[Federal Register: August 15, 2005 (Volume 70, Number 156)]
[Rules and Regulations]
[Page 47879-48006]
From the Federal Register Online via GPO Access [wais.access.gpo.gov]
[DOCID:fr15au05-9]
[[Page 47879]]
-----------------------------------------------------------------------
Part II
Department of Health and Human Services
-----------------------------------------------------------------------
Centers for Medicare & Medicaid Services
-----------------------------------------------------------------------
42 CFR Part 412
Medicare Program; Inpatient Rehabilitation Facility Prospective Payment
System for FY 2006; Final Rule
[[Page 47880]]
-----------------------------------------------------------------------
DEPARTMENT OF HEALTH AND HUMAN SERVICES
Centers for Medicare & Medicaid Services
42 CFR Part 412
[CMS-1290-F]
RIN 0938-AN43
Medicare Program; Inpatient Rehabilitation Facility Prospective
Payment System for FY 2006
AGENCY: Centers for Medicare & Medicaid Services (CMS), HHS.
ACTION: Final rule.
-----------------------------------------------------------------------
SUMMARY: This final rule will update the prospective payment rates for
inpatient rehabilitation facilities for Federal fiscal year 2006 as
required under section 1886(j)(3)(C) of the Social Security Act (the
Act). Section 1886(j)(5) of the Act requires the Secretary to publish
the classification and weighting factors for the inpatient
rehabilitation facilities case-mix groups and a description of the
methodology and data used in computing the prospective payment rates
for that fiscal year.
In addition, we are implementing new policies and are changing
existing policies regarding the prospective payment system within the
authority granted under section 1886(j) of the Act.
DATES: These regulations are effective October 1, 2005. The updated IRF
prospective payment rates are applicable for discharges on or after
October 1, 2005 and on or before September 30, 2006 (FY 2006).
FOR FURTHER INFORMATION CONTACT: Pete Diaz, (410) 786-1235. Susanne
Seagrave, (410) 786-0044. Mollie Knight, (410) 786-7948 for information
regarding the market basket and labor-related share. August Nemec,
(410) 786-0612 for information regarding the tier comorbidities. Zinnia
Ng, (410) 786-4587 for information regarding the wage index and Core-
Based Statistical Areas (CBSAs).
SUPPLEMENTARY INFORMATION:
Table of Contents
I. Background
A. General Overview of the Current Inpatient Rehabilitation
Facility Prospective Payment System (IRF PPS)
B. Requirements for Updating the Prospective Payment Rates for
IRFs
C. Operational Overview of the Current IRF PPS
D. Summary of the FY 2006 Proposed Update to the IRF PPS
II. Provisions of the Proposed Regulations
III. Analysis of and Responses to Public Comments
IV. Research to Support Refinements of the Current IRF PPS
V. Refinements to the Patient Classification System
A. Changes to the IRF Classification System
1. Development of the IRF Classification System
2. Description and Methodology Used To Develop the IRF
Classification System in the August 7, 2001 Final Rule
a. Rehabilitation Impairment Categories
b. Functional Status Measures and Age
c. Comorbidities
d. Development of CMG Relative Weights
e. Overview of Development of the CMG Relative Weights
B. Changes to the Existing List of Tier Comorbidities
1. Changes to Remove Codes That Are Not Positively Related to
Treatment Costs
2. Changes to Move Dialysis to Tier One
3. Changes to Move Comorbidity Codes Based on Their Marginal
Cost
C. Changes to the CMGs
1. Changes for Updating the CMGs
2. Use of a Weighted Motor Score Index and Correction to the
Treatment of Unobserved Transfer to Toilet Values
3. Changes for Updating the Relative Weights
VI. FY 2006 Federal Prospective Payment Rates
A. Reduction of the Standard Payment Amount to Account for
Coding Changes
B. Adjustments to Determine the FY 2006 Standard Payment
Conversion Factor
1. Market Basket Used for IRF Market Basket Index
a. Overview of the RPL Market Basket
b. Methodology for Operating Portion of the RPL Market Basket
c. Methodology for Capital Proportion of the RPL Market Basket
d. Labor-Related Share
2. Area Wage Adjustment
a. Revisions of the IRF PPS Geographic Classification
b. Current IRF PPS Labor Market Areas Based on MSAs
c. Core-Based Statistical Areas (CBSAs)
d. Revisions of the IRF PPS Labor Market Areas
i. New England MSAs
ii. Metropolitan Divisions
iii. Micropolitan Areas
e. Implementation of the CBSA-Based Labor Market Areas
f. Wage Index Data
3. Teaching Status Adjustment
4. Adjustment for Rural Location
5. Adjustment for Disproportionate Share of Low-Income Patients
6. Update to the Outlier Threshold Amount
7. Budget Neutrality Factor Methodology for Fiscal Year 2006
8. Description of the Methodology Used to Implement the Changes
in a Budget Neutral Manner
9. Description of the IRF Standard Payment Conversion Factor for
Fiscal Year 2006
10. Example of the Methodology for Adjusting the Federal
Prospective Payment Rates
VII. Quality of Care in IRFs
VIII. Miscellaneous Comments Within the Scope of the Proposed Rule
IX. Miscellaneous Comments Outside the Scope of the Proposed Rule
X. Provisions of the Final Regulations
XI. Collection of Information Requirements
XII. Regulatory Impact Analysis
Acronyms
Because of the many terms to which we refer by acronym in this
final rule, we are listing the acronyms used and their corresponding
terms in alphabetical order below.
ADC Average Daily Census
AHA American Hospital Association
AMI Acute Myocardial Infarction
BBA Balanced Budget Act of 1997 (BBA), Pub. L. 105-33
BBRA Medicare, Medicaid, and SCHIP [State Children's Health
Insurance Program] Balanced Budget Refinement Act of 1999, Pub. L.
106-113
BIPA Medicare, Medicaid, and SCHIP [State Children's Health
Insurance Program] Benefits Improvement and Protection Act of 2000,
Pub. L. 106-554
BLS Bureau of Labor Statistics
CART Classification and Regression Trees
CBSA Core-Based Statistical Areas
CCR Cost-to-charge ratio
CMGs Case-Mix Groups
CMI Case Mix Index
CMSA Consolidated Metropolitan Statistical Area
CPI Consumer Price Index
DSH Disproportionate Share Hospital
ECI Employment Cost Index
FI Fiscal Intermediary
FIM Functional Independence Measure (FIMTM is a
registered trademark of UDSMR)
FIM-FRGs Functional Independence Measures-Function Related Groups
FRG Function Related Group
FTE Full-time equivalent
FY Federal Fiscal Year
GME Graduate Medical Education
HCRIS Healthcare Cost Report Information System
HIPAA Health Insurance Portability and Accountability Act
HHA Home Health Agency
IME Indirect Medical Education
IFMC Iowa Foundation for Medical Care
IPF Inpatient Psychiatric Facility
IPPS Inpatient Prospective Payment System
IRF Inpatient Rehabilitation Facility
IRF-PAI Inpatient Rehabilitation Facility-Patient Assessment
Instrument
IRF-PPS Inpatient Rehabilitation Facility-Prospective Payment System
IRVEN Inpatient Rehabilitation Validation and Entry
LIP Low-income percentage
MEDPAR Medicare Provider Analysis and Review
MSA Metropolitan Statistical Area
NECMA New England County Metropolitan Area
NOS Not Otherwise Specified
NTIS National Technical Information Service
OMB Office of Management and Budget
OSCAR Online Survey, Certification, and Reporting
PAI Patient Assessment Instrument
PLI Professional Liability Insurance
[[Page 47881]]
PMSA Primary Metropolitan Statistical Area
PPI Producer Price Index
PPS Prospective Payment System
RIC Rehabilitation Impairment Category
RPL Rehabilitation Hospital, Psychiatric Hospital, and Long-Term
Care Hospital Market Basket
TEFRA Tax Equity and Fiscal Responsibility Act
TEP Technical Expert Panel
I. Background
We received approximately 55 timely items of correspondence on the
Inpatient Rehabilitation Facility Prospective Payment System for FY
2006 proposed rule (70 FR 30188). Summaries of the public comments and
our responses to those comments are set forth below under the
appropriate section heading of this final rule.
A. General Overview of the Current Inpatient Rehabilitation Facility
Prospective Payment System (IRF PPS)
Section 4421 of the Balanced Budget Act of 1997 (BBA) (Pub. L. 105-
33), as amended by section 125 of the Medicare, Medicaid, and SCHIP
[State Children's Health Insurance Program] Balanced Budget Refinement
Act of 1999 (BBRA) (Pub. L. 106-113), and by section 305 of the
Medicare, Medicaid, and SCHIP Benefits Improvement and Protection Act
of 2000 (BIPA) (Pub. L. 106-554), provides for the implementation of a
per discharge prospective payment system (PPS), through section 1886(j)
of the Social Security Act (the Act), for inpatient rehabilitation
hospitals and inpatient rehabilitation units of a hospital (hereinafter
referred to as IRFs).
Payments under the IRF PPS encompass inpatient operating and
capital costs of furnishing covered rehabilitation services (that is,
routine, ancillary, and capital costs) but not costs of approved
educational activities, bad debts, and other services or items outside
the scope of the IRF PPS. Although a complete discussion of the IRF PPS
provisions appears in the August 7, 2001 final rule, we are providing
below a general description of the IRF PPS.
The IRF PPS, as described in the August 7, 2001 final rule, uses
Federal prospective payment rates across 100 distinct case-mix groups
(CMGs). Ninety-five CMGs were constructed using rehabilitation
impairment categories, functional status (both motor and cognitive),
and age (in some cases, cognitive status and age may not be a factor in
defining a CMG). Five special CMGs were constructed to account for very
short stays and for patients who expire in the IRF.
For each of the CMGs, we developed relative weighting factors to
account for a patient's clinical characteristics and expected resource
needs. Thus, the weighting factors account for the relative difference
in resource use across all CMGs. Within each CMG, the weighting factors
were ``tiered'' based on the estimated effects that certain
comorbidities have on resource use.
The Federal PPS rates were established using a standardized payment
amount (previously referred to as the budget-neutral conversion
factor). The standardized payment amount was previously called the
budget neutral conversion factor because it reflected a budget
neutrality adjustment for FYs 2001 and 2002, as described in Sec.
412.624(d)(2) of our regulations. However, the statute requires a
budget neutrality adjustment only for FYs 2001 and 2002. Accordingly,
for subsequent years we believe it is more consistent with the statute
to refer to the standardized payment as the standardized payment
conversion factor, rather than refer to it as a budget neutral
conversion factor (see 68 FR 45674, 45684 and 45685). Therefore, we
will refer to the standardized payment amount in this final rule as the
standard payment conversion factor.
For each of the tiers within a CMG, the relative weighting factors
were applied to the standard payment conversion factor to compute the
unadjusted Federal prospective payment rates. Under the current system,
adjustments that accounted for geographic variations in wages (wage
index), the percentage of low-income patients, and location in a rural
area were applied to the IRF's unadjusted Federal prospective payment
rates. In addition, adjustments were made to account for the early
transfer of a patient, interrupted stays, and high cost outliers.
Lastly, the IRF's final prospective payment amount was determined
under the transition methodology prescribed in section 1886(j) of the
Act. Specifically, for cost reporting periods that began on or after
January 1, 2002 and before October 1, 2002, section 1886(j)(1) of the
Act and as specified in Sec. 412.626 provide that IRFs transitioning
into the PPS would receive a ``blended payment.'' For cost reporting
periods that began on or after January 1, 2002 and before October 1,
2002, these blended payments consisted of 66\2/3\ percent of the
Federal IRF PPS rate and 33\1/3\ percent of the payment that the IRF
would have been paid had the IRF PPS not been implemented. However,
during the transition period, an IRF with a cost reporting period
beginning on or after January 1, 2002 and before October 1, 2002 could
have elected to bypass this blended payment and be paid 100 percent of
the Federal IRF PPS rate. For cost reporting periods beginning on or
after October 1, 2002 (FY 2003), the transition methodology expired,
and payments for all IRFs consist of 100 percent of the Federal IRF PPS
rate.
We established a CMS Web site that contains useful information
regarding the IRF PPS. The Web site URL is http://www.cms.hhs.gov/providers/irfpps/default.asp
and may be accessed to download or view
publications, software, and other information pertinent to the IRF PPS.
B. Requirements for Updating the Prospective Payment Rates for IRFs
On August 7, 2001, we published a final rule entitled ``Medicare
Program; Prospective Payment System for Inpatient Rehabilitation
Facilities'' in the Federal Register (66 FR at 41316), that established
a PPS for IRFs as authorized under section 1886(j) of the Act and
codified at subpart P of part 412 of the Medicare regulations. In the
August 7, 2001 final rule, we set forth the per discharge Federal
prospective payment rates for fiscal year (FY) 2002 that provided
payment for inpatient operating and capital costs of furnishing covered
rehabilitation services (that is, routine, ancillary, and capital
costs) but not costs of approved educational activities, bad debts, and
other services or items that are outside the scope of the IRF PPS. The
provisions of the August 7, 2001 final rule were effective for cost
reporting periods beginning on or after January 1, 2002. On July 1,
2002, we published a correcting amendment to the August 7, 2001 final
rule in the Federal Register (67 FR at 44073). Any references to the
August 7, 2001 final rule in this final rule include the provisions
effective in the correcting amendment.
