Medicare and Medicaid Programs; CY 2018 Home Health Prospective Payment System Rate Update and Proposed CY 2019 Case-Mix Adjustment Methodology Refinements; Home Health Value-Based Purchasing Model; and Home Health Quality Reporting Requirements

AGENCY:

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

ACTION:

Proposed rule.

SUMMARY:

This proposed rule updates the home health prospective payment system (HH PPS) payment rates, including the national, standardized 60-day episode payment rates, the national per-visit rates, and the non-routine medical supply (NRS) conversion factor, effective for home health episodes of care ending on or after January 1, 2018. This rule also: updates the HH PPS case-mix weights using the most current, complete data available at the time of rulemaking; implements the 3rd-year of a 3-year phase-in of a reduction to the national, standardized 60-day episode payment to account for estimated case-mix growth unrelated to increases in patient acuity (that is, nominal case-mix growth) between CY 2012 and CY 2014; and discusses our efforts to monitor the potential impacts of the rebasing adjustments that were implemented in CY 2014 through CY 2017. This rule proposes case-mix methodology refinements, as well as a change in the unit of payment from 60-day episodes of care to 30-day periods of care, to be implemented for home health services beginning on or after January 1, 2019; and finally, this rule proposes changes to the Home Health Value-Based Purchasing (HHVBP) Model and to the Home Health Quality Reporting Program (HH QRP).

DATES:

To be assured consideration, comments must be received at one of the addresses provided below, no later than 5 p.m. on September 25, 2017.

ADDRESSES:

In commenting, please refer to file code CMS-1672-P. Because of staff and resource limitations, we cannot accept comments by facsimile (FAX) transmission.

You may submit comments in one of four ways (please choose only one of the ways listed):

1. Electronically. You may submit electronic comments on this regulation to http://www.regulations.gov. Follow the instructions under the “More Search Options” tab.

2. By regular mail. You may mail written comments to the following address ONLY: Centers for Medicare & Medicaid Services, Department of Health and Human Services, Attention: CMS-1672-P, P.O. Box 8016, Baltimore, MD 21244-8016.

Please allow sufficient time for mailed comments to be received before the close of the comment period.

3. By express or overnight mail. You may send written comments to the following address ONLY: Centers for Medicare & Medicaid Services, Department of Health and Human Services, Attention: CMS-1672-P, Mail Stop C4-26-05, 7500 Security Boulevard, Baltimore, MD 21244-1850.

4. By hand or courier. If you prefer, you may deliver (by hand or courier) your written comments before the close of the comment period to either of the following addresses:

a. For delivery in Washington, DC—Centers for Medicare & Medicaid Services, Department of Health and Human Services, Room 445-G, Hubert H. Humphrey Building, 200 Independence Avenue SW., Washington, DC 20201.

(Because access to the interior of the Hubert H. Humphrey Building is not readily available to persons without federal government identification, commenters are encouraged to leave their comments in the CMS drop slots located in the main lobby of the building. A stamp-in clock is available for persons wishing to retain a proof of filing by stamping in and retaining an extra copy of the comments being filed.)

b. For delivery in Baltimore, MD—Centers for Medicare & Medicaid Services, Department of Health and Human Services, 7500 Security Boulevard, Baltimore, MD 21244-1850.

If you intend to deliver your comments to the Baltimore address, please call (410) 786-7195 in advance to schedule your arrival with one of our staff members.

Comments mailed to the addresses indicated as appropriate for hand or courier delivery may be delayed and received after the comment period.

For information on viewing public comments, see the beginning of the SUPPLEMENTARY INFORMATION section.

FOR FURTHER INFORMATION CONTACT:

For general information about the HH PPS, please send your inquiry via email to: HomehealthPolicy@cms.hhs.gov.

For information about the HHVBP model, please send your inquiry via email to: HHVBPquestions@cms.hhs.gov.

Joan Proctor, (410) 786-0949 for information about the home health quality reporting program.

SUPPLEMENTARY INFORMATION:

Inspection of Public Comments: All comments received before the close of the comment period are available for viewing by the public, including any personally identifiable or confidential business information that is included in a comment. We post all comments received before the close of the comment period on the following Web site as soon as possible after they have been received: http://www.regulations.gov. Follow the search instructions on that Web site to view public comments.

Comments received timely will also be available for public inspection as they are received, generally beginning approximately 3 weeks after publication of a document, at the headquarters of the Centers for Medicare & Medicaid Services, 7500 Security Boulevard, Baltimore, Maryland 21244, Monday through Friday of each week from 8:30 a.m. to 4 p.m. EST. To schedule an appointment to view public comments, phone 1-800-743-3951.

Table of Contents

I. Executive Summary

A. Purpose

B. Summary of the Major Provisions

C. Summary of Costs and Benefits

II. Background

A. Statutory Background

B. Current System for Payment of Home Health Services

C. Updates to the Home Health Prospective Payment System

D. Report to Congress: Home Health Study on Access to Care for Vulnerable Patient Populations and Subsequent Research and Analyses

III. Provisions of the Proposed Rule: Payment Under the Home Health Prospective Payment System (HH PPS)

A. Monitoring for Potential Impacts—Affordable Care Act Rebasing Adjustments

B. Proposed CY 2018 HH PPS Case-Mix Weights

C. Proposed CY 2018 Home Health Payment Rate Update

D. Payments for High-Cost Outliers under the HH PPS

E. Proposed Implementation of the Home Health Groupings Model (HHGM) for CY 2019

IV. Proposed Provisions of the Home Health Value-Based Purchasing (HHVBP) Model

A. Background

B. Quality Measures

C. Quality Measures for Future Consideration

V. Proposed Updates to the Home Health Care Quality Reporting Program (HH QRP)

A. Background and Statutory Authority

B. General Considerations Used for the Selection of Quality Measures for the HH QRP

C. Accounting for Social Risk Factors in the HH QRP

D. Proposed Data Elements for Removal From OASIS

E. Proposed Collection of Standardized Patient Assessment Data Under the HH QRP

F. HH QRP Quality Measures Proposed Beginning With the CY 2020 HH QRP

G. HH QRP Quality Measures and Measure Concepts Under Consideration for Future Years

H. Proposed Standardized Patient Assessment Data

I. Proposals Relating to the Form, Manner, and Timing of Data Submission Under the HH QRP

J. Other Proposals for the CY 2019 HH QRP and Subsequent Years

K. Proposals and Policies Regarding Public Display of Quality Measure Data for the HH QRP

L. Proposed Mechanism for Providing Confidential Feedback Reports to HHAs

M. Home Health Care CAHPS® Survey (HHCAHPS)

VI. Request for Information on CMS Flexibilities and Efficiencies

VII. Collection of Information Requirements

A. Statutory Requirement for Solicitation of Comments

B. Collection of Information Requirements for the HH QRP

C. Submission of PRA-Related Comments

VIII. Response to Public Comments

IX. Regulatory Impact Analysis

A. Statement of Need

B. Overall Impact

C. Detailed Economic Analysis

D. Alternatives Considered

E. Accounting Statement and Table

F. Reducing Regulation and Controlling Regulatory Costs

G. Conclusion

X. Federalism Analysis

Regulation Text

Acronyms

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

ACH LOS Acute Care Hospital Length of Stay

ADL Activities of Daily Living

AM-PAC Activity Measure for Post-Acute Care

APU Annual Payment Update

ASPE Assistant Secretary for Planning and Evaluation

BBA Balanced Budget Act of 1997, Public Law 105-33

BBRA Medicare, Medicaid, and SCHIP Balanced Budget Refinement Act of 1999, (Pub. L. 106-113)

