
this assumption that behavior would remain unchanged under RCS-I. To 
the extent that commenters may believe that behavior could change under 
RCS-I, we would ask that the commenters describe the types of 
behavioral changes we should expect. Additionally, we solicit comments 
on what type of impact on states we should expect from implementing the 
revisions considered in this ANPRM.
    Another assumption made for these impacts is that, as with prior 
system transitions, we would implement the RCS-I case-mix system, along 
with the other policy changes discussed in section III of this ANPRM, 
in a budget neutral manner through application of a parity adjustment 
to the case-mix weights under the RCS-I model under consideration, as 
further discussed below. We make this assumption because, as with prior 
system transitions, in considering changes to the case-mix methodology, 
we do not intend to change the aggregate amount of Medicare payments to 
SNFs, but rather to utilize a case-mix methodology to classify 
residents in such a manner as to best ensure that payments made for 
specific residents are an accurate reflection of resource utilization 
without introducing potential incentives which could incentivize 
inappropriate care delivery, as we believe may exist under the current 
case-mix methodology. However, as we would not be required to implement 
RCS-I in a budget neutral manner, we solicit comment on whether we 
should consider implementing RCS-I in a manner that is not budget 
neutral.
    For illustrative purposes, the impact analysis presented here 
assumes implementation of these changes in a budget neutral manner 
without a behavioral change. The prior sections describe how case-mix 
weights are set to reflect relative resource use for each case-mix 
group. RCS-I payment before application of a parity adjustment is 
calculated using the unadjusted CMI for each component, the variable 
per diem payment adjustment schedule, the different base rates for 
urban and rural facilities, the labor-related share, and the geographic 
wage indexes. In applying a parity adjustment to the case-mix weights, 
we maintained the relative value of each CMI, but multiplied every CMI 
by a ratio to achieve parity in overall SNF PPS payments under the RCS-
I case-model and under the RUG-IV case-mix model. The multiplier is 
calculated through the following steps. First, we calculate total 
payment subtracted by pre-AIDS adjusted non-case mix payment under RUG-
IV. Second, we calculate what total payment would have been under RCS-I 
before application of the parity adjustment. Third, we subtract non-
case-mix component payments from both calculations, as this component 
does not change across systems. This subtraction does not include the 
temporary add-on for residents with HIV/AIDS in the RUG-IV system, 
therefore ensuring that the amount subtracted is the same for both RUG-
IV and potential RCS-I payments, given the replacement of the temporary 
add-on described in section III.E. Lastly, we divide the remaining 
total RUG-IV payments over the remaining total RCS-I payments prior to 
the parity adjustment. This division yields a ratio (parity adjustment) 
by which the RCS-I CMIs are multiplied so that total estimated payments 
under the RCS-I model under consideration would be equal to total 
estimated payments under RUG-IV, assuming no changes in the population, 
provider behavior, and coding. More details regarding this calculation 
and analysis are described in section 3.12 of the SNF PMR Technical 
Report. The impact analysis presented in this section focuses on how 
payments under the RCS-I model under consideration would be re-
allocated across different resident groups and among different facility 
types, assuming implementation in a budget neutral manner. We invite 
comments on this discussion and approach.
    The projected resident-level impacts are presented in Table 18. The 
first column identifies different resident subpopulations and the 
second column shows what percent of SNF stays are represented by the 
given subpopulation. The third column shows the average change in 
payment for residents in a given subpopulation, represented as a 
percentage change from payments made for that subpopulation under RUG-
IV versus those which would be made under the RCS-I model under 
consideration. Positive changes in this column represent a projected 
positive shift in payments for that subpopulation under the RCS-I model 
under consideration, while negative changes in this column represent 
projected negative shifts in payment for that subpopulation. More 
information on the construction of current payments under RUG-IV and 
payments under the RCS-I model for purposes of this impact analysis can 
be found in section 3.13 of the SNF PMR Technical Report available at 
https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. Based on the data presented in Table 18, we 
observe that the most significant shift in payments created by 
implementation of the RCS-I case-mix model would be to redirect 
payments away from residents who are receiving very high amounts of 
therapy under the current SNF PPS (which strongly incentivizes the 
provision of therapy) to residents with more complex clinical needs. 
Other resident types that may see higher relative payments under the 
RCS-I system are residents with high NTA costs, dual-eligible 
residents, residents with ESRD, and residents with longer qualifying 
inpatient stays.