Section 1886(j)(5) of the Act and Sec. 412.628 of the regulations
require the Secretary to publish the classifications and weighting
factors for the IRF CMGs and a description of the methodology and data
used in computing the prospective payment rates for the upcoming FY. On
August 1, 2002, we published a notice in the Federal Register (67 FR at
49928) to update the IRF Federal prospective payment rates from FY 2002
to FY 2003 using the methodology as described in Sec. 412.624. As
stated in the August 1, 2002 notice, we used the same classifications
and weighting factors for the IRF CMGs that were set forth in the
August 7, 2001 final rule to update the IRF Federal prospective payment
rates from FY 2002
[[Page 47882]]
to FY 2003. We have continued to update the prospective payment rates
each year in accordance with the methodology set forth in the August 7,
2001 final rule.
We published a proposed rule in the Federal Register (70 FR 30189)
to update the IRF Federal prospective payment rates from FY 2005 to FY
2006, and we proposed revisions to the methodology described in Sec.
412.624.
C. Operational Overview of the Current IRF PPS
As described in the August 7, 2001 final rule, upon the admission
and discharge of a Medicare Part A fee-for-service patient, the IRF is
required to complete the appropriate sections of a patient assessment
instrument, the Inpatient Rehabilitation Facility-Patient Assessment
Instrument (IRF-PAI). All required data must be electronically encoded
into the IRF-PAI software product. Generally, the software product
includes patient grouping programming called the GROUPER software. The
GROUPER software uses specific Patient Assessment Instrument (PAI) data
elements to classify (or group) the patient into a distinct CMG and
account for the existence of any relevant comorbidities.
The GROUPER software produces a 5-digit CMG number. The first digit
is an alpha-character that indicates the comorbidity tier. The last 4
digits represent the distinct CMG number. (Free downloads of the
Inpatient Rehabilitation Validation and Entry (IRVEN) software product,
including the GROUPER software, are available at the CMS Web site at
http://www.cms.hhs.gov/providers/irfpps/default.asp).
Once the patient is discharged, the IRF completes the Medicare
claim (UB-92 or its equivalent) using an alphanumeric CMG code and
sends it to the appropriate Medicare fiscal intermediary (FI). (Claims
submitted to Medicare must comply with both the Administrative
Simplification Compliance Act (ASCA), Pub. L. 107-105, and the Health
Insurance Portability and Accountability Act of 1996 (HIPAA), Pub. L.
104-191. Section 3 of ASCA requires the Medicare Program, subject to
subsection (H), to deny payment under Part A or Part B for any expenses
for items or services ``for which a claim is submitted other than in an
electronic form specified by the Secretary.'' Subsection (h) provides
that the Secretary shall waive such denial in two types of cases and
may also waive such denial ``in such unusual cases as the Secretary
finds appropriate.'' See also, 68 FR 48805 (August 15, 2003). Section 3
of ASCA operates in the context of the Administrative Simplification
provisions of HIPAA, which include, among others, the transactions and
code sets standards requirements codified as 45 CFR part 160 and 162,
subparts A and I through R (generally known as the Transactions Rule).
The Transactions Rule requires covered entities, including covered
providers, to conduct covered electronic transactions according to the
applicable transaction standards. See the program claim memoranda
issued and published by CMS at http://www.cms.hhs.gov/providers/edi/default.asp
(http://www.cms.hhs.gov/provider/edi/default.asp) and
listed in the addenda to the Medicare Intermediary Manual, Part 3,
section 3600. Instructions for the limited number of claims submitted
to Medicare on paper are located in section 3604 of Part 3 of the
Medicare Intermediary Manual.
The Medicare Fiscal Intermediary (FI) processes the claim through
its software system. This software system includes pricing programming
called the PRICER software. The PRICER software uses the CMG code,
along with other specific claim data elements and provider-specific
data, to adjust the IRF's prospective payment for interrupted stays,
transfers, short stays, and deaths and then applies the applicable
adjustments to account for the IRF's wage index, percentage of low-
income patients, rural location, and outlier payments.
D. Summary of the FY 2006 Proposed Update to the IRF PPS
In the FY 2006 proposed rule (70 FR 30188), we proposed a number of
refinements to the IRF PPS case-mix classification system (the CMGs and
the corresponding relative weights) and the case-level and facility-
level adjustments. The refinements that we proposed were based on
analyses by RAND using calendar year 2002 and FY 2003 data.
Several new developments warranted proposing these refinements,
including--(1) The availability of more recent 2002 and 2003 data; (2)
better coding of comorbidities and patient severity; (3) more complete
data; (4) new data sources for imputing missing values; and (5)
improved statistical approaches.
Our proposals included the following key changes:
The FY 2006 IRF PPS proposed rule (70 FR 30188, 30234 through
30241) included a proposal to adopt OMB's Core Based Statistical Area
(CBSA) market area definitions in a budget neutral manner. This
geographic adjustment is made using a 1-year lag of the pre-
reclassification hospital wage index (FY 2001 hospital wage data).
The FY 2006 proposed rule (70 FR 30188, 30222) also included a
proposal to implement a payment adjustment to account for changes in
coding. We proposed to reduce the standard payment amount by 1.9
percent to account for changes in coding following implementation of
the IRF PPS. The analysis conducted by CMS's contractor found that the
real change in the case-mix was between negative 2.4 percent and
positive 1.5 percent, with the rest of the change (between 1.9 percent
and 5.8 percent) attributable to coding changes. CMS proposed to reduce
the standard payment amount by the lowest of these estimates.
In addition, in the FY 2006 proposed rule (70 FR 30188), we
proposed modifications to the case mix groups, tier comorbidities, and
relative weights. The proposed rule included a number of adjustments to
the IRF classification system that are designed to improve the system's
ability to predict IRF costs. The new data indicate that moving or
eliminating some comorbidity codes from the tiers, redefining the case
mix groups, and other minor changes to the system could improve the
ability of the classification system to ensure that Medicare payments
to IRFs continue to be aligned with the costs of care.
In addition, the FY 2006 IRF PPS proposed rule (70 FR 30188, 30241)
contained a proposal to implement a new teaching status adjustment for
IRFs, similar to the one recently adopted for inpatient psychiatric
facilities. We proposed to implement the teaching status adjustment in
a budget neutral manner.
The FY 2006 IRF PPS proposed rule (70 FR 30188, 30222) also
contained a proposal to revise the market basket. We proposed to use a
new market basket reflecting the operating and capital cost structures
for rehabilitation, psychiatric, and long term care hospitals to update
IRF payment rates. The proposed new market basket excludes cancer
hospitals and children's hospitals. For the FY 2006 proposed rule (70
FR 30188), we proposed a market basket increase for FY 2006 of 3.1
percent.
In the FY 2006 proposed rule (70 FR 30188, 30244 through 30246), we
also proposed to update the rural adjustment (from 19.1 percent to 24.1
percent), the low-income patient adjustment (from an exponent of 0.484
to an exponent of 0.636), and the outlier threshold amount (from
$11,211 to $4,911). We proposed to implement the changes to the rural
and low-income percentage updates in a budget neutral manner.
[[Page 47883]]
Lastly, in the FY 2006 proposed rule (70 FR 30188), we estimated
that the proposed changes would increase costs to the Medicare program
for IRF services in FY 2006 by $180 million over FY 2005 levels. The
estimated increased cost to the Medicare program was due to the
estimated IRF market basket of 3.1 percent, the 1.9 percent reduction
to the standard payment amount to account for changes in coding that
affect total estimated aggregate payments, and the update to the
outlier threshold amount. We proposed to make the changes to the IRF
labor-related share and the wage indices, the case mix groups, tier
comorbidities, and relative weights, the new IME adjustment, the
updated rural adjustment, and the updated LIP adjustment in a budget
neutral manner. Thus, these proposed changes would have no overall
effect on estimated costs to the Medicare program.
II. Provisions of the Proposed Regulations
In the FY 2006 proposed update to the IRF PPS (70 FR 30188),
hereinafter referred to as the FY 2006 proposed rule, we proposed to
make revisions to the regulations to implement the proposed PPS for
IRFs for FY 2006 and subsequent fiscal years. Specifically, we proposed
to make conforming changes in 42 CFR part 412. These proposed revisions
and others are discussed in detail below.
A. Section 412.602 Definitions
In Sec. 412.602, we proposed to revise the definitions of ``Rural
area'' and ``Urban area'' to read as follows:
Rural area means: For cost-reporting periods beginning on or after
January 1, 2002, with respect to discharges occurring during the period
covered by such cost reports but before October 1, 2005, an area as
defined in Sec. 412.62(f)(1)(iii). For discharges occurring on or
after October 1, 2005, rural area means an area as defined in Sec.
412.64(b)(1)(ii)(C).
Urban area means: For cost-reporting periods beginning on or after
January 1, 2002, with respect to discharges occurring during the period
covered by such cost reports but before October 1, 2005, an area as
defined in Sec. 412.62(f)(1)(ii). For discharges occurring on or after
October 1, 2005, urban area means an area as defined in Sec.
412.64(b)(1)(ii)(A) and Sec. 412.64(b)(1)(ii)(B).
B. Section 412.622 Basis of Payment
In this section, we proposed to correct the cross references in
paragraphs (b)(1) and (b)(2)(i). In paragraph (b)(1), we proposed to
remove the cross references ``Sec. 413.85 and Sec. 413.86 of this
chapter'' and add in their place ``Sec. 413.75 and Sec. 413.85 of
this chapter.'' In paragraph (b)(2)(i), we proposed to remove the cross
reference ``Sec. 413.80 of this chapter'' and add in its place ``Sec.
413.89 of this chapter.''
C. Section 412.624 Methodology for Calculating the Federal Prospective
Payment Rates
In this section, we proposed to make the following revisions:
In paragraph (d)(1), remove the cross reference to
``paragraph (e)(4)'' and add in its place ``paragraph (e)(5).''
Add a new paragraph (d)(4).
Redesignate paragraphs (e)(4) and (e)(5) as paragraphs
(e)(5) and (e)(6).
Add a new paragraph (e)(4).
Revise newly redesignated paragraph (e)(5).
Revise newly redesignated paragraph (e)(6).
Add a new paragraph (e)(7).
In paragraph (f)(2)(v), remove the cross references to
``paragraphs (e)(1), (e)(2), and (e)(3) of this section'' and add in
their place ``paragraphs (e)(2), (e)(3), (e)(4), and (e)(7) of this
section.''
D. Additional Changes
We also proposed the following changes:
Reduce the standard payment amount by 1.9 percent to
account for coding changes.
Revise the comorbidity tiers and CMGs.
Use a weighted motor score index in assigning patients to
CMGs.
Update the relative weights.
Update payments for rehabilitation facilities using a
market basket reflecting the operating and capital cost structures for
the RPL market basket.
Provide the weights and proxies to use for the FY 2002-
based RPL market basket.
Indicate the methodology for the capital portion of the
RPL market basket.
Adopt the new geographic labor market area definitions as
specified in Sec. 412.64(b)(1)(ii)(A)-(C).
Use the New England MSAs as determined under the proposed
new CBSA-based labor market area definitions.
Implement a budget neutral 3 year hold harmless policy for
FY 2005 rural IRFs redesignated as urban in FY 2006.
Use FY 2001 acute care hospital wage data in computing the
FY 2006 IRF PPS payment rates.
Implement a teaching status adjustment.
Update the formulas used to compute the rural and the LIP
adjustments to IRF payments.
Update the outlier threshold amount to maintain total
estimated outlier payments at 3 percent of total estimated payments.
Revise the methodology for computing the standard payment
conversion factor (for FY 2006 only) to make the CMG and tier changes,
the teaching status adjustment, and the updates to the rural and LIP
adjustments in a budget neutral manner.
III. Analysis of and Responses to Public Comments
As stated above, we received approximately 55 timely items of
correspondence containing multiple comments on the FY 2006 proposed
rule (70 FR 30188) from providers, health industry organizations, the
Medicare Payment Advisory Commission, and others. In general,
commenters expressed some concerns about our proposals in light of
other changes occurring in the IRF PPS at this time and suggested that
we wait to implement the proposals until other recent IRF policy
changes are fully implemented. However, many commenters supported the
proposed changes to the facility-level adjustments. Summaries of the
public comments received on the proposed provisions and our responses
to those comments are provided in the appropriate sections of the
preamble of this final rule.