BIMS Brief Interview for Mental Status

BLS Bureau of Labor Statistics

CAD Coronary Artery Disease

CAH Critical Access Hospital

CAM Confusion Assessment Method

CARE Continuity Assessment Record and Evaluation

CASPER Certification and Survey Provider Enhanced Reports

CBSA Core-Based Statistical Area

CCN CMS Certification Number

CHF Congestive Heart Failure

CMI Case-Mix Index

CMP Civil Money Penalty

CMS Centers for Medicare & Medicaid Services

CoPs Conditions of Participation

COPD Chronic Obstructive Pulmonary Disease

CVD Cardiovascular Disease

CY Calendar Year

DM Diabetes Mellitus

DRA Deficit Reduction Act of 2005, Public Law 109-171, enacted February 8, 2006

DTI Deep Tissue Injury

EOC End of Care

FDL Fixed Dollar Loss

FI Fiscal Intermediaries

FR Federal Register

FY Fiscal Year

HAVEN Home Assessment Validation and Entry System

HCC Hierarchical Condition Categories

HCIS Health Care Information System

HH Home Health

HHA Home Health Agency

HHCAHPS Home Health Care Consumer Assessment of Healthcare Providers and Systems Survey

HH PPS Home Health Prospective Payment System

HHGM Home Health Groupings Model

HHQRP Home Health Quality Reporting Program

HHRG Home Health Resource Group

HHVBP Home Health Value-Based Purchasing

HIPPS Health Insurance Prospective Payment System

HVBP Hospital Value-Based Purchasing

IADL Instrumental Activities of Daily Living

ICD-9-CM International Classification of Diseases, Ninth Revision, Clinical Modification

ICD-10-CM International Classification of Diseases, Tenth Revision, Clinical Modification

IH Inpatient Hospitalization

IMPACT Act Improving Medicare Care Transformation Act of 2014 (Pub. L. 113-185)

IPR Interim Performance Report

IRF Inpatient Rehabilitation Facility

IRF-PAI IRF Patient Assessment Instrument

IV Intravenous

LCDS LTCH CARE Data Set

LEF Linear Exchange Function

LTCH Long-Term Care Hospital

LUPA Low-Utilization Payment Adjustment

MACRA Medicare Access and CHIP Reauthorization Act of 2015

MAP Measure Applications Partnership

MDS Minimum Data Set

MEPS Medical Expenditures Panel Survey

MFP Multifactor productivity

MMA Medicare Prescription Drug, Improvement, and Modernization Act of 2003, Pub. L. 108-173, enacted December 8, 2003

MSA Metropolitan Statistical Area

MSS Medical Social Services

NQF National Quality Forum

NQS National Quality Strategy

NRS Non-Routine Supplies

OASIS Outcome and Assessment Information Set

OBRA Omnibus Budget Reconciliation Act of 1987, Pub. L. 100-2-3, enacted December 22, 1987

OCESAA Omnibus Consolidated and Emergency Supplemental Appropriations Act, Pub. L. 105-277, enacted October 21, 1998

OES Occupational Employment Statistics

OIG Office of Inspector General

OLS Ordinary Least Squares

OT Occupational Therapy

OMB Office of Management and Budget

PAC Post-Acute Care

PAC-PRD Post-Acute Care Payment Reform Demonstration

PAMA Protecting Access to Medicare Act of 2014

PEP Partial Episode Payment Adjustment

PHQ-2 Patient Health Questionnaire-2

PPOC Primary Point of Contact

PPS Prospective Payment System

PRA Paperwork Reduction Act

PRRB Provider Reimbursement Review Board

PT Physical Therapy

PY Performance Year

QAP Quality Assurance Plan

QIES Quality Improvement Evaluation System

QRP Quality Reporting Program

RAP Request for Anticipated Payment

RF Renal Failure

RFA Regulatory Flexibility Act, Pub. L. 96—354

RHHIs Regional Home Health Intermediaries

RIA Regulatory Impact Analysis

ROC Resumption of Care

SAF Standard Analytic File

SLP Speech-Language Pathology

SN Skilled Nursing

SNF Skilled Nursing Facility

SOC Start of Care

SSI Surgical Site Infection

TEP Technical Expert Panel

TPS Total Performance Score

UMRA Unfunded Mandates Reform Act of 1995.

VAD Vascular Access Device

VBP Value-Based Purchasing

I. Executive Summary

A. Purpose

This proposed rule would update the payment rates for home health agencies (HHAs) for calendar year (CY) 2018, as required under section 1895(b) of the Social Security Act (the Act). This proposed rule would update the case-mix weights under section 1895(b)(4)(A)(i) and (b)(4)(B) of the Act for CY 2018 and implement a 0.97 percent reduction to the national, standardized 60-day episode payment amount to account for case-mix growth unrelated to increases in patient acuity (that is, nominal case-mix growth) between CY 2012 and CY 2014, under the authority of section 1895(b)(3)(B)(iv) of the Act. For home health services beginning on or after January 1, 2019, this rule also proposes case-mix methodology refinements under the authority set out at sections 1895(b)(4)(A)(i) and (b)(4)(B) of the Act, and a change in the unit of payment from a 60-day episode of care to a 30-day period of care under the authority set out at section 1895(b)(2) of the Act. Additionally, this rule proposes changes to: The Home Health Value Based Purchasing (HHVBP) model under the authority of section 1115A of the Act; and the Home Health Quality Reporting Program (HH QRP) requirements under the authority of section 1895(b)(3)(B)(v) of the Act.

B. Summary of the Major Provisions

Section III.A of this rule discusses our efforts to monitor for potential impacts due to the rebasing adjustments implemented in CY 2014 through CY 2017, as mandated by section 3131(a) of the Patient Protection and Affordable Care Act of 2010 (Pub. L. 111-148, enacted March 23, 2010) as amended by the Health Care and Education Reconciliation Act of 2010 (Pub. L. 111-152, enacted March 30, 2010), collectively referred to as the “Affordable Care Act”. In the CY 2015 HH PPS final rule (79 FR 66072), we finalized our proposal to recalibrate the case-mix weights every year with the most current and complete data available at the time of rulemaking. In section III.B of this rule, we are recalibrating the HH PPS case-mix weights, using the most current cost and utilization data available, in a budget neutral manner. Also in section III.B of this rule, as finalized in the CY 2016 HH PPS final rule (80 FR 68624), we are implementing a reduction to the national, standardized 60-day episode payment rate for CY 2018 of 0.97 percent to account for estimated case-mix growth unrelated to increases in patient acuity (that is, nominal case-mix growth) between CY 2012 and CY 2014.

In section III.C of this proposed rule, we would update the payment rates under the HH PPS by 1 percent for CY 2018 in accordance with section 411(d) of the Medicare Access and CHIP Reauthorization Act of 2015 (MACRA) (Pub. L. 114-10, enacted April 16, 2015) which amended section 1895(b)(3)(B) of the Act. Additionally, section III.C of this rule, would update the CY 2018 home health wage index using FY 2014 hospital cost report data. In section III.D of this proposed rule, we note that the fixed-dollar loss ratio would remain 0.55 for CY 2018 to pay up to, but no more than, 2.5 percent of total payments as outlier payments, as required by section 1895(b)(5)(A) of the Act.

In section III.E of this rule we are proposing to implement case-mix methodology refinements and a change in the unit of payment from a 60-day episode of care to a 30-day period of care, effective for home health services beginning on or after January 1, 2019. The proposed home health groupings model (HHGM) relies more heavily on clinical characteristics and other patient information to place patients into meaningful payment categories, while eliminating therapy service use thresholds that are currently used to case-mix adjust payments under the HH PPS. This includes proposed changes in the episode timing categories, the addition of an admission source category, the creation of six clinical groups used to categorize patients based on their primary reason for home health care, revised functional levels and corresponding OASIS items, the addition of a comorbidity adjustment, and a proposed change in the Low-Utilization Payment Adjustment (LUPA) threshold. The LUPA add-on policy, the partial [episode] payment adjustment policy, and the methodology used to calculate payments for high-cost outliers would remain unchanged except for occurring on a 30-day basis rather than a 60-day basis.