             Table 18--RCS-I Impact Analysis, Resident-Level
------------------------------------------------------------------------
                                            Percent of
        Resident characteristics               stays      Percent change
------------------------------------------------------------------------
All stays...............................           100.0             0.0
Sex:
    Female..............................            62.1            -0.7
    Male................................            37.9             1.2
Age:
    <65 years...........................             9.6             5.4

[[Page 21010]]

 
    65-74 years.........................            21.3             2.7
    75-84 years.........................            34.0            -0.3
    85-89 years.........................            19.3            -2.3
    90+ years...........................            15.7            -2.8
Race/Ethnicity:
    White...............................            85.2            -0.1
    Black...............................            10.6             0.4
    Hispanic............................             1.6            -0.2
    Asian...............................             1.2            -0.8
    Native American.....................             0.4             6.6
    Other or unknown....................             1.1             0.7
Medicare/Medicaid Dual Status:
    Dually enrolled.....................            35.2             2.9
    Not dually enrolled.................            64.8            -1.9
Original Reason for Medicare Enrollment:
    Aged................................            76.6            -1.2
    Disabled............................            22.5             3.9
    ESRD................................             0.9            10.0
    Unknown.............................             0.0            -3.3
Number of Utilization Days:
    1-15 days...........................            33.3            15.9
    16-30 days..........................            31.6             0.6
    31+ days............................            35.1            -2.5
Number of Utilization Days = 100:
    No..................................            97.4             0.3
    Yes.................................             2.6            -2.7
Length of Qualifying Inpatient Stay:
    3 days..............................            22.5            -2.3
    4-30 days...........................            73.6             0.5
    31+ days............................             1.8             4.6
Presence of Complications in MS-DRG of
 Qualifying Inpatient Stay:
    No Complication.....................            37.9            -2.3
    CC/MCC..............................            62.1             1.4
Stroke:
    No..................................            87.5            -0.1
    Yes.................................            12.5             0.7
CFS Level:
    Cognitive Intact....................            54.3            -0.5
    Mildly Impaired.....................            22.8             1.6
    Moderately Impaired.................            18.2            -1.8
    Severely Impaired...................             4.6             6.1
HIV:
    No..................................            99.7             0.2
    Yes.................................             0.3           -40.0
IV Medication:
    No..................................            91.4            -2.0
    Yes.................................             8.6            22.9
Diabetes:
    No..................................            65.0            -2.8
    Yes.................................            35.0             5.2
Wound Infection:
    No..................................            97.8            -0.4
    Yes.................................             2.2            17.9
Amputation/Prosthesis Care:
    No..................................           100.0             0.0
    Yes.................................             0.0             4.7
Most Common Therapy Level:
    RU..................................            54.0            -9.1
    RV..................................            22.7             9.3
    RH..................................             7.7            24.4
    RM..................................             3.7            36.9
    RL..................................             0.1            49.3
    Non-Rehabilitation..................            11.7            44.5
Number of Therapy Disciplines Used:
    0...................................             5.4            20.0
    1...................................             3.3            37.3
    2...................................            51.4             1.6
    3...................................            39.9            -3.9
Physical Therapy Utilization:
    No..................................             7.3            24.2
    Yes.................................            92.7            -1.0

[[Page 21011]]

 
Occupational Therapy Utilization:
    No..................................             8.6            24.8
    Yes.................................            91.4            -1.2
Speech Language Pathology Utilization:
    No..................................            58.4             3.2
    Yes.................................            41.6            -3.1
Therapy Utilization:
    PT+OT+SLP...........................            39.9            -3.9
    PT+OT Only..........................            50.4             1.2
    PT+SLP Only.........................             0.6            22.9
    OT+SLP Only.........................             0.5            25.6
    PT Only.............................             1.9            34.9
    OT Only.............................             0.7            41.8
    SLP Only............................             0.7            39.2
    Non-therapy.........................             5.4            20.0
NTA Costs:
    $0-$10..............................            10.9            -2.6
    $10-$50.............................            44.1            -3.2
    $50-$150............................            32.1             3.5
    $150+...............................             9.4            19.2
    Unknown.............................             3.5             3.3
Extensive Services Level:
    Tracheostomy and Ventilator/                     0.4            18.1
     Respirator.........................
    Tracheostomy or Ventilator/                      0.6             3.1
     Respirator.........................
    Infection Isolation.................             1.3             8.9
    Neither.............................            97.8            -0.3
------------------------------------------------------------------------