IV. Research To Support Refinements of the Current IRF PPS
As described in the August 7, 2001 final rule, we contracted with
the RAND Corporation to analyze IRF data to support our efforts in
developing the CMG patient classification system and the IRF PPS. Since
then, we have continued our contract with RAND to support us in
developing potential refinements to the classification system and the
PPS. RAND has also developed a system to monitor the effects of the IRF
PPS on patients' access to IRF care and other post-acute care services.
1. History of RAND's Research on the IRF PPS
In 1995, RAND began extensive research, sponsored by us, on the
development of a per-discharge based PPS using a patient classification
system known as Functional Independence Measures--Function Related
Groups (FIM-FRGs) for IRFs. The results of RAND's earliest research,
using 1994 data, were released in September 1997 and are contained in
two reports available through the National Technical Information
Service (NTIS). The reports are: Classification System
[[Page 47884]]
for Inpatient Rehabilitation Patients--A Review and Proposed Revisions
to the Function Independence Measure--Function Related Groups, NTIS
order number PB98-105992INZ, and Prospective Payment System for
Inpatient Rehabilitation, NTIS order number PB98-106024INZ.
In July 1999, we contracted with RAND to update its earlier
research. The update included an analysis of Functional Independence
Measure (FIM) data, the Function Related Groups (FRGs), and the model
rehabilitation PPS using 1996 and 1997 data. The purpose of updating
the earlier research was to develop the underlying data necessary to
support the Medicare IRF PPS based on CMGs for the November 3, 2000
proposed rule (65 FR at 66313). RAND expanded the scope of its earlier
research to include the examination of several payment elements, such
as comorbidities, facility-level adjustments, and implementation
issues, including evaluation and monitoring. Then, to develop the
provisions of the August 7, 2001 final rule (66 FR 41316, 41323), RAND
did similar analysis on calendar year 1998 and 1999 Medicare Provider
Analysis and Review (MedPAR) files and patient assessment data.
We have continued to contract with RAND to help us identify
potential refinements to the IRF PPS. The refinements we proposed to
make to the IRF PPS, and which we are finalizing in this final rule,
are based on the analyses and recommendations from RAND. In addition,
RAND sought advice from a technical expert panel (TEP), which reviewed
their methodology and findings.
2. Data Files Used for Analysis of the Current IRF PPS
RAND conducted updated analyses of the patient classification
system, case mix and coding changes, and facility-level adjustments for
the IRF PPS using data from calendar year 2002 and FY 2003. This is the
first time CMS or RAND has had data generated by IRFs after the
implementation of the IRF PPS that are available for data analysis.
Public comments and our responses on RAND's research to support the
proposed refinements are summarized below:
Comment: Several commenters expressed concerns about basing the
refinements that we proposed in the FY 2006 proposed rule (70 FR 30188)
on analyses of calendar year 2002 and FY 2003 data, which do not
reflect IRF case mix changes currently taking place in response to our
recent enforcement of the classification criterion, commonly known as
the ``75 percent rule.'' These commenters suggested that we wait for
analysis of future data (CY 2005 or beyond) to become available before
implementing refinements to the IRF PPS.
Response: As discussed in the August 7, 2001 final rule (66 FR
41316), we used RAND's analysis of calendar year 1998 and 1999 Medicare
Provider Analysis and Review (MedPAR) files and patient assessment data
to develop the initial classification system and prospective payment
amounts for the IRF PPS. These data were from a period of time before
the IRF PPS when IRFs' reimbursement was based on costs, subject to
certain limits, rather than on prospective payment amounts.
Furthermore, we used the best available 1998 and 1999 data from a time
period that also preceded enforcement of the 75 percent rule
requirements. Today, we have 2002 and 2003 data that represents all
Medicare-covered IRF cases in a post-PPS environment and, therefore,
portrays a recent and complete picture of IRFs' patient populations. In
addition, the IRF payment system has undergone a major transformation
since the 1998 and 1999 data in the form of a change from a cost-based
payment system to a PPS that became effective with the cost reporting
periods beginning on or after January 1, 2002. Because of this
transformation, we believe the data we have on which to base
refinements to the IRF PPS will help ensure that IRF PPS payments
accurately reflect the costs of care in an IRF.
This is because these data allow RAND to obtain precision in their
analyses, and ensures that the data are not over- or under-representing
particular types of facilities or patients. We believe it is
appropriate and necessary to implement refinements to the IRF PPS at
this time, based on the best available data we have from calendar year
2002 and FY 2003. Since analysis of this data indicates that we have an
opportunity at this time, through the proposed refinements, to improve
the alignment between IRF payments and the cost of care, we believe it
is important to proceed with the refinements discussed in this final
rule.
However, we agree with the commenters that we should continue to
collect the best available data we can to monitor the IRF PPS and
ensure that IRF payments are appropriately aligned with costs of care
and that Medicare patients continue to have appropriate access to IRF
services. We will, whenever necessary, use the best data available in
the future to propose appropriate refinements that will further improve
the alignment between IRF payments and the costs of care. Thus, to the
extent changes in case mix occur due to enforcement of the 75 percent
rule, these changes should appear in later data that we will use to
propose refinements in the future.
Comment: Several commenters noted that 98 IRF providers in RAND's
analysis data affiliated with HealthSouth decided to omit home office
cost data from the 2002 and 2003 cost reports that were filed with us.
The commenters questioned whether this omission might have affected the
results of RAND's analysis and, therefore, our proposed policies.
Response: After publication of the FY 2006 proposed rule (70 FR
30188), we learned that 98 providers in our data file that were
affiliated with HealthSouth omitted home office cost data from the 2002
and 2003 cost reports that were filed with us and that RAND used in the
analysis of the FY 2006 proposed rule (70 FR 30188). These data were a
voluntary omission on the part of these providers, but nevertheless
affect some of the distributional policies (that is, the proposed
teaching status adjustment, the proposed changes to the rural and LIP
adjustments, and the proposed change to the outlier threshold)
contained in the proposed rule. However, because RAND used the
hospital-specific relative value method (that is, the methodology that
effectively controls for inter-hospital variation while estimating the
relative costs of different types of patients within each hospital) for
all of the proposed changes to the classification system described in
section V of this final rule (that is, the proposed changes to the tier
comorbidities, the proposed changes to the CMG definitions, the
proposed weighted motor score methodology, the proposed change to the
coding of the transfer-to-toilet item, and the proposed update of the
relative weights), these proposed changes would not have been affected
by the omission of the home office cost data. In other words, RAND
examined the relative costs of patients within each IRF, so the fact
that the omission of HealthSouth's home office costs caused total costs
to be understated in the cost report data would not have mattered for
the proposed classification system changes described in section V of
this final rule.
In addition, the omission of the home office cost data would have
no effect on the proposed 1.9 percent reduction to the standard payment
amount (discussed in section VI.A of this final rule) because cost
report data were not
[[Page 47885]]
used in the analysis that supports this proposed reduction.
Although the omission of the home office cost data, in theory,
could have had some effect on the estimates of the proposed FY 2002-
based RPL market basket (discussed in section VI.B.1 of this final
rule), our Office of the Actuary conducted some preliminary analyses of
the effects on the market basket calculation and, based on these
analyses, determined that these effects would likely be small. Home
office costs represent only one of many cost categories (including, but
not limited to, salaries, benefits, professional liability insurance,
and pharmacueticals) that are used to develop the cost category
weights. We believe the absence of HealthSouth home office costs in
this market basket has a minor impact on the distribution of these
weights and, by extension, the final market basket update itself. Thus,
we did not believe it was necessary to recalculate the market basket.
Finally, since the facility-level adjustments we proposed in the FY
2006 proposed rule (70 FR 30188) were calculated using regression
analysis based on the relative total costs associated with care in
different types of IRFs (that is, urban/rural, teaching/non-teaching,
low DSH percentage/high DSH percentage), the omission of HealthSouth's
home office costs had some effect on the results of these analyses. The
largest example is for the cost differential between urban and rural
facilities in our analysis. Since the providers that omitted the home
office cost data were largely urban facilities, their lower reported
total cost data caused the differential between urban and rural
facilities to be larger in the initial analyses. The same was true, to
a lesser extent, with the teaching status adjustment and the LIP
adjustment.
Furthermore, the omission of the home office cost data caused
overall reported costs to be lower in these facilities and, therefore,
affected the cost-to-charge ratios computed for these facilities for
FYs 2002 and 2003. We used these cost-to-charge ratios to determine the
proposed update to the outlier threshold amount. Therefore, analysis of
the data indicates that the outlier threshold amount we proposed in the
FY 2006 proposed rule (70 FR 30188) was affected by the omission of the
home office cost data.
Given that the facility-level adjustments, such as the rural, LIP,
and teaching status adjustments, and the outlier threshold amount for
all IRFs were likely affected by the decision of this one large for-
profit chain provider to omit home office cost data from the FY 2002
and FY 2003 cost reports, we believe it is appropriate for us to
recalculate the values for these adjustments and for the outlier
threshold using data that accounts for the omitted home office costs.
Thus, we obtained the FY 2004 HealthSouth home office cost statement
and, from this cost report statement, compiled the home office cost
data for each of the individual HealthSouth IRF providers listed. Of
the 98 providers that omitted home office cost data for FYs 2002 and
2003, 92 of the providers have had home office cost data reported on
the FY 2004 home office cost statement; and six providers did not have
any home office cost information for FY 2004.
We considered several options with respect to incorporating the
missing HealthSouth home office costs into the data RAND used to
conduct the analyses for this final rule. First, we considered the
option of removing all of the HealthSouth cost report data from the
analysis and re-computing the facility-level adjustments (that is, the
rural adjustment, the LIP adjustment, and the teaching status
adjustment) and the outlier threshold without the HealthSouth cost
report data. Dropping all of the cost report data for 98 of the 1,188
facilities in RAND's analysis file, especially when they are large
urban facilities, would seem to skew the data even further because we
would be leaving out a substantial amount of cost report data connected
with one specific type of IRF provider (i.e., urban IRFs). Leaving out
the data for these facilities would make other types of IRFs that are
left in the data appear to have more of an effect on the regression
analysis than they actually do. Since we were hoping to reduce the bias
in the data, rather than increase the bias, we generally rejected this
option.
The second option we considered was to update the analysis using FY
2004 data for all providers and re-compute the facility-level
adjustments and the outlier threshold using the FY 2004 cost report
data. Unfortunately, the FY 2004 data have only recently been submitted
by all IRF providers, and it would have been impossible for RAND and
CMS to have completed all the necessary re-analysis of all of the
proposed policies with the FY 2004 cost report data for all IRF
providers in time for the proposed policies to be implemented in FY
2006.
The third option we considered was to use the FY 2004 home office
cost data that we were able to obtain from the HealthSouth home office
cost statement for 92 of the 98 HealthSouth IRF providers, standardize
all of the other cost report data from FY 2003 for the 98 HealthSouth
providers and the other non-HealthSouth providers using the most recent
market basket for FY 2004, and fill in the FY 2004 home office cost
data for the 92 HealthSouth providers for which we had data. This
option enabled us to meet the October 1 implementation date of our
updates as well as to make those updates and payment adjustments as
accurate as possible. Next, we considered two options for treating the
six HealthSouth facilities for which we did not have FY 2004 home
office cost data: We considered leaving those six IRFs' cost data as
is, without adding any home office cost data since we had none from FY
2004 to add. The other option we considered for treating these six
facilities was to take the average home office costs as a percentage of
total costs for the 92 facilities (which came to approximately 13
percent) and use this as an estimate of home office costs for the 6
facilities. We chose the second of the two options, which meant that we
inflated total costs for those six facilities by the average of about
13 percent, because it seemed inappropriate to ignore the fact that
cost data was missing for these six facilities and 13 percent appeared
to be a reasonable estimate of home office costs generally for IRFs
(from the general analysis we were able to perform).
Because we believe the data file that results from the third option
is more complete than the data RAND previously used to compute the
proposed facility-level adjustments and the proposed outlier threshold
amount for the FY 2006 proposed rule (70 FR 30188), we used the data
from the third option described above to re-compute the values for the
teaching status adjustment (described in more detail in section VI.B.3
of this final rule), the rural adjustment (described in more detail in
section VI.B.4 of this final rule), the LIP adjustment (described in
more detail in section VI.B.5 of this final rule), and the outlier
threshold amount (described in more detail in section VI.B.6 of this
final rule). Because the values of these adjustments have changed, we
also re-computed the budget neutrality factors and, thus, the standard
payment conversion factor.
Comment: Several commenters requested that we make IRF claims data,
IRF-PAI data, patient-specific CMG data, and cost report files
available to the public so that the public would have the opportunity
to recreate the analyses used in developing the proposed refinements
for the FY 2006 proposed rule (70 FR 30188).
Response: The data files mentioned by the commenters are generally
available (and were generally available
[[Page 47886]]
during the comment period for the FY 2006 proposed rule (70 FR 30188))
to the public through CMS's standard data distribution systems. More
information on CMS's data distribution policies is available on CMS's
website at http://www.cms.hhs.gov/researchers/statsdata.asp.