In section IV of this rule, we are proposing changes to the Home Health Value-Based Purchasing (HHVBP) Model implemented January 1, 2016. We are proposing to amend the definition of “applicable measure” to specify that the HHA would have to submit a minimum of 40 completed surveys for Home Health Care Consumer Assessment of Healthcare Providers and Systems (HHCAHPS) measures, for purposes of receiving a performance score for any of the HHCAHPS measures, and for performance year (PY) 3 and subsequent years, to remove the Outcome and Assessment Information Set (OASIS)-based measure, Drug Education on All Medications Provided to Patient/Caregiver during All Episodes of Care, from the set of applicable measures. We are also soliciting public comments on composite quality measures for future consideration.

In section V of this rule, we propose updates to the Home Health Quality Reporting Program, including: The replacement of one quality measure, the adoption of two new quality measures, the reporting of standardized patient assessment data in five categories described under the IMPACT Act, data submission requirements, exception and extension requirements, and reconsideration and appeals procedures.

C. Summary of Costs and Benefits

II. Background

A. Statutory Background

The Balanced Budget Act of 1997 (BBA) (Pub. L. 105-33, enacted August 5, 1997), significantly changed the way Medicare pays for Medicare HH services. Section 4603 of the BBA mandated the development of the HH PPS. Until the implementation of the HH PPS on October 1, 2000, HHAs received payment under a retrospective reimbursement system.

Section 4603(a) of the BBA mandated the development of a HH PPS for all Medicare-covered HH services provided under a plan of care (POC) that were paid on a reasonable cost basis by adding section 1895 of the Act, entitled “Prospective Payment For Home Health Services.” Section 1895(b)(1) of the Act requires the Secretary to establish a HH PPS for all costs of HH services paid under Medicare.

Section 1895(b)(3)(A) of the Act requires the following: (1) The computation of a standard prospective payment amount include all costs for HH services covered and paid for on a reasonable cost basis and that such amounts be initially based on the most recent audited cost report data available to the Secretary; and (2) the standardized prospective payment amount be adjusted to account for the effects of case-mix and wage levels among HHAs.

Section 1895(b)(3)(B) of the Act addresses the annual update to the standard prospective payment amounts by the HH applicable percentage increase. Section 1895(b)(4) of the Act governs the payment computation. Sections 1895(b)(4)(A)(i) and (b)(4)(A)(ii) of the Act require the standard prospective payment amount to be adjusted for case-mix and geographic differences in wage levels. Section 1895(b)(4)(B) of the Act requires the establishment of an appropriate case-mix change adjustment factor for significant variation in costs among different units of services.

Similarly, section 1895(b)(4)(C) of the Act requires the establishment of wage adjustment factors that reflect the relative level of wages, and wage-related costs applicable to HH services furnished in a geographic area compared to the applicable national average level. Under section 1895(b)(4)(C) of the Act, the wage-adjustment factors used by the Secretary may be the factors used under section 1886(d)(3)(E) of the Act.

Section 1895(b)(5) of the Act gives the Secretary the option to make additions or adjustments to the payment amount otherwise paid in the case of outliers due to unusual variations in the type or amount of medically necessary care. Section 3131(b)(2) of the Affordable Care Act revised section 1895(b)(5) of the Act so that total outlier payments in a given year would not exceed 2.5 percent of total payments projected or estimated. The provision also made permanent a 10 percent agency-level outlier payment cap.

In accordance with the statute, as amended by the BBA, we published a final rule in the July 3, 2000 Federal Register (65 FR 41128) to implement the HH PPS legislation. The July 2000 final rule established requirements for the new HH PPS for HH services as required by section 4603 of the BBA, as subsequently amended by section 5101 of the Omnibus Consolidated and Emergency Supplemental Appropriations Act for Fiscal Year 1999 (OCESAA), (Pub. L. 105-277, enacted October 21, 1998); and by sections 302, 305, and 306 of the Medicare, Medicaid, and SCHIP Balanced Budget Refinement Act of 1999, (BBRA) (Pub. L. 106-113, enacted November 29, 1999). The requirements include the implementation of a HH PPS for HH services, consolidated billing requirements, and a number of other related changes. The HH PPS described in that rule replaced the retrospective reasonable cost-based system that was used by Medicare for the payment of HH services under Part A and Part B. For a complete and full description of the HH PPS as required by the BBA, see the July 2000 HH PPS final rule (65 FR 41128 through 41214).

Section 5201(c) of the Deficit Reduction Act of 2005 (DRA) (Pub. L. 109-171, enacted February 8, 2006) added new section 1895(b)(3)(B)(v) to the Act, requiring HHAs to submit data for purposes of measuring health care quality, and links the quality data submission to the annual applicable percentage increase. This data submission requirement is applicable for CY 2007 and each subsequent year. If an HHA does not submit quality data, the HH market basket percentage increase is reduced by 2 percentage points. In the November 9, 2006 Federal Register (71 FR 65884, 65935), we published a final rule to implement the pay-for-reporting requirement of the DRA, which was codified at § 484.225(h) and (i) in accordance with the statute. The pay-for-reporting requirement was implemented on January 1, 2007.

The Affordable Care Act made additional changes to the HH PPS. One of the changes in section 3131 of the Affordable Care Act is the amendment to section 421(a) of the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (MMA)(Pub. L. 108-173, enacted on December 8, 2003) as amended by section 5201(b) of the DRA. Section 421(a) of the MMA, as amended by section 3131 of the Affordable Care Act, requires that the Secretary increase, by 3 percent, the payment amount otherwise made under section 1895 of the Act, for HH services furnished in a rural area (as defined in section 1886(d)(2)(D) of the Act) with respect to episodes and visits ending on or after April 1, 2010, and before January 1, 2016.

Section 210 of the MACRA amended section 421(a) of the MMA to extend the rural add-on for 2 more years. Section 421(a) of the MMA, as amended by section 210 of the MACRA, requires that the Secretary increase, by 3 percent, the payment amount otherwise made under section 1895 of the Act, for HH services provided in a rural area (as defined in section 1886(d)(2)(D) of the Act) with respect to episodes and visits ending on or after April 1, 2010, and before January 1, 2018. Section 411(d) of MACRA amended section 1895(b)(3)(B) of the Act such that for home health payments for CY 2018, the market basket percentage increase shall be 1 percent.

B. Current System for Payment of Home Health Services

Generally, Medicare currently makes payment under the HH PPS on the basis of a national, standardized 60-day episode payment rate that is adjusted for the applicable case-mix and wage index. The national, standardized 60-day episode rate includes the six HH disciplines (skilled nursing, HH aide, physical therapy, speech-language pathology, occupational therapy, and medical social services). Payment for non-routine supplies (NRS) is not part of the national, standardized 60-day episode rate, but is computed by multiplying the relative weight for a particular NRS severity level by the NRS conversion factor. Payment for durable medical equipment covered under the HH benefit is made outside the HH PPS payment system. To adjust for case-mix, the HH PPS uses a 153-category case-mix classification system to assign patients to a home health resource group (HHRG). The clinical severity level, functional severity level, and service utilization are computed from responses to selected data elements in the OASIS assessment instrument and are used to place the patient in a particular HHRG. Each HHRG has an associated case-mix weight which is used in calculating the payment for an episode. Therapy service use is measured by the number of therapy visits provided during the episode and can be categorized into nine visit level categories (or thresholds): 0-5; 6; 7-9; 10; 11-13; 14-15; 16-17; 18-19; and 20 or more visits.

For episodes with four or fewer visits, Medicare pays national per-visit rates based on the discipline(s) providing the services. An episode consisting of four or fewer visits within a 60-day period receives what is referred to as a low-utilization payment adjustment (LUPA). Medicare also adjusts the national standardized 60-day episode payment rate for certain intervening events that are subject to a partial episode payment adjustment (PEP adjustment). For certain cases that exceed a specific cost threshold, an outlier adjustment may also be available.