    Projected facility-level impacts are presented in Table 19. The 
first column identifies different facility subpopulations and the 
second column shows the percentage of SNFs represented by the given 
subpopulation. The third column shows the average change in payment for 
facilities in a given subpopulation, represented as a percentage change 
from payments made for that subpopulation under RUG-IV versus those 
which would be made under the RCS-I model under consideration. Positive 
changes in this column represent a projected positive shift in payments 
for that subpopulation under the RCS-I model under consideration, while 
negative changes in this column represent projected negative shifts in 
payment for that subpopulation. More information on the construction of 
current payments under RUG-IV and payments under the RCS-I model for 
purposes of this impact analysis can be found in section 3.13 of the 
SNF PMR Technical Report available at https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/SNFPPS/therapyresearch.html. Based on 
the data presented in Table 19, we observe that the most significant 
shift in Medicare payments created by implementation of the RCS-I case-
mix model would be from facilities with a high proportion of 
rehabilitation residents (more specifically, facilities with high 
proportions of Ultra-High Rehabilitation residents), to facilities with 
high proportions of non-rehabilitation residents. Other facility types 
that may see higher relative payments under the RCS-I system that we 
describe here are small facilities, non-profit facilities, government-
owned facilities, and hospital-based and swing-bed facilities.

             Table 19--RCS-I Impact Analysis, Facility-Level
------------------------------------------------------------------------
                                            Percent of        Percent
        Provider characteristics             providers        change
------------------------------------------------------------------------
All stays...............................           100.0             0.0
Institution type:
    Freestanding........................            95.0            -0.5
    Hospital-Based/Swing Bed............             5.0            15.8
Ownership:
    For-profit..........................            71.2            -1.1
    Non-profit..........................            23.9             3.1
    Government..........................             5.0             7.6
Location:
    Urban...............................            70.6            -0.8
    Rural...............................            29.4             3.7
Bed Size:
    0-49................................            11.2             6.7
    50-99...............................            37.1             0.3
    100-149.............................            34.3            -0.6
    150-199.............................            11.2            -0.5
    200+................................             6.1            -0.7
Census division:

[[Page 21012]]

 
    New England.........................             6.2             2.1
    Middle Atlantic.....................            11.2            -1.3
    East North Central..................            19.9             0.2
    West North Central..................            12.8             6.9
    South Atlantic......................            15.4            -0.8
    East South Central..................             6.6             1.0
    West South Central..................            13.2            -1.5
    Mountain............................             4.7             0.9
    Pacific.............................            10.1            -1.3
% of Stays with 100 Utilization Days:
    0-10%...............................            90.4             0.3
    10-25%..............................             8.6            -3.2
    25-100%.............................             1.0            -3.9
% of Stays with Medicare/Medicaid Dual
 Enrollment:
    0-10%...............................             8.4            -1.7
    10-2%...............................            17.2            -0.7
    25-50%..............................            35.5             0.6
    50-75%..............................            26.5             0.8
    75-90%..............................             8.5            -0.4
    90-100%.............................             3.8            -0.5
% of Utilization Days Billed as RU:
    0-10%...............................            12.5            28.4
    10-25%..............................             9.8            13.6
    25-50%..............................            25.5             5.6
    50-75%..............................            37.2            -1.9
    75-90%..............................            13.0            -7.1
    90-100%.............................             2.1            -9.9
 % of Utilization Days Billed as Non-
 Rehabilitation:
    0-10%...............................            70.4            -2.2
    10-25%..............................            23.2             6.3
    25-50%..............................             4.6            20.2
    50-75%..............................             1.0            45.6
    75-90%..............................             0.2            44.8
    90-100%.............................             0.7            38.4
------------------------------------------------------------------------

    In addition to the impacts discussed throughout this section, we 
would also note that we expect a significant reduction in regulatory 
burden under the SNF PPS, due to the changes we are considering in the 
MDS assessment schedule, as discussed above in section III.C.1 of this 
ANPRM. We invite comments on the impact analysis presented here.

IV. Collection of Information Requirements

    This ANPRM solicits comment on several options pertaining to the 
SNF PPS payment methodology. Since it does not propose any new or 
revised information collection requirements or burden, it need not be 
reviewed by the Office of Management and Budget (OMB) under the 
authority of the Paperwork Reduction Act of 1995 (44 U.S.C. 3501 et 
seq.). Should the outcome of the ANPRM result in any new or revised 
information collection requirements or burden, the requirements and 
burden will be submitted to OMB for approval. Interested parties will 
also be provided an opportunity to comment on such information through 
subsequent proposed and final rulemaking documents.

V. Response to Comments

    Because of the large number of public comments we normally receive 
on Federal Register documents, we are not able to acknowledge or 
respond to them individually. We will review all comments we receive by 
the date and time specified in the DATES section of this preamble, as 
we continue to consider the model presented in this ANPRM.

    Dated: April 21, 2017.
Seema Verma
Administrator, Centers for Medicare & Medicaid Services.
    Dated: April 21, 2017.
Thomas E. Price
Secretary, Department of Health and Human Services.
[FR Doc. 2017-08519 Filed 4-27-17; 4:15 pm]
BILLING CODE 4120-01-P