Comment: A few commenters requested that we make available RAND's
research using FY 2003 data. They noted that 3 of the 4 reports
published on RAND's website for public access are based on analysis of
calendar year 2002 data. One of RAND's publicly available reports is
based on analysis of FY 2003 data.
Response: We asked RAND to use the best available, most current
data possible for the analyses contained in the FY 2006 proposed rule
(70 FR 30188) and this final rule. This was generally FY 2003 data.
The updated analysis is generally not contained in RAND's reports,
and RAND has indicated to CMS that they have no plans to publish the
updated analyses (using the FY 2003 data) after publication of the
final rule. However, RAND informed us that, in all of the FY 2003
analyses for the FY 2006 proposed rule (70 FR 30188) and for this final
rule, they used the identical methodologies presented in the reports
available on RAND's website and reviewed by RAND's technical expert
panel. The only change was that RAND used updated data from FY 2003
(and FY 2004 HealthSouth home office cost data, as discussed above).
Thus, interested parties should examine the reports available on RAND's
website for the detailed methodology used to develop the proposed and
final revisions. In addition, interested parties may contact RAND
directly for more information regarding the analysis of FY 2003 data.
Comment: One commenter asked whether a large number of short period
cost reports for periods ending in 2001 might have affected RAND's
research findings and, if so, how RAND handled this issue in the data.
Response: We were unable to find any reasons for the unusually
large number of short period cost reports the commenter is indicating
for cost report periods ending in 2001. However, since some of RAND's
analysis for this final rule was based on calendar year 2002 data, and
the majority of RAND's analysis for this final rule was based on FY
2003 data, we do not believe that a spike in the number of short period
cost reports in 2001 would have had an effect on RAND's analyses.
V. Refinements to the Patient Classification System
A. Changes to the IRF Classification System
1. Development of the IRF Classification System
Section 1886(j)(2)(A)(i) of the Act, as amended by section 125 of
the Medicare, Medicaid, and SCHIP Balanced Budget Refinement Act of
1999 requires the Secretary to establish ``classes of patient
discharges of rehabilitation facilities by functional-related groups
(each referred to as a case-mix group or CMG), based on impairment,
age, comorbidities, and functional capability of the patients, and such
other factors as the Secretary deems appropriate to improve the
explanatory power of functional independence measure-function related
groups.'' In addition, the Secretary is required to establish a method
of classifying specific patients in IRFs within these groups as
specified in Sec. 412.620.
In the August 7, 2001 final rule (66 FR at 41342), we implemented a
methodology to establish a patient classification system using CMGs.
The CMGs are based on the FIM-FRG methodology and reflect refinements
to that methodology.
In general, a patient is first placed in a major group called a
rehabilitation impairment category (RIC) based on the patient's primary
reason for inpatient rehabilitation, (for example, a stroke). The
patient is then placed into a CMG within the RIC, based on the
patient's ability to perform specific activities of daily living, and
sometimes the patient's cognitive ability and/or age. Other special
circumstances, such as the occurrence of very short stays, or cases
where the patient expired, are also considered in determining the
appropriate CMG.
We explained in the August 7, 2001 final rule that further analysis
of FIM and Medicare data may result in refinements to CMGs. In the
August 7, 2001 final rule, we used the most recent FIM and Medicare
data available at that time (that is 1998 and 1999 data). Developing
the CMGs with the 1998 and 1999 data resulted in 95 CMGs based on the
FIM-FRG methodology. The data also supported the establishment of five
additional special CMGs that improved the explanatory power of the FIM-
FRGs. We established one additional special CMG to account for very
short stays and four additional special CMGs to account for cases where
the patient expired. In addition, we established a payment of an
additional amount for patients with at least one relevant comorbidity
in certain CMGs.
2. Description and Methodology Used To Develop the IRF Classification
System in the August 7, 2001 Final Rule
a. Rehabilitation Impairment Categories
In the first step to develop the CMGs, the FIM data from 1998 and
1999 were used to group patients into RICs. Specifically, the
impairment code from the assessment instrument used by clients of UDSmr
and Healthsouth indicates the primary reason for the inpatient
rehabilitation admission. This impairment code is used to group the
patient into a RIC. Currently, we use 21 RICs for the IRF PPS.
b. Functional Status Measures and Age
After using the RIC to define the first division among the
inpatient rehabilitation groups, we used functional status measures and
age to partition the cases further. In the August 7, 2001 final rule,
we used 1998 and 1999 Medicare bills with corresponding FIM data to
create the CMGs and more thoroughly examine each item of the motor and
cognitive measures. Based on the data used for the August 7, 2001 final
rule, we found that we could improve upon the CMGs by making a slight
modification to the motor measure. We modified the motor measure by
removing the transfer to tub/shower item because we found that an
increase in a patient's ability to perform functional tasks with less
assistance for this item was associated with an increase in cost,
whereas an increase in other functional items decreased costs. We
describe below the statistical methodology (Classification and
Regression Trees (CART)) that we used to incorporate a patient's
functional status measures (modified motor score and cognitive score)
and age into the construction of the CMGs in the August 7, 2001 final
rule.
We used the CART methodology to divide the rehabilitation cases
further within each RIC. (Further information regarding the CART
methodology can be found in the seminal literature on CART
(Classification and Regression Trees, Leo Breiman, Jerome Friedman,
Richard Olshen, Charles Stone, Wadsworth Inc., Belmont CA, 1984: pp.
78-80).) We chose to use the CART method because it is useful in
identifying statistical relationships among data and, using these
relationships, constructing a predictive model for organizing and
separating a large set of data into smaller, similar groups. Further,
in constructing the CMGs, we analyzed the extent to which the
independent
[[Page 47887]]
variables (motor score, cognitive score, and age) helped predict the
value of the dependent variable (the log of the cost per case). The
CART methodology creates the CMGs that classify patients with
clinically distinct resource needs into groups. CART is an iterative
process that creates initial groups of patients and then searches for
ways to divide the initial groups to decrease the clinical and cost
variances further and to increase the explanatory power of the CMGs.
Our current CMGs are based on historical data. In order to develop a
separate CMG, we need to have data on a sufficient number of cases to
develop coherent groups. Therefore, we are removing these codes from
the tiers that increase payment.
c. Comorbidities
Under the statutory authority of section 1886(j)(2)(C)(i) of the
Act, we proposed to make several changes to the comorbidity tiers
associated with the CMGs for comorbidities that are not positively
related to treatment costs, or their excessive use is questionable, or
their condition could not be differentiated from another condition.
Specifically, section 1886(j)(2)(C)(i) of the Act provides the
following: The Secretary shall from time to time adjust the
classifications and weighting factors established under this paragraph
as appropriate to reflect changes in treatment patterns, technology,
case mix, number of payment units for which payment is made under this
title and other factors that may affect the relative use of resources.
The adjustments shall be made in a manner so that changes in aggregate
payments under the classification system are a result of real changes
and are not a result of changes in coding that are unrelated to real
changes in case mix.
A comorbidity is a specific patient condition that is secondary to
the patient's principal diagnosis or impairment that is used to place a
patient into a RIC. A patient could have one or more comorbidities
present during the inpatient rehabilitation stay. Our analysis for the
August 7, 2001 final rule found that the presence of a comorbidity
could have a major effect on the cost of furnishing inpatient
rehabilitation care. We also stated that the effect of comorbidities
varied across RICs, significantly increasing the costs of patients in
some RICs, while having no effect in others. Therefore, for the August
7, 2001 final rule, we linked frequently occurring comorbidities to
impairment categories in order to ensure that all of the chosen
comorbidities were not an inherent part of the diagnosis that assigns
the patient to the RIC.
Furthermore, in the August 7, 2001 final rule, we indicated that
comorbidities can affect cost per case for some of the CMGs, but not
all. When comorbidities substantially increased the average cost of the
CMG and were determined to be clinically relevant (not inherent in the
diagnosis in the RIC), we developed CMG relative weights adjusted for
comorbidities (Sec. 412.620(b)).
d. Development of CMG Relative Weights
Section 1886(j)(2)(B) of the Act requires that an appropriate
relative weight be assigned to each CMG. Relative weights account for
the variance in cost per discharge and resource utilization among the
payment groups and are a primary element of a case-mix adjusted PPS.
The establishment of relative weights helps ensure that beneficiaries
have access to care and receive the appropriate services that are
commensurate to other beneficiaries that are classified in the same
CMG. In addition, prospective payments that are based on relative
weights encourage provider efficiency and, hence, help ensure a fair
distribution of Medicare payments. Accordingly, as specified in Sec.
412.620(b)(1), we calculate a relative weight for each CMG that is
proportional to the resources needed by an average inpatient
rehabilitation case in that CMG. For example, cases in a CMG with a
relative weight of 2, on average, will cost twice as much as cases in a
CMG with a relative weight of 1. We discuss the details of developing
the relative weights below.
As indicated in the August 7, 2001 final rule, we believe that the
RAND analysis has shown that CMGs based on function-related groups
(adjusted for comorbidities) are effective predictors of resource use
as measured by proxies such as length of stay and costs. The use of
these proxies is necessary in developing the relative weights because
data that measure actual nursing and therapy time spent on patient
care, and other resource use data, are not available.
e. Overview of Development of the CMG Relative Weights
As indicated in the August 7, 2001 final rule, to calculate the
relative weights, we estimate operating (routine and ancillary
services) and capital costs of IRFs. For this final rule as we
indicated in the FY 2006 proposed rule (70 FR 30188), we use the same
method for calculating the cost of a case that we outlined in the
August 7, 2001 final (66 FR at 41351 through 43153). We obtained cost-
to-charge ratios for ancillary services and per diem costs for routine
services from the most recent available cost report data. We then
obtain charges from Medicare bill data and derived corresponding
functional measures from the FIM data. We omit data from rehabilitation
facilities that are classified as all-inclusive providers from the
calculation of the relative weights, as well as from the parameters
that we use to define transfer cases, because these facilities are paid
a single, negotiated rate per discharge and therefore do not maintain a
charge structure. For ancillary services, we calculate both operating
and capital costs by converting charges from Medicare claims into costs
using facility-specific, cost-center specific cost-to-charge ratios
obtained from cost reports. Our data analysis for the August 7, 2001
final rule showed that some departmental cost-to-charge ratios were
missing or found to be outside a range of statistically valid values.
For anesthesiology, a value greater than 10, or less than 0.01, is
found not to be statistically valid. For all other cost centers, values
greater than 10 or less than 0.5 are found not to be statistically
valid. In the August 7, 2001 final rule, we replaced individual cost-
to-charge ratios outside of these thresholds. The replacement value
that we used for these aberrant cost-to-charge ratios was the mean
value of the cost-to-charge ratio for the cost-center within the same
type of hospital (either freestanding or unit). For routine services,
per diem operating and capital costs are used to develop the relative
weights. In addition, per diem operating and capital costs for special
care services are used to develop the relative weights. (Special care
services are furnished in intensive care units. We note that less than
1 percent of rehabilitation days are spent in intensive care units.)
Per diem costs are obtained from each facility's Medicare cost report
data. We use per diem costs for routine and special care services
because, unlike for ancillary services, we could not obtain cost-to-
charge ratios for these services from the cost report data. To estimate
the costs for routine and special care services included in developing
the relative weights, we sum the product of routine cost per diem and
Medicare inpatient days and the product of the special care per diem
and the number of Medicare special care days.
In the August 7, 2001 final rule, we used a hospital specific
relative value method to calculate relative weights.
[[Page 47888]]
For the FY 2006 proposed rule (70 FR 30188) and this final rule, we
used the following basic steps to calculate the relative weights as
indicated in the August 7, 2001 final rule (at 66 FR 41316, 41351
through 41352).
The first step in calculating the CMG weights is to estimate the
effect that comorbidities have on costs. The second step required us to
adjust the cost of each Medicare discharge (case) to reflect the
effects found in the first step. In the third step, the adjusted costs
from the second step were used to calculate ``relative adjusted
weights'' in each CMG using the hospital-specific relative value
method. The final steps are to calculate the CMG relative weights by
modifying the ``relative adjusted weight'' with the effects of the
existence of the comorbidity tiers (explained below) and normalizing
the weights to 1.
Our methodology for determining the IRF classification system
remains unchanged from the August 7, 2001 final rule.
B. Changes to the Existing List of Tier Comorbidities
1. Changes To Remove Codes That Are Not Positively Related to Treatment
Costs
While our methodology for this final rule for determining the tiers
remains unchanged from the August 7, 2001 final rule, as we indicated
in the FY 2006 proposed rule (70 FR 30188), RAND's analysis indicates
that 1.6 percent of FY 2003 cases received a tier payment (often in
tier one) that was not justified by any higher cost for the case.
Therefore, under statutory authority section 1886(j)(2)(C)(i) of the
Act, as we proposed in the FY 2006 proposed rule (70 FR 30188) we are
implementing several technical changes to the comorbidity tiers
associated with the CMGs. Specifically, the RAND analysis found that
the first 17 diagnoses shown in Table 1 below are no longer positively
related to treatment cost after controlling for CMG. The additional two
codes were also problematic. According to RAND, code 410.91 (AMI, NOS,
Initial) was not specific enough to be differentiated from other
related codes and code 260, Kwashiorkor, was found to be
unrealistically represented in the data according to the RAND technical
expert panel.