C. Updates to the Home Health Prospective Payment System

As required by section 1895(b)(3)(B) of the Act, we have historically updated the HH PPS rates annually in the Federal Register. The August 29, 2007 final rule with comment period set forth an update to the 60-day national episode rates and the national per-visit rates under the HH PPS for CY 2008. The CY 2008 HH PPS final rule included an analysis performed on CY 2005 HH claims data, which indicated a 12.78 percent increase in the observed case-mix since 2000. Case-mix represents the variations in conditions of the patient population served by the HHAs. Subsequently, a more detailed analysis was performed on the 2005 case-mix data to evaluate if any portion of the 12.78 percent increase was associated with a change in the actual clinical condition of HH patients. We identified 8.03 percent of the total case-mix change as real, and therefore, decreased the 12.78 percent of total case-mix change by 8.03 percent to get a final nominal case-mix increase measure of 11.75 percent (0.1278 * (1−0.0803) = 0.1175).

To account for the changes in case-mix that were not related to an underlying change in patient health status, we implemented a reduction, over 4 years, to the national, standardized 60-day episode payment rates. That reduction was to be 2.75 percent per year for 3 years beginning in CY 2008 and 2.71 percent for the fourth year in CY 2011. In the CY 2011 HH PPS final rule (76 FR 68532), we updated our analyses of case-mix change and finalized a reduction of 3.79 percent, instead of 2.71 percent, for CY 2011 and deferred finalizing a payment reduction for CY 2012 until further study of the case-mix change data and methodology was completed.

In the CY 2012 HH PPS final rule (76 FR 68526), we updated the 60-day national episode rates and the national per-visit rates. In addition, as discussed in the CY 2012 HH PPS final rule (76 FR 68528), our analysis indicated that there was a 22.59 percent increase in overall case-mix from 2000 to 2009 and that only 15.76 percent of that overall observed case-mix percentage increase was due to real case-mix change. As a result of our analysis, we identified a 19.03 percent nominal increase in case-mix. At that time, to fully account for the 19.03 percent nominal case-mix growth identified from 2000 to 2009, we finalized a 3.79 percent payment reduction in CY 2012 and a 1.32 percent payment reduction for CY 2013.

In the CY 2013 HH PPS final rule (77 FR 67078), we implemented a 1.32 percent reduction to the payment rates for CY 2013 to account for nominal case-mix growth from 2000 through 2010. When taking into account the total measure of case-mix change (23.90 percent) and the 15.97 percent of total case-mix change estimated as real from 2000 to 2010, we obtained a final nominal case-mix change measure of 20.08 percent from 2000 to 2010 (0.2390 * (1 − 0.1597) = 0.2008). To fully account for the remainder of the 20.08 percent increase in nominal case-mix beyond that which was accounted for in previous payment reductions, we estimated that the percentage reduction to the national, standardized 60-day episode rates for nominal case-mix change would be 2.18 percent. Although we considered proposing a 2.18 percent reduction to account for the remaining increase in measured nominal case-mix, we finalized the 1.32 percent payment reduction to the national, standardized 60-day episode rates in the CY 2012 HH PPS final rule (76 FR 68532).

Section 3131(a) of the Affordable Care Act requires that, beginning in CY 2014, we apply an adjustment to the national, standardized 60-day episode rate and other amounts that reflect factors such as changes in the number of visits in an episode, the mix of services in an episode, the level of intensity of services in an episode, the average cost of providing care per episode, and other relevant factors. Additionally, we must phase in any adjustment over a 4-year period in equal increments, not to exceed 3.5 percent of the amount (or amounts) as of the date of enactment of the Affordable Care Act, and fully implement the rebasing adjustments by CY 2017. The statute specifies that the maximum rebasing adjustment is to be no more than 3.5 percent per year of the CY 2010 rates. Therefore, in the CY 2014 HH PPS final rule (78 FR 72256) for each year, CY 2014 through CY 2017, we finalized a fixed-dollar reduction to the national, standardized 60-day episode payment rate of $80.95 per year, increases to the national per-visit payment rates per year, and a decrease to the NRS conversion factor of 2.82 percent per year. We also finalized three separate LUPA add-on factors for skilled nursing, physical therapy, and speech-language pathology and removed 170 diagnosis codes from assignment to diagnosis groups in the HH PPS Grouper. In the CY 2015 HH PPS final rule (79 FR 66032), we implemented the 2nd year of the 4 year phase-in of the rebasing adjustments to the HH PPS payment rates and made changes to the HH PPS case-mix weights. In addition, we simplified the face-to-face encounter regulatory requirements and the therapy reassessment timeframes.

In the CY 2016 HH PPS final rule (80 FR 68624), we implemented the 3rd year of the 4-year phase-in of the rebasing adjustments to the national, standardized 60-day episode payment amount, the national per-visit rates and the NRS conversion factor (as outlined above). In the CY 2016 HH PPS final rule, we also recalibrated the HH PPS case-mix weights, using the most current cost and utilization data available, in a budget neutral manner and finalized reductions to the national, standardized 60-day episode payment rate in CY 2016, CY 2017, and CY 2018 of 0.97 percent in each year to account for estimated case-mix growth unrelated to increases in patient acuity (that is, nominal case-mix growth) between CY 2012 and CY 2014. Finally, section 421(a) of the MMA, as amended by section 210 of the MACRA, extended the payment increase of 3 percent for HH services provided in rural areas (as defined in section 1886(d)(2)(D) of the Act) to episodes or visits ending before January 1, 2018.

In the CY 2017 HH PPS final rule (81 FR 76702), we implemented the last year of the 4-year phase-in of the rebasing adjustments to the national, standardized 60-day episode payment amount, the national per-visit rates and the NRS conversion factor (as outlined above). We also finalized changes to the methodology used to calculate outlier payments under the authority of section 1895(b)(5) of the Act. Lastly, in accordance with section 1834(s) of the Act, as added by section 504(a) of the Consolidated Appropriations Act, 2016 (Pub. L. 114-113, enacted December 18, 2015), we implemented changes in payment for furnishing Negative Pressure Wound Therapy (NPWT) using a disposable device for patients under a home health plan of care for which payment would otherwise be made under section 1895(b) of the Act.

D. Report to Congress: Home Health Study on Access to Care for Vulnerable Patient Populations and Subsequent Research and Analyses

Section 3131(d) of the Affordable Care Act required CMS to conduct a study on home health agency costs involved with providing ongoing access to care to low-income Medicare beneficiaries or beneficiaries in medically underserved areas, and in treating beneficiaries with varying levels of severity of illness and submit a report to Congress. As discussed in the CY 2016 HH PPS proposed rule (80 FR 39840) and the CY 2017 HH PPS proposed rule (81 FR 43744), the findings from the Report to Congress on the “Medicare Home Health Study: An Investigation on Access to Care and Payment for Vulnerable Patient Populations”, found that payment accuracy could be improved under the current payment system, particularly for patients with certain clinical characteristics requiring more nursing care than therapy.

The research for the Report to Congress, released in December 2014, consisted of extensive analysis of both survey and administrative data. The CMS-developed surveys were given to physicians who referred vulnerable patient populations to Medicare home health and to Medicare-certified HHAs. The response rates were 72 percent and 59 percent for the HHA and physician surveys, respectively. The results of the survey revealed that over 80 percent of respondent HHAs and over 90 percent of respondent physicians reported that access to home health care for Medicare fee-for-service beneficiaries in their local area was excellent or good. When survey respondents reported access issues, specifically their inability to place or admit Medicare fee-for-service patients into home health, the most common reason reported (64 percent of respondent HHAs surveyed) was that the patients did not qualify for the Medicare home health benefit. HHAs and physicians also cited family or caregiver issues as an important contributing factor in the inability to admit or place patients. Only 17.2 percent of HHAs and 16.7 percent of physicians reported insufficient payment as an important contributing factor in the inability to admit or place patients. The results of the CMS-conducted surveys suggested that CMS' ability to improve access for certain vulnerable patient populations through payment policy may be limited. However, we are able to revise the case-mix system to minimize differences in payment that could potentially be serving as a barrier to receiving care. In this rule, we propose to better align payment with resource use so that it reduces HHAs' financial incentives to select certain patients over others.

However, we also performed an analysis of Medicare administrative data (CY 2010 Medicare claims and cost report data) and calculated margins for episodes of care. This was done because margin differences associated with patient clinical and social characteristics can indicate whether financial incentives exist in the current HH PPS to provide home health care for certain types of patients over others. Lower margins, if systematically associated with care for vulnerable patient populations, may indicate financial disincentives for HHAs to admit these patients, potentially creating access to care issues. The findings from the data analysis found that certain patient characteristics appear to be strongly associated with margin levels, and thus may create financial incentives to select certain patients over others. Margins were estimated to be lower for patients who required parenteral nutrition, who had traumatic wounds or ulcers, or required substantial assistance in bathing. For example, in CY 2010, episodes for patients with parenteral nutrition were, on average, associated with a $178.53 lower margin than episodes for patients without parenteral nutrition. Given that these variables are already included in the HH PPS case-mix system, the results indicated that modifications to the way the current case-mix system accounts for resource use differences may be needed to mitigate any financial incentives to select certain patients over others. Margins were also lower for beneficiaries who were admitted after acute or post-acute stays or who had certain poorly-controlled clinical conditions, such as poorly-controlled pulmonary disorders, indicating that accounting for additional patient characteristic variables in the HH PPS case-mix system may also reduce financial incentives to select certain types of patients over others. More information on the results from the Home Health Study required by section 3131(d) of the Affordable Care Act can be found in the Report to Congress on the “Medicare Home Health Study: An Investigation on Access to Care and Payment for Vulnerable Patient Populations” available at https://www.cms.gov/center/provider-Type/home-Health-Agency-HHA-Center.html.

Section 3131(d)(5) of the Affordable Care Act allowed for the Secretary to determine whether a Medicare demonstration project is appropriate to conduct based on the result of the Home Health Study. If the Secretary determined it was appropriate to conduct the demonstration project under this subsection, the Secretary was to conduct the project for a four year period beginning not later than January 1, 2015. We did not determine that it was appropriate to conduct a demonstration project based on the findings from the Home Health Study. Rather, the findings from the Home Health Study suggested that follow-on work should be conducted to better align payments with costs under the authority of section 1895 of the Act.

In addition to the findings from the Report to Congress on the “Medicare Home Health Study: An Investigation on Access to Care and Payment for Vulnerable Patient Populations”, concerns have also been raised about the use of therapy thresholds in the current payment system. Under the current payment system, HHAs receive higher payments for providing more therapy visits once certain thresholds are reached. As a result, the average number of therapy visits per 60-day episode of care have increased since the implementation of the HH PPS, while the number of skilled nursing and home health aide visits have decreased over the same time period as shown in Figure 3 in section III.A of this rule. A study examining an option of using predicted, rather than actual, therapy visits in the HH found that in 2013, 58 percent of home health episodes included some therapy services, and these episodes accounted for 72 percent of all Medicare home health payments. Figure 1 from that study demonstrates that the percentage of episodes, and the average episode payment by the number of therapy visits for episodes with at least one therapy visit in 2013 increased sharply in therapy provision just over payment thresholds at 6, 7, and 16. According to the study, the presence of sharp increases in the percentage of episodes just above payment thresholds suggests a response to financial incentives in the home health payment system. Similarly, between 2008 and 2013, MedPAC reported a 26 percent increase in the number of episodes with at least 6 therapy visits, compared with a 1 percent increase in the number of episodes with five or fewer therapy visits. CMS analysis demonstrates that the average share of therapy visits across all 60-day episodes of care increased from 9 percent of all visits in 1997, prior to the implementation of the HH PPS (see 64 FR 58151), to 39 percent of all visits in 2015 (see Table 2 in section III.A. of this proposed rule).

Figure 1 suggests that HHAs may be responding to financial incentives in the home health payment system when making care plan decisions. Additionally, an investigation into the therapy practices of the four largest publically-traded home health companies, conducted by the Senate Committee on Finance in 2010, found that three out of the four companies investigated “encouraged therapists to target the most profitable number of therapy visits, even when patient need alone may not have justified such patterns”. The Committee on Finance investigation also highlighted the abrupt and dramatic responses the home health industry has taken to maximize reimbursement under the therapy threshold models (both the original 10-visit threshold model and under the revised thresholds implemented in the CY 2008 HH PPS final rule (72 FR 49762)). Under the HH PPS, the report noted that HHAs have broad discretion over the number of therapy visits to provide patients and therefore have control of the single-largest variable in determining reimbursement and overall margins. The report recommended that CMS closely examine a future payment approach that focuses on patient well-being and health characteristics, rather than the numerical utilization measures.

MedPAC also continues to recommend the removal of the therapy thresholds used for determining payment from the HH PPS, as it believes that such thresholds run counter to the goals of a prospective payment system, create financial incentives that detract from a focus on patient characteristics and care needs when agencies are setting plans of care for their patients, and incentivize unnecessary therapy utilization. For the average HHA, according to MedPAC, the increase in payment for therapy visits rises faster than costs resulting in financial incentives for HHAs to overprovide therapy services. HHAs that provide more therapy episodes tend to be more profitable and this higher profitability and rapid growth in the number of therapy episodes suggest that financial incentives are causing agencies to favor therapy services when possible. Eliminating therapy as a payment factor would base home health payment solely on patient characteristics, which is a more patient-focused approach to payment, as recommended by both MedPAC and previously by the Senate Committee on Finance.

After considering the findings from the Report to Congress and recommendations from MedPAC and the Senate Committee on Finance, CMS, along with our contractor, conducted additional research on ways to improve the payment accuracy under the current payment system. Exploring all options and different models ultimately led us to further develop the Home Health Groupings Model (HHGM) proposal. The HHGM proposal uses 30-day periods, rather than 60-day episodes, and relies more heavily on clinical characteristics and other patient information (for example, principal diagnosis, functional level, comorbid conditions, admission source, and timing) to place patients into meaningful payment categories, rather than the current therapy driven system. We believe this patient-centered approach is consistent with how clinicians differentiate between home health patients and would improve payment accuracy and access for medically complex cases and not just cases receiving therapy. The HHGM proposal leverages many of the same aspects of the current system; however, the major differences between the current system and the HHGM proposal include a change from a 60-day to a 30-day billing cycle and the elimination of the therapy thresholds in the case-mix system.

We shared the analyses and development of the HHGM with both internal and external stakeholders via technical expert panels, clinical workgroups, special open door forums, and in the CY 2016 HH PPS proposed rule (80 FR 39840) and the CY 2017 HH PPS proposed rule (81 FR 43744). Most recently, we posted a detailed technical report on the CMS Web site in December of 2016. After posting the technical report for the public to review, we also held additional technical expert panel and clinical workgroup webinars to garner feedback from the industry and conducted a National Provider call that occurred in January 2017 to solicit feedback from external stakeholders. The feedback we received during the National Provider call on the HHGM was positive. We discuss the HHGM proposal further below, in section III.E, and seek public comment on this proposal and the underlying analyses.

III. Provisions of the Proposed Rule: Payment Under the Home Health Prospective Payment System (HH PPS)

A. Monitoring for Potential Impacts—Affordable Care Act Rebasing Adjustments

1. Analysis of FY 2015 HHA Cost Report Data

As part of our efforts in monitoring the potential impacts of the rebasing adjustments finalized in the CY 2014 HH PPS final rule (78 FR 72293), we continue to update our analysis of home health cost report and claims data. Previous years' cost report and claims data analyses and results can be found in the CY 2017 HH PPS proposed rule (81 FR 43719 through 43720). For this proposed rule, we analyzed 2015 HHA cost report data and 2015 HHA claims data. To determine the 2015 average cost per visit per discipline, we applied the same trimming methodology outlined in the CY 2014 HH PPS proposed rule (78 FR 40284) and weighted the costs per visit from the 2015 cost reports by size, facility type, and urban/rural location so the costs per visit were nationally representative according to 2015 claims data. The 2015 average number of visits was taken from 2015 claims data. We estimated the cost of a 60-day episode in CY 2015 to be $2,449.01 using 2015 cost report data as shown in Table 2. However, the national, standardized 60-day episode payment amount in CY 2015 was $2,961.38. For CY 2015, on average, payments were 21 percent higher than costs (($2,961.38—$2,449.01)/$2,449.01).