With respect to the eighteenth code in Table One, (410.X1) Specific
AMI, initial), we note that RAND found there is no clinical reason to
believe that this code differs in a rehabilitation environment from all
of the specific codes for initial AMI of the form 410.X1, where X is an
numeric digit. In other words, this code is indistinguishable from the
seventeenth code in Table One (410.91 AMI, NOS, initial). Following
this observation, RAND tested the other initial AMI codes as a single
group and found that they have no positive effect on case cost. Thus,
as we indicated in the FY 2006 proposed rule (70 FR 30188), we proposed
to remove ``AMI, NOS, initial'' from the tier list because it is not
positively related to treatment cost after controlling for the CMG. In
addition, for similar reasons, we proposed in the FY 2006 proposed rule
(70 FR 30188) to remove ``Specific AMI, initial from the tier list
since it is indistinguishable from ``AMI, NOS, initial.''
As we proposed in the FY 2006 proposed rule (70 FR 30188), with
respect to the last code in Table One (Kwashiorkor), we are removing
this code from the tier list as well. This comorbidity is positively
related to cost in our data. However, RAND's technical expert panel
(TEP) found the large number of cases coded with this rare disease to
be unrealistic and recommended that it be removed from the tier list.
Table 1 contains two malnutrition codes, and as we proposed in the
FY 2006 proposed rule (70 FR 30188), we are removing these two
malnutrition codes. As we stated in the FY 2006 Proposed Rule (70 FR
30188), removal of these codes where use is concentrated in specific
hospitals is particularly important because these hospitals are likely
receiving unwarrantedly high payments due to the tier one assignment of
these cases. Thus, because we believe the excess use of these two
comorbid conditions is inappropriate based on the findings of RAND's
TEP, they will be removed.
The data indicate large variation in the rate of increase from the
1999 data to the 2003 data across the conditions that make up the
tiers. The greatest increases were for miscellaneous throat conditions
and malnutrition, each of which were more than 10 times as frequent in
2003 as in 1999. The growth in these two conditions was far larger than
for any other condition. Many conditions, however, more than doubled in
frequency, including dialysis, cachexia, obesity, and the non-renal
complications of diabetes. The condition with the least growth, renal
complications of diabetes, may have been affected by improved coding of
dialysis.
As we proposed in the FY 2006 proposed rule (70 FR 30188), we are
finalizing changes to our initial list of diagnoses that deal with
tracheostomy cases. These rare cases were excluded from the pulmonary
RIC 15 in the August 7, 2001 final rule. The new data indicate that
they are more expensive than other cases in the same CMG in RIC 15, as
well as in other RICs. Therefore, we believe the data demonstrate that
tracheostomy cases should be added to the tier list for RIC 15 in order
to receive a higher payment. Finally, the new data indicate that DX
V55.0, ``attention to tracheostomy'' should be part of this condition
as these cases were and are as expensive as other tracheostomy cases.
Thus, since ``attention to tracheostomy'' is as expensive as other
tracheostomy cases, it is logical to group such similar cases together.
Therefore, we are finalizing our proposal to remove the RIC 15
exclusion for code V55.0 (attention to tracheostomy) so that code V55.0
can receive appropriate payment for the additional costs it incurs.
As we stated in the FY 2006 proposed rule (70 FR 30188), we believe
that the data provided by RAND support the removal of the codes in
Table 1 below because they either have no impact on cost after
controlling for their CMG or are indistinguishable from other codes or
are unrealistically overrepresented. Therefore, we are finalizing our
proposed policy to remove these codes from the tier list.
Table 1.--List of Codes To Be Removed From the Tier List
------------------------------------------------------------------------
ICD-9-CM code Abbreviated code title Condition
------------------------------------------------------------------------
235.1............. Unc behav neo oral/phar.. Miscellaneous throat
conditions.
933.1............. Foreign body in larynx... Miscellaneous throat
conditions.
934.1............. Foreign body bronchus.... Miscellaneous throat
conditions.
530.0............. Achalasia & cardiospasm.. Esophegeal conditions.
530.3............. Esophageal stricture..... Esophageal conditions.
530.6............. Acquired esophag Esophageal conditions.
diverticulum.
[[Page 47889]]
V46.1 *........... Dependence on respirator. Ventilator status.
799.4............. Cachexia................. Cachexia.
V49.75............ Status amputation below Amputation of LE.
knee.
V49.76............ Status amputation above Amputation of LE.
knee.
V49.77............ Status amputation hip.... Amputation of LE.
356.4............. Idiopathic progressive Meningitis and
polyneuropathy. encephalitis.
250.90............ Diabetes II, w Non-renal complications
unspecified of diabetes.
complications, not
stated as uncontrolled.
250.93............ Diabetes I, w unspecified Non-renal complications
complications, of diabetes.
uncontrolled.
261............... Nutritional Marasmus..... Malnutrition.
262............... Other severe protein Malnutrition.
calorie deficiency.
410.91............ AMI, NOS, initial........ Major comorbidities.
410.X1............ Specific AMI, initial.... Major comorbidities.
260............... Kwashiorkor.............. Malnutrition.
------------------------------------------------------------------------
* V46.11 and V46.12 were not in existence when the data used in the
analysis was collected. Since these codes are subcategories of code
V46.1 (the code we proposed to remove from the tiers that make
additional payment), they will be removed from the comorbidity tiers
as well.
We received numerous comments on the proposed changes to the
existing list of tier comorbidities which are summarized below:
Comment: One commenter remarked that kwashiorkor should be omitted
from the list of comorbidities to be deleted from the list of
comorbidities that increase the payment rate of the CMG because some of
the software packages used by the industry allow this code to be used
for the coding of the inpatient's comorbidities.
Response: We disagree with the commenter. Kwashiorkor is a severe
malnutrition of infants and young children, primarily in tropical and
subtropical regions, caused by deficiency in the quality and quantity
of protein in the diet. It is characterized by anemia, edema, potbelly,
loss of pigment in the skin, hair loss or change in hair color,
hypoalbuminemia, and bulky stools containing undigested food. In
addition, an inpatient with this condition most likely would not be
able to receive the three hours of intensive rehabilitation that is a
qualifying guideline to be an inpatient within an IRF. While protein
deficiencies may be noted in patients within an IRF, by definition, the
incidence of Kwashiorkor could not be as high as reported. Also, as
previously stated, RAND's TEP reported that the data indicate large
variation in the rate of increase across conditions. However, coding of
malnutrition increased by more than 10 times, and RAND found the large
number of cases coded with this rare disease to be unrealistic and
recommended that it be removed from the tier list. Consequently,
kwashiorkor will be eliminated from the list of comorbidities that
increase the payment rate of the CMG.
Comment: One commenter wrote that code V46.1 is listed in the
proposed list of codes to be removed from the tier list. Since this
code contains two other codes, the commenter wanted to know if it is
our intention to remove both codes in this category, namely V46.11
(Dependence on respirator, status) and V46.12 (Encounter for respirator
dependence during power failure) or just one of these codes.
Response: First, we want to explain how codes V46.11 and V46.12
became codes that are used to increase the CMG payment rate. In the
August 7, 2001 final rule (66 FR 41316), we published Appendix C that
listed the ICD-9-CM comorbid condition codes which are used to increase
the CMG payment rate. The ICD-9-CM codes of the comorbid conditions are
recorded by the IRF's staff on the IRF-PAI, and that data as well as
some other data recorded on the IRF-PAI is used to classify an
inpatient into a CMG payment rate. One of the codes we published as
part of Appendix C was V46.1. Each year the codes used in the ICD-9-CM
coding system undergo a review resulting in updates to some of the
existing codes. In accordance with a review that updated the ICD-9-CM
coding system V46.11 and V46.12 were added to the ICD-9-CM coding
system as subcategories of V46.1. We believe that the comorbid
condition represented by the code V46.11 or V46.12 is a derivative of
the comorbid condition represented by the code V46.1. Therefore, in
2005 we updated the CMG grouper software which resulted in the CMG
payment being increased by the same amount if the IRF-PAI data of an
inpatient included codes V46.1, or V46.11, or V46.12.
The analysis that our data contractor performed, using certain data
after the IRF PPS was implemented, shows that the comorbid condition
represented by code V46.1 does not have an effect upon treatment cost
after controlling for the CMG. Therefore, code V46.1 and its derivative
codes that comprise it (V46.11 and V46.12) will be removed from the
list of codes that are used by the IRF PPS to increase the CMG payment
rate.
Comment: Several commenters urged us to consider not removing codes
V49.75, V49.76, and V49.77 from the list of comorbidity codes that
increase the CMG payment because of concerns with the complexity of a
patient with an amputation.
Response: After controlling for the CMG, RAND found that these
codes do not impact cost. Further, IRFs do not incur additional costs
to treat these comorbidities after controlling for the CMG. This means
that the CMG to which the inpatient is assigned, already accounts for
the costs associated with the treatment of inpatients with an
amputation and no additional payment is needed beyond the CMG amount to
adequately reimburse for such a case. Therefore we are removing these
codes from the list of comorbidities that increase the CMG payment.
Comment: Several commenters mentioned a concern with the code
V497.7 in the table of codes to be removed. They believed it to be a
typographical error where the actual code to be removed is V49.77.
Response: We agree with the commenters and have made the correction
to the typographical error. The corrected code to be removed is V49.77.
Comment: Several commenters noted that there is a discrepancy with
code 428.3 (vocal cord paralysis, not otherwise specified) in CMS' list
of
[[Page 47890]]
codes being reassigned based on their marginal cost in the Comorbidity
Tier Reassignment Changes File found at http://www.cms.hhs.gov/providers/irfpps/fy06nprm.asp.
They stated that it should actually be
code 478.30 (vocal cord paralysis, not otherwise specified).
Response: We agree with the commenters and shall make the
appropriate corrections to the typographical error within the file.
Comment: Several commenters noted an error with the description of
meningitis and encephalitis for code 356.4 in the Comorbidity Tier
Reassignment Changes File found at http://www.cms.hhs.gov/providers/irfpps/fy06nprm.asp
.
Response: We agree with the commenters and the description will be
amended to read idiopathic progressive polyneuropathy for code 356.4.
Comment: Commenters expressed concern for the removal of codes
530.0 (achalasia and cardiospasm), 530.3 (stricture and stenosis of
esophagus) and 530.6 (diverticulum of esophagus) that are used to
record esophageal conditions because of costs associated with these
conditions and requested that they not be removed from the tier list
which increases payment for these comorbidities.
Response: After controlling for the CMG, RAND found that these
comorbidities do not positively impact costs, meaning that the CMG
encompasses sufficient payment to compensate for these comorbidities.
Therefore, we are removing codes 530.0, 530.3 and 530.6 from the list
of comorbidities that increase CMG payment.
Comment: Several commenters agreed with CMS' proposed policy to
remove malnutrition codes 261 (nutritional marasmus) and 262 (other
severe protein-calorie malnutrition), while others opposed the proposed
policy to remove these codes. In addition, several commenters suggested
that CMS examine the impact of malnutrition on increasing the length of
stay within an IRF.
Response: We acknowledge both opinions as expressed by the
different commenters. The RAND TEP, and our Medical Officers, believes
these codes are drastically overstated and inpatients with these levels
of malnutrition would not be candidates for three hours of intensive
therapy. In addition, after controlling for the CMG, both of these
codes do not positively affect payment. Therefore we believe it is
appropriate to remove malnutrition codes 261 and 262 from the list of
comorbidity codes that are used to increase the CMG payment rate.
Additionally, we will continue to examine the impact of comorbidities,
including malnutrition, upon IRF Medicare-covered inpatients.
Comment: One commenter suggested adding codes 250.91 and 250.92 to
the list of comorbidities to be removed from the list of codes used to
increase payment because they believe those codes to be similar in
description to codes 250.90 and 250.93.
Response: Only the first 17 codes within Table 1 were found to have
no positive effect on cost after controlling for the CMG. The data
analysis performed by RAND does not indicate that at this time 250.91
and 250.93 should be removed from the list of codes used to increase
the CMG payment rate because they continue to positively affect costs.
Therefore we believe it is inappropriate to remove them from the list
of comorbidities that impact cost. Consequently, we are not removing
any other codes from the list of codes used to increase the CMG payment
rate.
Comment: One commenter recommended that several codes be added to
our comorbidity tier system based upon suggestions from the RAND TEP,
namely codes 428.0 (congestive heart failure), V43.3 (heart valve
replacement), 250.1 (insulin dependent diabetes without mention of
complications, not stated as controlled) and 438.2X (hemi-paresis due
to an old stroke).
Response: After examining the RAND recommendations, our Medical
Officers felt that codes V43.3 and 438.2X were too vague and non-
descript to capture the necessary information needed for these codes to
be added to the list of codes used to increase the CMG payment rate.
However, in response to the comments our Medical Officers re-evaluated
the effect on cost by the comorbid condition represented by code 250.1
(insulin dependent diabetes without mention of complications, not
stated as controlled). They determined that code 250.1 should be added
to the list of codes used to increase the CMG payment rate. They also
determined that the code should be a tier 3 code because the other 250
series of codes related to diabetes are in tier 3. Therefore, this code
will be added as a tier 3 code to the list of codes used to increase
the CMG payment rate. There will be no excluded RICs with code 250.1.
After examining the comments, our Medical Officers continue to believe
that 428.9 (heart failure, unspecified), was too non-descript and
should not be added to the list of codes that can increase payment.
However, our Medical Officers agree with the commenter regarding other
numerous congestive heart failure codes including Code 428.1--Left
Heart Failure, Code 428.20--Systolic Heart Failure Unspecified, Code
428.21--Systolic Heart Failure Acute, Code 428.22--Systolic Heart
Failure Chronic, Code 428.23--Systolic Hear Failure Acute on Chronic,
Code 428.30--Diastolic Heart Failure Unspecified, Code 428.31--
Diastolic Heart Failure Acute, Code 428.32--Diastolic Heart Failure
Chronic, Code 428.33--Diastolic Heart Failure Acute on Chronic, Code
428.40--Combined Systolic and Diastolic Heart Failure Unspecified, Code
428.41--Combined Systolic and Diastolic Heart Failure Acute, Code
428.42--Combined Systolic and Diastolic Heart Failure Chronic, and Code
428.43--Combined Systolic and Diastolic Heart Failure Acute on Chronic,
largely due to the increased costs associated with these codes.
Therefore, these 428 cardiac codes will be added to the list of codes
used to increase the CMG payment rate as tier 3 codes because of their
similarity to certain cardiac codes with respect to resource
utilization. However, these codes will not be used to increase the CMG
payment rate if the CMG code is one of the CMG codes derived from RIC
14 (the cardiac RIC) because these cardiac codes costs have been
accounted for in the CMGs associated with RIC 14.
Comment: A commenter believes that the CMG payment rate should
include an adjustment for mental health problems, such as a depression.
The commenter believes that a patient's mental health status has an
effect on the patient treatment costs an IRF incurs.
Response: The significance and appropriateness of a patient's state
of mental health in response to an impairment that requires a patient
to undergo intensive inpatient rehabilitation is a subject that we
believe requires further study. Additional study will help to determine
the effect of the patient's state of mental health on treatment costs.
An ICD-9-CM code may be used to show that a patient is exhibiting signs
that a rehabilitation clinician believes indicate a mental disorder.
However, quantifying by use of ICD-9-CM codes the association between a
patient's state of mental health and how it affects a patient's
response to rehabilitation treatment is at best limited. For example,
we believe that in response to a stroke or hip fracture, or some other
impairment, a situational depression may be a rational response.
However, that does not mean that the IRF will incur additional costs
that were not already taken into account when the CMG payment rates
were developed. In addition, mental disorders vary greatly
[[Page 47891]]
in severity as does how a patient's functioning is affected by a mental
disorder.
There would have to be multiple factors taken into consideration
before any type of mental disorder could be added to the list of
comorbidities that would increase payment of the CMG. The data for a
complete psychiatric evaluation must be made available to correctly
code for these comorbidities. In addition, this is a budget neutral
system, and no additional funding will be added to the system. Under
our final rule, funds will not be added but simply be redistributed
among the comorbidities among the tiers that increase payment. This is
because the changes associated with the comorbidity tiers and CMGs are
done in a budget neutral manner. On the assumption that there is an
even distribution of these psychiatric patients among IRFs, and these
patients may receive the redistributed payment, the addition of these
codes may not contribute to an increased payment for inpatients with
these comorbid conditions and may affectively lower payments for CMG's
with other comorbid conditions because the same amount of funding is
distributed across more comorbid conditions. Also, few IRFs have
psychiatric personnel and rehabilitation doctors rarely have the time
required to observe the patient to make a complete psychiatric
evaluation and thus some codes may be assigned (or not assigned) in
error. In addition, RAND's TEP believed that it would be inappropriate
to use ICD-9-CM diagnoses to identify patients with affective
disorders. Therefore, in this final rule, we are not adding codes for
depression and mental disorders to the list of codes used to increase
payment.
Comment: We received comments to both challenge and support the
removal of certain comorbidity codes from the tier list including code
799.4 Cachexia, and code 933.1 (foreign body in larynx). Commenters
stated that these conditions required more resources, and thus
increased treatment costs. The other commenter stated that the CMG
already covered these costs.
Response: The data analysis did not show that the comorbid
conditions indicated by these codes increased the costs of treating an
inpatient with these comobidities after controlling for the CMG because
their CMG payment rate covers costs associated with their corresponding
treatment. The more recent RAND analysis found that after controlling
for the CMG, these comobidities do not impact cost. Therefore, we are
removing them from the comorbidity tiers that would increase payment.
Comment: One commenter made a general statement stating that the
list of comorbidities that comprise the tiers do not reflect the
challenges that contribute to higher costs in the rehabilitation
setting.
Response: We disagree with the commenter because the RAND
regression analyses show that the comorbid conditions that comprise the
tiers positively impact cost and provide additional payments for
services not included in the payment associated with the CMG.
Final Decision: In this final rule, we are adopting the proposal to
remove the comorbidity tier codes set forth in Table 1 of the FY 2006
proposed rule (70 FR 30188). We are also removing codes V46.11 and
V46.12 because they are subcategories of code V46.1, which has been
found to have no impact on cost after controlling for the CMG. We are
adding several codes that the RAND analyses found to positively impact
costs. We chose to add codes 250.1 (insulin dependent diabetes without
mention of complications, not stated as controlled), as well as
numerous congestive heart failure codes including Code 428.1--Left
Heart Failure, Code 428.20--Systolic Heart Failure Unspecified, Code
428.21--Systolic Heart Failure Acute, Code 428.22--Systolic Heart
Failure Chronic, Code 428.23--Systolic Heart Failure Acute on Chronic,
Code 428.30--Diastolic Heart Failure Unspecified, Code 428.31--
Diastolic Heart Failure Acute, Code 428.32--Diastolic Heart Failure
Chronic, Code 428.33--Diastolic Heart Failure Acute on Chronic, Code
428.40--Combined Systolic and Diastolic Heart Failure Unspecified, Code
428.41--Combined Systolic and Diastolic Heart Failure Acute, Code
428.42--Combined Systolic and Diastolic Heart Failure Chronic, and Code
428.43--Combined Systolic and Diastolic Heart Failure Acute on Chronic,
which our Medical Officers believe were specific enough to be used in
our list of codes that are used to increase the CMG payment amount.
2. Changes To Move Dialysis to Tier One
As we proposed in the FY 2006 proposed rule (70 FR 30188), we are
finalizing the movement of dialysis from comorbidity tier two to
comorbidity tier one, which is the tier associated with the highest
payment. The data from the RAND analysis show that patients on dialysis
cost more than the tier payment to which dialysis is currently
assigned, and should be moved into the highest paid tier because this
tier would more closely align payment with the cost of a case. Based on
RAND's analysis using 2003 data, a patient with dialysis costs 31
percent more than a non-dialysis patient in the same CMG and with the
same other accompanying comorbidities.
Overall, the largest increase in the cost of a condition occurs
among patients on dialysis, where the coefficient in the cost
regression increases by 93 percent, from 0.1400 to 0.2697. Part of the
explanation for the increased coefficient could be that some IRFs had
not borne all dialysis costs for their patients in the pre-PPS period,
which was the previous data analysis time period(because providers were
previously permitted to bill for dialysis separately). It is likely
that, in the 1999 data, some IRFs had not borne all dialysis costs for
their patients. Because the fraction of cases coded with dialysis
increased by 170 percent, it is also likely that improved coding was
part of the explanation for the increased coefficient. We believe a 170
percent increase is such a dramatic increase that it would be highly
unlikely that in the time periods used for the data analysis, 170
percent more patients needed dialysis when compared to the time period
before the implementation of the IRF PPS. We also believe that the
improved coding is likely due to the fact that higher costs are
associated with dialysis patients, and therefore IRFs, in an effort to
ensure that their payments cover these higher expenses better and more
carefully coded comorbidities whose presence resulted in higher PPS
payments.
Therefore we are moving dialysis patients to comorbidity tier one
will more adequately compensate IRFs for the extra cost of those
patients and thereby maintain or increase access to these services.
Comment: A number of commenters supported our decision to move
dialysis patients to tier one due to the increase cost of dialysis
patients.
Response: We agree with these commenters. The data analyses
performed by RAND found evidence that suggested that a dialysis patient
cost 31 percent more than a non-dialysis patient in the same CMG.
Therefore, as proposed in the FY 2006 proposed rule (70 FR 30188), we
are moving dialysis to tier 1 because the additional payment associated
with tier 1 more closely approximate the additional costs associated
with the treatment of an inpatient with this condition.
Final Decision: As proposed in the FY 2006 proposed rule (70 FR
30188), we are adopting the decision to move dialysis patients to
comorbidity tier one.
[[Page 47892]]
3. Changes To Move Comorbidity Codes Based on Their Marginal Cost
Under section 1886(j)(2)(C)(i) of the Act, as was proposed in the
FY 2006 proposed rule (70 FR 30188), we are refining how we pay for a
comorbidity based on marginal cost. A commonly understood definition of
marginal cost is the increase or decrease in costs as a result of one
higher or lower unit of a good or service. In this situation, we are
reassigning comorbidities to tiers based on their marginal costs, and
by this we mean the increase or decrease in costs as a result of one
higher or lower comorbidity tier. Payment for several comorbidities
would be more accurate if their tier assignments were changed, and
after examining RAND's data, we believe that of the FY 2003 cases, a
full 4 percent of cases should be associated with comorbidity tiers
that have a lower payment than the comorbidity tiers to which they were
assigned. Therefore, comorbidities would be more accurate if their tier
assignments were more appropriately based on their marginal costs.
As we proposed in the FY 2006 proposed rule (70 FR 30188),
comorbidity tier assignments in this final rule are based on the
results of statistical analyses RAND has performed under contract with
CMS, using as independent variables only the CMGs and conditions for
tiers. As we proposed in the FY 2006 proposed rule (70 FR 30188), tier
assignments of each of these conditions for the final rule are
determined based on the magnitude of their coefficients in RAND's
statistical analysis.
We believe the IRF PPS led to substantial changes in coding of
comorbidities between 1999 (pre-implementation of the IRF PPS) and 2003
(post-implementation of the IRF PPS). The percentage of cases with one
or more comorbidities increased from 16.79 percent according to the
data used to define the comorbidity tiers (1998 through 1999) to 25.51
percent in FY 2003. This is an increase of 52 percent in tier incidence
(52 = 100 x (25.51-16.79)/16.79). The recording of a tier one
comorbidity, the highest paid of the tiers, almost quadrupled during
this same time period. Although, improved coding likely increased the
recording of comorbidities, those coding the comorbidities may have
been motivated by the objective to use coding changes as a means to
increase the CMG payment.
The 2003 data provides an excellent comprehensive picture of the
costs that are associated with each of the comorbidities. We believe
this because CMS has data for 100 percent of the Medicare-covered IRF
cases. Therefore, as we indicated in the FY 2006 proposed rule, we
believe that using the 2003 data to assign the comorbidities to a
payment tier ensures heightened accuracy with respect to the matching
of payments to relative costs of a case.
We received several comments on the proposed changes to the
existing list identifying which tier is associated with a particular
comorbidity. The public comments are summarized below.
Comment: One commenter suggested that we postpone reassigning
comorbidity tiers based on their marginal costs, and again instead
perform the data analysis used to reassign the comorbidity codes based
on marginal costs using more current data.
Response: This final rule reflects the most recent analysis of
data. In the future, we will continue to perform data analyses and, as
necessary, adjust the payment rates to achieve the most accurate
payment. In this final rule, we are adopting the policy we proposed in
the FY 2006 proposed rule (70 FR 30188), and reassigning comorbidities
to tiers based on their marginal cost because we believe that this
reassignment is based on the best comprehensive post-PPS implementation
data that are available at this time.
Comment: One commenter recommended that we not reassign any
comorbidity codes based on their marginal costs under the premise that
there is no concrete evidence of upcoding.
Response: Taking into consideration that we believe that there has
been improved coding due to prospective payment based system, the
recommendations of RAND's technical expert panel, and the guidance of
our Medical Officers, we believe that the comorbidity codes should be
assigned based on their marginal costs in order to increase the
association between costs and payment.
Final Decision: In summary, we are adopting all of the proposals
set forth in the FY 2006 proposed rule (70 FR 30188), with regard to
the removal of the list of codes from comorbidity tiers that increase
payment, the movement of dialysis patients to tier one, the code V55.0
will no longer be excluded from RIC 15, and comorbidity codes will now
be reassigned based on their marginal costs.