2. Analysis of CY 2016 HHA Claims Data

In the CY 2014 HH PPS final rule (78 FR 72283), some commenters expressed concern that the rebasing of the HH PPS payment rates would result in HHA closures and would therefore diminish access to home health services. In addition to examining more recent cost report data, for this proposed rule we examined home health claims data from the first 3 years of the 4-year phase-in of the rebasing adjustments (CY 2014, CY 2015, and CY 2016), the first calendar year of the HH PPS (CY 2001), and claims data for 2 years before implementation of the rebasing adjustments (CY 2012 and CY2013). Analysis of CY 2016 home health claims data indicates that the number of episodes and the number of home health users that received at least one episode of care remained virtually the same (change of less than 1 percent) from 2015 to 2016, while the number of FFS beneficiaries increased 2 percent from 2015 to 2016. Between 2013 and 2014 there appears to be a net decrease in the number of HHAs billing Medicare for home health services of 1.6 percent, a continued decrease of 1.7 percent from 2014 to 2015, and a decrease of 2.5 percent from 2015 to 2016. The number of home health users, as a percentage of FFS beneficiaries, appears to have slightly decreased from 9.0 percent in 2012 to 8.7 percent in 2016, but remains higher than the 6.9 percent in 2001. In CY 2016, there were 2.9 HHAs per 10,000 FFS beneficiaries, which is still markedly higher than the 1.9 HHAs per 10,000 FFS beneficiaries observed close to the implementation of the HH PPS in 2001 (see Table 3). Therefore, the rebasing adjustments made to the HH PPS payment rates in CYs 2014 through 2016 do not appear to have resulted in significant HHA closures or otherwise diminished access to home health services.

In addition to examining home health claims data from the first three years of the implementation of rebasing adjustments required by the Affordable Care Act, we examined trends in home health utilization for all years starting in CY 2001 and up through CY 2016. Figure 2, displays the average number of visits per 60-day episode of care and the average payment per visit. While the average payment per visit has steadily increased from approximately $116 in CY 2001 to $167 for CY 2016, the average total number of visits per 60-day episode of care has declined, most notably between CY 2009 (21.7 visits per episode) and CY 2010 (19.8 visits per episode), which was the first year that the 10 percent agency-level cap on HHA outlier payments was implemented. The average of total visits per episode has steadily decreased from 21.7 in 2009 to 17.9 in 2016.

Figure 3 displays the average number of visits by discipline type for a 60-day episode of care and shows that the number of therapy visits per 60-day episode of care has increased steadily. However, the number of skilled nursing visits has decreased from 10.7 in 2009 to 8.7 in 2016. The number of home health aide visits has decreased from 5.6 average visits in 2009 to 1.5 visits in 2016. The results of the home health study required by section 3131(d) of the Affordable Care Act suggest that the current home health payment system may discourage HHAs from serving patients with clinically complex and/or poorly controlled chronic conditions who do not qualify for therapy but require a large number of skilled nursing visits. The home health study results seem to be consistent with the recent trend in the decreased number of visits per episode of care driven by decreases in skilled nursing and home health aide services evident in Figures 2 and 3.

As part of our monitoring efforts, we also examined the trends in episode timing and service use over time. The first and second episodes are considered “early” episodes, while third and later episodes are considered “late” episodes. Specifically, we examined the percentage of early episodes with 0 to 19 therapy visits, late episodes with 0 to 19 therapy visits, and episodes with 20+ therapy visits from CY 2008 to CY 2016. In CY 2008, we implemented refinements to the HH PPS case-mix system. As part of those refinements, we added additional therapy thresholds and differentiated between early and late episodes for those episodes with less than 20+ therapy visits. When the case-mix system first differentiated payments between early and late episodes of care, late episodes of care tended to have higher case-mix weights compared to early episodes of care. Table 4 shows that while there was a substantial increase in the number of late episodes between 2008 and 2009 (8 percentage points), since 2011 the number of late episodes as a percentage of total episodes has decreased over time. In 2015, the case-mix weights for the third and later episodes of care with 0 to 19 therapy visits decreased as a result of the CY 2015 recalibration of the case-mix weights. The recalibration of the HH PPS case-mix weights, beginning in CY 2015, does not seem to have substantially impacted the percentage of early versus late episodes of care.

The case-mix weights for episodes with 20+ therapy visits are not determined based on the timing of the episode of care. The percentage of episodes with 20+ therapy visits increased from 4.6 percent in CY 2008 to 7.0 percent in CY 2016. The increase in the percentage of episodes with 20+ therapy visits is consistent with the overall observed increase in therapy visits provided during a 60-day episode of care (see Figure 3).

We also examined trends in admission source for home health episodes over time. Specifically, we examined the admission source for the “first or only” episodes of care (first episodes in a sequence of adjacent episodes of care or the only episode of care) from CY 2008 through CY 2016 (Figure 4). The percentage of first or only episodes with an acute admission source, defined as episodes with an inpatient hospital stay within the 14 days prior to a home health episode, has decreased from 38.6 percent in CY 2008 to 33.9 percent in CY 2016. The percentage of first or only episodes with a post-acute admission source, defined as episodes which had a stay at a skilled nursing facility (SNF), inpatient rehabilitation facility (IRF), or long term care hospital (LTCH) within 14 days prior to the home health episode, slightly increased from 16.5 percent in CY 2008 to 17.5 percent in CY 2016. The percentage of first or only episodes with a community admission source, defined as episodes which did not have an acute or post-acute stay in the 14 days prior to the home health episode, increased from 37.4 percent in CY 2008 to 42.6 percent in CY 2016. Our findings on the trends in admission source are consistent with MedPAC's as outlined in their 2015 Report to the Congress. MedPAC examined admission source trends from 2002 up through 2013 and concluded that “there has been tremendous growth in the use of home health for patients residing in the community, episodes not preceded by a prior hospitalization. The high rates of volume growth for these types of episodes, which have more than doubled since 2001, suggest there is significant potential for overuse, particularly since Medicare does not currently require any cost sharing for home health care.”

We will continue to monitor for potential impacts due to the rebasing adjustments required by section 3131(a) of the Affordable Care Act and other policy changes in the future. Independent effects of any one policy may be difficult to discern in years where multiple policy changes occur in any given year.

B. Proposed CY 2018 HH PPS Case-Mix Weights

In the CY 2015 HH PPS final rule (79 FR 66072), we finalized a policy to annually recalibrate the HH PPS case-mix weights—adjusting the weights relative to one another—using the most current, complete data available. To recalibrate the HH PPS case-mix weights for CY 2018, we will use the same methodology finalized in the CY 2008 HH PPS final rule (72 FR 49762), the CY 2012 HH PPS final rule (76 FR 68526), and the CY 2015 HH PPS final rule (79 FR 66032). Annual recalibration of the HH PPS case-mix weights ensures that the case-mix weights reflect, as accurately as possible, current home health resource use and changes in utilization patterns.

To generate the proposed CY 2018 HH PPS case-mix weights, we used CY 2016 home health claims data (as of March 17, 2017) with linked OASIS data. These data are the most current and complete data available at this time. We will use CY 2016 home health claims data (as of June 30, 2017 or later) with linked OASIS data to generate the CY 2018 HH PPS case-mix weights in the CY 2018 HH PPS final rule. The process we used to calculate the HH PPS case-mix weights are outlined below.

Step 1: Re-estimate the four-equation model to determine the clinical and functional points for an episode using wage-weighted minutes of care as our dependent variable for resource use. The wage-weighted minutes of care are determined using the CY 2015 Bureau of Labor Statistics national hourly wage plus fringe rates for the six home health disciplines and the minutes per visit from the claim. The points for each of the variables for each leg of the model, updated with CY 2016 home health claims data, are shown in Table 5. The points for the clinical variables are added together to determine an episode's clinical score. The points for the functional variables are added together to determine an episode's functional score.

In updating the four-equation model for CY 2018, using 2016 home health claims data (the last update to the four-equation model for CY 2017 used CY 2015 home health claims data), there were few changes to the point values for the variables in the four-equation model. These relatively minor changes reflect the change in the relationship between the grouper variables and resource use between CY 2015 and CY 2016. The CY 2018 four-equation model resulted in 120 point-giving variables being used in the model (as compared to the 124 variables for the CY 2017 recalibration). There were 8 variables that were added to the model and 12 variables that were dropped from the model due to the absence of additional resources associated with the variable. Of the variables that were in both the four-equation model for CY 2017 and the four-equation model for CY 2018, the points for 14 variables increased in the CY 2018 four-equation model and the points for 48 variables decreased in the CY 2018 4-equation model. There were 50 variables with the same point values.

Step 2: Re-defining the clinical and functional thresholds so they are reflective of the new points associated with the CY 2018 four-equation model. After estimating the points for each of the variables and summing the clinical and functional points for each episode, we look at the distribution of the clinical score and functional score, breaking the episodes into different steps. The categorizations for the steps are as follows:

• Step 1: First and second episodes, 0-13 therapy visits.
• Step 2.1: First and second episodes, 14-19 therapy visits.
• Step 2.2: Third episodes and beyond, 14-19 therapy visits.
• Step 3: Third episodes and beyond, 0-13 therapy visits.
• Step 4: Episodes with 20+ therapy visits.

We then divide the distribution of the clinical score for episodes within a step such that a third of episodes are classified as low clinical score, a third of episodes are classified as medium clinical score, and a third of episodes are classified as high clinical score. The same approach is then done looking at the functional score. It was not always possible to evenly divide the episodes within each step into thirds due to many episodes being clustered around one particular score. Also, we looked at the average resource use associated with each clinical and functional score and used that as a guide for setting our thresholds. We grouped scores with similar average resource use within the same level (even if it meant that more or less than a third of episodes were placed within a level). The new thresholds, based off the CY 2018 four-equation model points are shown in Table 6.

Step 3: Once the clinical and functional thresholds are determined and each episode is assigned a clinical and functional level, the payment regression is estimated with an episode's wage-weighted minutes of care as the dependent variable. Independent variables in the model are indicators for the step of the episode as well as the clinical and functional levels within each step of the episode. Like the four-equation model, the payment regression model is also estimated with robust standard errors that are clustered at the beneficiary level. Table 7 shows the regression coefficients for the variables in the payment regression model updated with CY 2016 home health claims data. The R-squared value for the payment regression model is 0.5073 (an increase from 0.4919 for the CY 2017 recalibration).

Step 4: We use the coefficients from the payment regression model to predict each episode's wage-weighted minutes of care (resource use). We then divide these predicted values by the mean of the dependent variable (that is, the average wage-weighted minutes of care across all episodes used in the payment regression). This division constructs the weight for each episode, which is simply the ratio of the episode's predicted wage-weighted minutes of care divided by the average wage-weighted minutes of care in the sample. Each episode is then aggregated into one of the 153 home health resource groups (HHRGs) and the “raw” weight for each HHRG was calculated as the average of the episode weights within the HHRG.

Step 5: The raw weights associated with 0 to 5 therapy visits are then increased by 3.75 percent, the weights associated with 14-15 therapy visits are decreased by 2.5 percent, and the weights associated with 20+ therapy visits are decreased by 5 percent. These adjustments to the case-mix weights were finalized in the CY 2012 HH PPS final rule (76 FR 68557) and were done to address MedPAC's concerns that the HH PPS overvalues therapy episodes and undervalues non-therapy episodes and to better align the case-mix weights with episode costs estimated from cost report data.

Step 6: After the adjustments in Step 5 are applied to the raw weights, the weights are further adjusted to create an increase in the payment weights for the therapy visit steps between the therapy thresholds. Weights with the same clinical severity level, functional severity level, and early/later episode status were grouped together. Then within those groups, the weights for each therapy step between thresholds are gradually increased. We do this by interpolating between the main thresholds on the model (from 0-5 to 14-15 therapy visits, and from 14-15 to 20+ therapy visits). We use a linear model to implement the interpolation so the payment weight increase for each step between the thresholds (such as the increase between 0-5 therapy visits and 6 therapy visits and the increase between 6 therapy visits and 7-9 therapy visits) are constant. This interpolation is identical to the process finalized in the CY 2012 HH PPS final rule (76 FR 68555).

Step 7: The interpolated weights are then adjusted so that the average case-mix for the weights is equal to 1.0000. This last step creates the proposed CY 2018 case-mix weights shown in Table 8.

To ensure the changes to the HH PPS case-mix weights are implemented in a budget neutral manner, we then apply a case-mix budget neutrality factor to the proposed CY 2018 national, standardized 60-day episode payment rate (see section III.C.3. of this proposed rule). The case-mix budget neutrality factor is calculated as the ratio of total payments when the CY 2018 HH PPS case-mix weights (developed using CY 2016 home health claims data) are applied to CY 2016 utilization (claims) data to total payments when CY 2017 HH PPS case-mix weights (developed using CY 2015 home health claims data) are applied to CY 2016 utilization data. This produces a case-mix budget neutrality factor for CY 2018 of 1.0159.

C. Proposed CY 2018 Home Health Payment Rate Update

1. Proposed CY 2018 Home Health Market Basket Update

Section 1895(b)(3)(B) of the Act requires that the standard prospective payment amounts for CY 2018 be increased by a factor equal to the applicable HH market basket update for those HHAs that submit quality data as required by the Secretary. The home health market basket was rebased and revised in CY 2013. A detailed description of how we derive the HHA market basket is available in the CY 2013 HH PPS final rule (77 FR 67080 through 67090).

Section 1895(b)(3)(B)(vi) of the Act, requires that, in CY 2015 (and in subsequent calendar years, except CY 2018 (under section 411(c) of the Medicare Access and CHIP Reauthorization Act of 2015 (MACRA) (Pub. L. 114-10, enacted April 16, 2015)), the market basket percentage under the HHA prospective payment system as described in section 1895(b)(3)(B) of the Act be annually adjusted by changes in economy-wide productivity. Section 1886(b)(3)(B)(xi)(II) of the Act defines the productivity adjustment to be equal to the 10-year moving average of change in annual economy-wide private nonfarm business multifactor productivity (MFP) (as projected by the Secretary for the 10-year period ending with the applicable fiscal year, calendar year, cost reporting period, or other annual period) (the “MFP adjustment”). The Bureau of Labor Statistics (BLS) is the agency that publishes the official measure of private nonfarm business MFP. Please see http://www.bls.gov/mfp, to obtain the BLS historical published MFP data.

Prior to the enactment of the MACRA, which amended section 1895(b)(3)(B) of the Act, the proposed home health update percentage for CY 2018 would have been based on the estimated home health market basket update of 2.7 percent (based on IHS Global Insight Inc.'s first-quarter 2017 forecast with historical data through fourth-quarter 2016). Due to the requirements specified at section 1895(b)(3)(B)(vi) of the Act prior to the enactment of MACRA, the estimated CY 2018 home health market basket update of 2.7 percent would have been reduced by a MFP adjustment as mandated by the Affordable Care Act (currently estimated to be 0.5 percentage point for CY 2018). In effect, the proposed home health payment update percentage for CY 2018 would have been 2.2 percent. However, section 411(c) of the MACRA amended section 1895(b)(3)(B) of the Act, such that for home health payments for CY 2018, the market basket percentage increase is required to be 1 percent.

Section 1895(b)(3)(B) of the Act requires that the home health update be decreased by 2 percentage points for those HHAs that do not submit quality data as required by the Secretary. For HHAs that do not submit the required quality data for CY 2018, the home health payment update would be -1 percent (1 percent minus 2 percentage points).