C. Changes to the CMGs
Section 1886(j)(2)(C)(i) of the Act requires the Secretary from
time to time to adjust the classifications and weighting factors of
patients under the IRF PPS to reflect changes in treatment patterns,
technology, case mix, number of payment units for which payment is
made, and other factors that may affect the relative use of resources.
These adjustments shall be made in a manner so that changes in
aggregate payments under the classification system are the result of
real changes and not the result of changes in coding that are unrelated
to real changes in case mix.
In the FY 2006 proposed rule (70 FR 30188, 30196), in accordance
with section 1886(j)(2)(C)(i) of the Act and as specified in Sec.
412.620(c) and based on the research conducted by RAND, we proposed to
update the CMGs used to classify IRF patients for purposes of
establishing payment amounts. We also proposed to update the relative
weights associated with the payment groups based on FY 2003 Medicare
bill and patient assessment data. We proposed replacing the current
unweighted motor score index used to assign patients to CMGs with a
weighted motor score index that would improve our ability to accurately
predict the costs of caring for IRF patients, as described in detail
below. However, we proposed not to change the methodology for computing
the cognitive score index.
As described in the August 7, 2001 final rule, we contracted with
RAND to analyze IRF data to support our efforts in developing our
patient classification system and the IRF PPS. We continued our
contract with RAND to support us in developing potential refinements to
the classification system and the PPS. As part of this research, we
asked RAND to examine possible refinements to the CMGs to identify
potential improvements in the alignment between Medicare payments and
actual IRF costs. In conducting its research, RAND used a technical
expert panel (TEP) made up of experts from industry groups, other
government entities, academia, and other interested parties. The
technical expert panel reviewed RAND's methodologies and advised RAND
on many technical issues.
Several recent developments make significant improvements in the
alignment between Medicare payments and actual IRF costs possible.
First, when the IRF PPS was implemented in 2002, a new assessment
instrument was used to collect patient data, the IRF Patient Assessment
Instrument (IRF-PAI). The new instrument contained items that improved
the quality of the patient-level information available to researchers.
Second, more recent data are available on a larger patient
population. Until now, the design of the IRF PPS was based entirely on
1999 data on Medicare
[[Page 47893]]
rehabilitation patients from just a sample of hospitals (the best
available data at the time). Now, we have post-PPS data from 2002 and
2003 that describe the entire universe of Medicare-covered
rehabilitation patients.
Finally, we believe that improvements in the algorithms that
produced the initial CMGs, as described below, should lead to new CMGs
that better predict treatment costs in the IRF PPS.
Using the inpatient rehabilitation facility assessment instrument
before the PPS, which is commonly referred to as the FIM, and Medicare
data from 1998 and 1999, RAND helped us develop the original structure
of the IRF PPS. IRFs became subject to the PPS beginning with cost
reporting periods starting on or after January 1, 2002. The PPS is
based on assigning patients to particular CMGs that are designed to
predict the costs of treating particular Medicare patients according to
how well they function in four general categories: Transfers, sphincter
control, self-care (for example, grooming, eating), and locomotion.
Patient functioning is measured according to 18 categories of activity:
13 motor tasks, such as putting on clothing, and 5 cognitive tasks,
such as memory. The PPS is intended to align payments to IRFs as
closely as possible with the actual costs of treating patients. If the
PPS ``underpays'' for some kinds of care, IRFs have incentives to limit
access for patients requiring that kind of care because payments for a
particular case would be less than the costs of providing care, so an
IRF may try to limit its financial ``losses''; conversely, if the PPS
overpays, resources are wasted because IRFs' payments exceed the costs
of providing care for a particular case.
The fiscal year 2003 data file currently available for refining the
CMGs contains many more IRF cases and represents the universe of
Medicare-covered IRF cases, rather than a sample. The best available
data that CMS and RAND had for analysis in 1999 contained 390,048 IRF
cases, representing 64 percent of all Medicare-covered patients in
participating IRFs. The more recent data contain 523,338 IRF cases
(fiscal year 2003), representing all Medicare-covered patients in
participating IRFs. The larger file enables RAND to obtain greater
precision in the analysis and portrays a more recent and complete
picture of patients under the IRF PPS.
Also, the fiscal year 2003 data include more detailed information
about patients' level of functioning. For example, new variables are
included in the more recent data that provide further details on
patient functioning. Standard bowel and bladder scores on the FIM
instrument (used to assess patients before the IRF PPS), for example,
measured some combination of the level of assistance required and the
frequency of accidents (that is, soiling of clothes and surroundings).
New variables on the IRF-PAI instrument measure the level and the
frequency separately. Since measures of the level of assistance
required and the frequency of accidents contain slightly different
information about the expected costliness of an IRF patient, having
measures for these two variables separately provides additional
information to researchers.
Furthermore, additional optional information is recorded on the
health status of patients in the more recent data (for example,
shortness of breath, presence of ulcers, inability to balance).
1. Changes for Updating the CMGs
In the FY 2006 proposed rule (70 FR 30188), we proposed to revise
the definitions of the CMGs based on regression analysis by RAND of the
FY 2003 data. As described in the August 7, 2001 final rule, RAND
developed the original list of CMGs using FIM data from 1998 and 1999
(see the FY 2006 proposed rule (70 FR 30188, 30198 through 30202) for a
table of the original CMG listing).
Given the availability of more recent, post-PPS data, we asked RAND
to examine possible refinements to the CMGs to identify potential
improvements in the alignment between Medicare payments and actual IRF
costs. In addition to analyzing fiscal year 2003 data, RAND also
convened a TEP, made up of researchers from industry, provider
organizations, government, and academia, to provide support and
guidance through the process of developing possible refinements to the
PPS. Members of the TEP reviewed drafts of RAND's reports, offered
suggestions for additional analyses, and provided clinicians' views of
the importance and significance of various findings.
As we explained in the FY 2006 proposed rule (70 FR 30188), RAND's
analysis of the FY 2003 data, along with the support and guidance of
the TEP, strongly suggested the need to update the CMGs to better align
payments with costs under the IRF PPS. The other option we considered
before proposing to update the CMGs with the fiscal year 2003 data was
to maintain the same CMG structure but recalculate the relative weights
for the current CMGs using the 2003 data. After carefully reviewing the
results of RAND's regression analysis, which compared the predictive
ability of the CMGs under 3 scenarios (not updating the CMGs or the
relative weights, updating only the relative weights and not the CMGs,
and updating both the relative weights and the CMGs), as we stated in
the FY 2006 proposed rule (70 FR 30188), we believed and continue to
believe (based on RAND's analysis) that updating both the relative
weights and the CMGs will allow the classification system to do a
better job of reflecting changes in treatment patterns, technology,
case mix, and other factors which may affect the relative use of
resources.
We continue to believe it is appropriate to update both the CMGs
and the relative weights at this time because the 2003 data we now have
represent a more recent and broader set of data elements. The more
recent data include all Medicare-covered IRF cases rather than a
subset, allowing us to base the CMG changes on a complete picture of
the types of patients in IRFs. In designing the IRF PPS, we used the
best available data, but those data may not have contained a complete
picture of the types of patients in IRFs. Also, the improved clinical
coding of patient conditions in IRFs is better reflected in the more
recent data than it was in the best available data we had to design the
IRF PPS. In addition, changes in treatment patterns, technology, case
mix, and other factors affecting the relative use of resources in IRFs
since the IRF PPS was implemented likely require an update to the
classification system.
Prior to the finalization of the proposed changes contained in this
final rule, we paid IRFs based on 95 CMGs and 5 special CMGs developed
using the CART algorithm applied to 1999 data. The CART algorithm that
was used in designing the IRF PPS assigned patients to RICs according
to their age and their motor and cognitive FIM scores. CART produced
the partitions so that the reported wage-adjusted rehabilitation cost
of the patients was relatively constant within partitions. Then, a
subjective decision-making process was used to decrease the number of
CMGs (to ensure that the payment system did not become unduly
complicated), to enforce certain constraints on the CMGs (to ensure
that, for instance, IRFs were not paid more for patients who had fewer
comorbidities than for patients with more comorbidities), and to fit
the comorbidity tiers. Although the use of a subjective decision-making
process (rather than a computer algorithm) was very useful, there were
limitations. For example, it made it difficult to explore
[[Page 47894]]
the implications of variations to the CART models because an individual
person is not able to examine as many variations of a model in as short
a period of time as a computer program. Furthermore, the computer is
more efficient at accounting for all of the possible combinations and
interactions between important variables that affect patient costs.
In analyzing potential refinements to the IRF PPS, RAND created a
new algorithm that would be very useful in constructing the CMGs (the
new algorithm would be based on the CART methodology described in
detail in section V.A.2.b of this final rule). RAND applied the new
algorithm to the fiscal year 2003 IRF data. In the FY 2006 proposed
rule (70 FR 30188), we proposed to use RAND's new algorithm for
refinements to the CMGs. The algorithm is based entirely on an
iterative computerized process to decrease the number of CMGs, enforce
constraints on the CMGs, and assign the comorbidity tiers. At each step
in the process, the new CART algorithm produces all of the possible
combinations of CMGs using all available variables. It then selects the
variables and the CMG constructions that offer the best predictive
ability, as measured by the greatest decrease in the mean-squared
error. We proposed to place the following constraints on the algorithm,
based on RAND's analysis: (1) Neighboring CMGs would have to differ by
at least $1,500, unless eliminating the CMG would change the estimated
costs of patients in that CMG by more than $1,000; (2) estimated costs
for patients with lower motor or cognitive index scores (more
functionally dependent) would always have to be higher than estimated
costs for patients with higher motor or cognitive index scores (less
functionally dependent). We believe that the PPS should not pay more
for a patient who is less functionally dependent than for one who is
more functionally dependent; and (3) each CMG must contain at least 50
observations (for statistical validity).
RAND's technical expert panel, which included representatives from
industry groups, other government entities, academia, and other
researchers, reviewed and commented on these constraints and the rest
of RAND's proposed methodology (developed based on RAND's analysis of
the data) for updating the CMGs as RAND developed the improvements to
the CART methodology.
The following are the most substantial differences between the CMGs
used prior to October 1, 2005 and the proposed new CMGs for FY 2006:
Fewer CMGs than before (87 now compared with 95 in the
prior system). The 5 special CMGs for very short stay cases and cases
in which the patient expires would remain unchanged.
The number of CMGs under the RIC for stroke patients (RIC
1) would decrease from 14 to 10.
The cognitive index score would affect patient
classification in two of the RICs (RICs 1 and 2), whereas it previously
affected RICs 1, 2, 5, 8, 12, and 18.
A patient's age would now affect assignment for CMGs in
RICs 1, 4, and 8, whereas it previously affected assignment for CMGs in
RICs 1 and 4.
The primary objective in updating the CMGs is to better align IRF
payments with the costs of caring for IRF patients, given more recent
information. This requires that we improve the ability of the system to
predict patient costs. RAND's analysis suggests that the proposed new
CMGs clearly improve the ability of the payment system to predict
patient costs. The proposed new CMGs would greatly improve the
explanation of variance in the system.
Public comments and our responses on the proposed changes for
updating the CMGs are summarized below.
Comment: Several commenters raised concerns that the FY 2003 data
used to update the CMGs did not reflect the full enforcement of the 75
percent rule and that CMS should, therefore, wait until the data
reflect full enforcement before making any changes to the CMGs.
Response: We agree that additional changes to the CMGs may
potentially be necessary in the future if enforcement of the 75 percent
rule results in substantial changes to IRFs' patient populations.
However, we believe it is now appropriate to begin refining the system
because several recent developments make significant improvements in
the alignment between Medicare payments and actual IRF costs possible.
First, when the IRF PPS was implemented for cost reporting periods
beginning on or after January 1, 2002, a new recording instrument
called the IRF-PAI was used to collect patient data. The new instrument
contained questions that improved the quality of the patient-level
information available to researchers. The 2003 data used in the
proposed refinements reflects this data.
Second, more recent data are available on a larger patient
population. Until now, the design of the IRF PPS was based entirely on
1999 data on Medicare rehabilitation patients from just a sample of
hospitals. Even though this was the best available data at the time, we
now have post-PPS data from 2002 and 2003 that describe the entire
universe of Medicare-covered rehabilitation patients.
Finally, we believe that proposed improvements in the algorithms
that produced the initial CMGs, as described above, lead to new CMGs
that better predict treatment costs in the IRF PPS.
We further note that making refinements to the IRF patient
classification system now, based on post-PPS data, does not preclude us
from making future refinements to the system if IRFs' case mix and care
practices change over time. We will continue to monitor the IRF PPS,
and make refinements as needed, to ensure that IRF payments are aligned
as closely as possible with the costs of providing care.
Comment: One commenter believed that the proposed changes to the
CMGs would make IRF quality measurement more difficult over time
because the proposed changes to the CMG definitions would mean that a
case classified into a particular CMG (such as CMG 0107) before October
1, 2005 (when the proposed changes would be implemented) would not
necessarily be classified into CMG 0107 after October 1, 2005. Thus,
people attempting to create a one-for-one crosswalk between the CMGs
before October 1, 2005 and the proposed CMGs after October 1, 2005
would be unable to do so. The commenter noted that many quality
measurement tools currently being used by IRFs require such a one-for-
one crosswalk.
Response: We recognize the importance of monitoring IRF quality of
care over time. However, we do not believe that the proposed changes to
the CMGs inhibit the ability to monitor quality in IRFs over time.
Quality of care is not measured by a payment rate, but by data
reflecting various indicators of the treatment patients receive. In the
FY 2006 proposed rule (70 FR 30188), we did not propose changes to the
patient assessment form itself or changes to the coding of the
underlying data that is used to classify patients into CMGs. Therefore,
comparisons of the underlying patient classification data could still
be used to monitor quality in these facilities over time.
Comment: One commenter expressed concerns that the cognitive scores
are not used as often in the definitions of the proposed revisions to
the CMGs as they were in the original CMGs defined in the August 7,
2001 final rule. This commenter stated that the cognitive scores are
important predictors of how costly patients are likely to be in the IRF
setting. The commenter also stated that,
[[Page 47895]]
if cognitive scores are not used as often as motor scores for assigning
patients to CMGs, the reason may be that measures of patients'
cognitive abilities may not currently be as well developed as measures
of patients' motor abilities. Therefore, this commenter recommended
that we develop more sensitive measures that have better predictive
qualities.
Response: As we noted previously, the cognitive score used to
classify IRF patients into CMGs is made up of cognitive items from the
IRF-PAI. These cognitive items are generally indications of the
patient's mental functioning level, and are related to the patient's
ability to process and respond to empirical factual information, use
judgment, and accurately perceive what is happening. Patients'
cognitive functioning clearly affects their expected costliness in an
IRF. However, RAND's regression analysis, in which they explored the
relationship of the FIM motor and cognitive scores to cost, showed that
patients' cognitive scores generally did not predict patients' expected
costliness above and beyond what patients' motor scores already were
able to predict. Thus, we see no reason to use cognitive scores in CMG
definitions for which they do not add predictive ability. When the
cognitive scores add information that increases the predictive ability
of the classification system, we make use of this information in the
CMG assignment.
We agree with one of the commenter's points that the cognitive
score may not predict costs as well as the motor score because the
cognitive items may not be as sensitive to patients' cognitive status
as the motor items are to patients' physical functioning. We further
agree with the commenter that more work could be done to better
identify measures of cognitive functioning. Along these lines, CMS has
awarded a contract to the Research Triangle Institute (RTI) to perform
research and data analysis to support possible changes to the IRF-PAI
instrument that would better capture physical and cognitive functioning
information on IRF patients. CMS remains open to examining well-
constructed peer-reviewed studies by other types of providers,
researchers, and other interested parties in order to improve upon the
cognitive assessment functioning measures for the Medicare population.
Until then, we will use the best cognitive functioning information
available for IRF patients to classify patients into the most
appropriate CMGs so IRF payments align as closely as possible with the
costs of care in IRFs.
Final Decision: After carefully considering all the comments we
received on the proposed changes to the CMG definitions, we are
finalizing our decision to adopt the CMG definitions presented below in
Table 2. Based on RAND's regression analysis of FY 2003 data, the best
data available for analysis, we believe these changes will increase the
accuracy of IRF PPS payments.
Table 2.--Case Mix Groups (CMGs), With the Associated Rehabilitation Impairment Categories (RICs)
[Beginning with discharges on or after October 1, 2005]
--------------------------------------------------------------------------------------------------------------------------------------------------------
RIC CMG No. CMG description
--------------------------------------------------------------------------------------------------------------------------------------------------------
01 Stroke (Stroke)........................ 0101 Motor >51.05.
0102 Motor >44.45 & Motor <51.05 & Cognitive >18.5.
0103 Motor >44.45 & Motor <51.05 & Cognitive <18.5.
01 Stroke (Stroke)........................ 0104 Motor >38.85 & Motor <44.45.
0105 Motor >34.25 & Motor <38.85.
0106 Motor >30.05 & Motor <34.25.
0107 Motor >26.15 & Motor <30.05.
0108 Motor <26.15 & Age >84.5.
0109 Motor >22.35 & Motor <26.15 & Age <84.5.
0110 Motor < 22.35 & Age < 84.5.
02 Traumatic brain injury (TBI)........... 0201 Motor >53.35 & Cognitive >23.5.
0202 Motor >44.25 & Motor <53.35 & Cognitive >23.5.
0203 Motor >44.25 & Cognitive <23.5.
0204 Motor >40.65 & Motor <44.25.
0205 Motor >28.75 & Motor <40.65.
0206 Motor >22.05 & Motor <28.75.
0207 Motor < 22.05.
03 Nontraumatic brain injury (NTBI)....... 0301 Motor >41.05.
0302 Motor >35.05 & Motor <41.05.
0303 Motor >26.15 & Motor <35.05.
0304 Motor < 26.15.
04 Traumatic spinal cord injury (TSCI).... 0401 Motor >48.45.
0402 Motor >30.35 & Motor <48.45.
0403 Motor >16.05 & Motor <30.35.
0404 Motor <16.05 & Age >63.5.
0405 Motor < 16.05 & Age < 63.5.
05 Nontraumatic spinal cord injury (NTSCI) 0501 Motor >51.35.
05 Nontraumatic spinal cord injury (NTSCI) 0502 Motor >40.15 & Motor <51.35.
0503 Motor >31.25 & Motor <40.15.
0504 Motor >29.25 & Motor <31.25.
0505 Motor >23.75 & Motor <29.25.
0506 Motor < 23.75.
06 Neurological (Neuro)................... 0601 Motor >47.75.
0602 Motor >37.35 & Motor <47.75.
0603 Motor >25.85 & Motor <37.35.
0604 Motor < 25.85.
07 Fracture of LE (FracLE)................ 0701 Motor >42.15.
0702 Motor >34.15 & Motor <42.15.
0703 Motor >28.15 & Motor <34.15.
0704 Motor < 28.15.
08 Replacement of LE joint (RepLE)........ 0801 Motor >49.55.
[[Page 47896]]
0802 Motor >37.05 & Motor <49.55.
0803 Motor >28.65 & Motor <37.05 & Age >83.5.
0804 Motor >28.65 & Motor <37.05 & Age <83.5.
0805 Motor >22.05 & Motor <28.65.
0806 Motor < 22.05.
09 Other orthopedic(Ortho)................ 0901 Motor >44.75.
0902 Motor >34.35 & Motor <44.75.
0903 Motor >24.15 & Motor <34.35.
0904 Motor < 24.15.
10 Amputation, lower extremity (AMPLE).... 1001 Motor >47.65.
1002 Motor >36.25 & Motor <47.65.
1003 Motor < 36.25.
11 Amputation, other (AMP-NLE)............ 1101 Motor >36.35.
11 Amputation, other (AMP-NLE)............ 1102 Motor < 36.35.
12 Osteoarthritis (OsteoA)................ 1201 Motor >37.65.
1202 Motor >30.75 & Motor <37.65.
1203 Motor < 30.75.
13 Rheumatoid, other arthritis (RheumA)... 1301 Motor >36.35.
1302 Motor >26.15 & Motor <36.35.
1303 Motor < 26.15.
14 Cardiac (Cardiac)...................... 1401 Motor >48.85.
1402 Motor >38.55 & Motor <48.85.
1403 Motor >31.15 & Motor <38.55.
1404 Motor < 31.15.
15 Pulmonary (Pulmonary).................. 1501 Motor >49.25.
1502 Motor >39.05 & Motor <49.25.
1503 Motor >29.15 & Motor <39.05.
1504 Motor < 29.15.
16 Pain Syndrome (Pain)................... 1601 Motor >37.15.
1602 Motor >26.75 & Motor <37.15.
1603 Motor < 26.75.
17 Major multiple trauma, no brain injury 1701 Motor >39.25.
or spinal cord injury (MMT-NBSCI).
1702 Motor >31.05 & Motor <39.25.
1703 Motor >25.55 & Motor <31.05.
1704 Motor < 25.55.
18 Major multiple trauma, with brain or 1801 Motor >40.85.
spinal cord injury (MMT-BSCI).
1802 Motor >23.05 & Motor <40.85.
1803 Motor < 23.05.
19 Guillian Barre (GB).................... 1901 Motor >35.95.
19 Guillian Barre (GB..................... 1902 Motor >18.05 & Motor <35.95
1903 Motor < 18.05.
20 Miscellaneous (Misc)................... 2001 Motor >49.15.
2002 Motor >38.75 & Motor <49.15.
2003 Motor >27.85 & Motor <38.75.
2004 Motor < 27.85.
21 Burns (Burns).......................... 2101 Motor >0.
Special CMGs.............................. 5001 Short-stay cases, length of stay is 3 days or fewer.
5101 Expired, orthopedic, length of stay is 13 days or fewer.
5102 Expired, orthopedic, length of stay is 14 days or more.
5103 Expired, not orthopedic, length of stay is 15 days or fewer.
5104 Expired, not orthopedic, length of stay is 16 days or more.
--------------------------------------------------------------------------------------------------------------------------------------------------------
Note: CMG definitions use weighted motor scores, as defined below.
2. Use of a Weighted Motor Score Index and Change to the Treatment of
Unobserved Transfer to Toilet Values
In the FY 2006 proposed rule (70 FR 30188, 30210), we proposed to
use a weighted motor score index in assigning patients to CMGs, instead
of the motor score index previously used that treated all components
equally. We also proposed to change how the IRF PPS GROUPER software
would assign a value for the transfer-to-toilet item when it is coded
by the provider with a 0. We proposed that the software would assign
this item a value of 2 instead of a 1 when the activity is coded by the
provider with a 0. However, we proposed not to change the cognitive
score index. As described in detail below, we continue to believe that
a weighted motor score index, with the change to the scoring of the
transfer to toilet item when the provider records a 0 value for the
activity on the IRF-PAI, will improve the classification of patients
into CMGs, which in turn will improve the accuracy of payments to IRFs.
To classify a patient into a CMG, IRFs use the admission assessment
data from the IRF-PAI to score a patient's functional independence
measures. The
[[Page 47897]]
functional independence measures consist of what are termed ``motor''
items and ``cognitive'' items. In addition to the functional
independence measures, the patient's age may also influence the
patient's CMG classification. The motor items are generally indications
of the patient's physical functioning level. The cognitive items are
generally indications of the patient's mental functioning level, and
are related to the patient's ability to process and respond to
empirical factual information, use judgment, and accurately perceive
what is happening. The motor items are eating, grooming, bathing,
dressing upper body, dressing lower body, toileting, bladder
management, bowel management, transfer to bed/chair/wheelchair,
transfer to toilet, transfer to tub or shower, walking or wheelchair
use, and stair climbing. The cognitive items are comprehension,
expression, social interaction, problem solving, and memory. (The CMS
IRF-PAI manual includes more information on these items.) Each item is
generally recorded on the IRF-PAI and scored on a scale of 0 to 7, with
a 7 indicating complete independence in this area of functioning, a 1
indicating that a patient is very impaired in this area of functioning,
and a 0 indicating that the activity did not occur.
As explained in the August 7, 2001 final rule (66 FR 41349), the
instructions for the IRF-PAI required that providers record an 8 for an
item to indicate that the activity did not occur, as opposed to a 1
through 7 indicating that the activity occurred and the estimated level
of function connected with that activity. However, when the IRF-PAI
form was finalized, the code 8 had been removed and was replaced with
the code 0. Therefore, facilities now record a 0 when an activity does
not occur.
To determine the appropriate payment for patients for whom an
activity is coded as 0 (that is, the activity did not occur), we needed
to decide an appropriate way of changing the 0 to another code for
which payment could be assigned. As discussed in the August 7, 2001
final rule (66 FR at 41349), for purposes of classifying patients into
CMGs, we decided to assign a code of 1 (indicating that the patient
needed ``total assistance'') whenever a code of 0 appeared for one of
the items on the IRF-PAI used to determine payment. This was the most
conservative approach we could have taken based on the best available
data at the time because a value of 1 indicates that the patient needed
total assistance performing the task. The result of recoding a 0 as a 1
and using that value to classify a patient into a CMG is that the
provider might receive a higher payment for that item (although it
might not be the highest payment overall, depending on the patient's
other functional abilities and/or comorbidities).
In the FY 2006 proposed rule (70 FR 30188), we proposed to change
the way we treat a code of 0 on the IRF-PAI for the transfer to toilet
item. This is the only item that we proposed to change at this time
because RAND's regression analysis demonstrated that, of all the motor
score values, the evidence supporting a change in the motor score
values was the strongest with respect to this item. We proposed to
assign a code of 2, instead of a code of 1, to patients for whom a 0 is
recorded on the IR