2. Proposed CY 2018 Home Health Wage Index

Sections 1895(b)(4)(A)(ii) and (b)(4)(C) of the Act require the Secretary to provide appropriate adjustments to the proportion of the payment amount under the HH PPS that account for area wage differences, using adjustment factors that reflect the relative level of wages and wage-related costs applicable to the furnishing of HH services. Since the inception of the HH PPS, we have used inpatient hospital wage data in developing a wage index to be applied to HH payments. We propose to continue this practice for CY 2018, as we continue to believe that, in the absence of HH-specific wage data, using inpatient hospital wage data is appropriate and reasonable for the HH PPS. Specifically, we propose to continue to use the pre-floor, pre-reclassified hospital wage index as the wage adjustment to the labor portion of the HH PPS rates. For CY 2018, the updated wage data are for hospital cost reporting periods beginning on or after October 1, 2013, and before October 1, 2014 (FY 2014 cost report data). We would apply the appropriate wage index value to the labor portion of the HH PPS rates based on the site of service for the beneficiary (defined by section 1861(m) of the Act as the beneficiary's place of residence).

To address those geographic areas in which there are no inpatient hospitals, and thus, no hospital wage data on which to base the calculation of the CY 2018 HH PPS wage index, we propose to continue to use the same methodology discussed in the CY 2007 HH PPS final rule (71 FR 65884) to address those geographic areas in which there are no inpatient hospitals. For rural areas that do not have inpatient hospitals, we would use the average wage index from all contiguous Core Based Statistical Areas (CBSAs) as a reasonable proxy. Currently, the only rural area without a hospital from which hospital wage data could be derived is Puerto Rico. However, for rural Puerto Rico, we would not apply this methodology due to the distinct economic circumstances that exist there (for example, due to the close proximity to one another of almost all of Puerto Rico's various urban and non-urban areas, this methodology would produce a wage index for rural Puerto Rico that is higher than that in half of its urban areas). Instead, we would continue to use the most recent wage index previously available for that area. For urban areas without inpatient hospitals, we would use the average wage index of all urban areas within the state as a reasonable proxy for the wage index for that CBSA. For CY 2018, the only urban area without inpatient hospital wage data is Hinesville, GA (CBSA 25980).

On February 28, 2013, OMB issued Bulletin No. 13-01, announcing revisions to the delineations of MSAs, Micropolitan Statistical Areas, and CBSAs, and guidance on uses of the delineation of these areas. In the CY 2015 HH PPS final rule (79 FR 66085 through 66087), we adopted the OMB's new area delineations using a 1-year transition. The most recent bulletin (No. 15-01) concerning the revised delineations was published by the OMB on July 15, 2015.

The proposed CY 2018 wage index is available on the CMS Web site at http://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/HomeHealthPPS/Home-Health-Prospective-Payment-System-Regulations-and-Notices.html.

3. Proposed CY 2018 Annual Payment Update

a. Background

The Medicare HH PPS has been in effect since October 1, 2000. As set forth in the July 3, 2000 final rule (65 FR 41128), the base unit of payment under the Medicare HH PPS is a national, standardized 60-day episode payment rate. As set forth in § 484.220, we adjust the national, standardized 60-day episode payment rate by a case-mix relative weight and a wage index value based on the site of service for the beneficiary.

To provide appropriate adjustments to the proportion of the payment amount under the HH PPS to account for area wage differences, we apply the appropriate wage index value to the labor portion of the HH PPS rates. The labor-related share of the case-mix adjusted 60-day episode rate would continue to be 78.535 percent and the non-labor-related share would continue to be 21.465 percent as set out in the CY 2013 HH PPS final rule (77 FR 67068). The CY 2018 HH PPS rates would use the same case-mix methodology as set forth in the CY 2008 HH PPS final rule with comment period (72 FR 49762) and would be adjusted as described in section III.B of this rule. The following are the steps we take to compute the case-mix and wage-adjusted 60-day episode rate:

(1) Multiply the national 60-day episode rate by the patient's applicable case-mix weight.

(2) Divide the case-mix adjusted amount into a labor (78.535 percent) and a non-labor portion (21.465 percent).

(3) Multiply the labor portion by the applicable wage index based on the site of service of the beneficiary.

(4) Add the wage-adjusted portion to the non-labor portion, yielding the case-mix and wage adjusted 60-day episode rate, subject to any additional applicable adjustments.

In accordance with section 1895(b)(3)(B) of the Act, this document proposes the annual update of the HH PPS rates. Section 484.225 sets forth the specific annual percentage update methodology. In accordance with § 484.225(i), for a HHA that does not submit HH quality data, as specified by the Secretary, the unadjusted national prospective 60-day episode rate is equal to the rate for the previous calendar year increased by the applicable HH market basket index amount minus 2 percentage points. Any reduction of the percentage change would apply only to the calendar year involved and would not be considered in computing the prospective payment amount for a subsequent calendar year.

Medicare pays the national, standardized 60-day case-mix and wage-adjusted episode payment on a split percentage payment approach. The split percentage payment approach includes an initial percentage payment and a final percentage payment as set forth in § 484.205(b)(1) and (b)(2). We may base the initial percentage payment on the submission of a request for anticipated payment (RAP) and the final percentage payment on the submission of the claim for the episode, as discussed in § 409.43. The claim for the episode that the HHA submits for the final percentage payment determines the total payment amount for the episode and whether we make an applicable adjustment to the 60-day case-mix and wage-adjusted episode payment. The end date of the 60-day episode as reported on the claim determines which calendar year rates Medicare would use to pay the claim.

We may also adjust the 60-day case-mix and wage-adjusted episode payment based on the information submitted on the claim to reflect the following:

• A low-utilization payment adjustment (LUPA) is provided on a per-visit basis as set forth in §§ 484.205(c) and 484.230.
• A partial episode payment (PEP) adjustment as set forth in §§ 484.205(d) and 484.235.
• An outlier payment as set forth in §§ 484.205(e) and 484.240.

b. Proposed CY 2018 National, Standardized 60-Day Episode Payment Rate

Section 1895(3)(A)(i) of the Act requires that the 60-day episode base rate and other applicable amounts be standardized in a manner that eliminates the effects of variations in relative case-mix and area wage adjustments among different home health agencies in a budget neutral manner. To determine the CY 2018 national, standardized 60-day episode payment rate, we would apply a wage index budget neutrality factor; a case-mix budget neutrality factor described in section III.B. of this proposed rule; a reduction of 0.97 percent to account for nominal case-mix growth from 2012 to 2014, as finalized in the CY 2016 HH PPS final rule (80 FR 68646); and the home health payment update percentage discussed in section III.C.1 of this proposed rule.

To calculate the wage index budget neutrality factor, we simulated total payments for non-LUPA episodes using the proposed CY 2018 wage index and compared it to our simulation of total payments for non-LUPA episodes using the CY 2017 wage index. By dividing the total payments for non-LUPA episodes using the proposed CY 2018 wage index by the total payments for non-LUPA episodes using the CY 2017 wage index, we obtain a wage index budget neutrality factor of 1.0001. We would apply the wage index budget neutrality factor of 1.0001 to the calculation of the proposed CY 2018 national, standardized 60-day episode rate.

As discussed in section III.B. of this proposed rule, to ensure the changes to the case-mix weights are implemented in a budget neutral manner, we would apply a case-mix weight budget neutrality factor to the CY 2018 national, standardized 60-day episode payment rate. The case-mix weight budget neutrality factor is calculated as the ratio of total payments when CY 2018 case-mix weights are applied to CY 2016 utilization (claims) data to total payments when CY 2017 case-mix weights are applied to CY 2016 utilization data. The case-mix budget neutrality factor for CY 2018 would be 1.0159 as described in section III.B of this proposed rule.

Next, we would apply a reduction of 0.97 percent to the national, standardized 60-day payment rate for CY 2018 to account for nominal case-mix growth between CY 2012 and CY 2014. Lastly, we would update the proposed payment rates by the proposed CY 2018 home health payment update percentage of 1 percent as mandated by section 1895(b)(3)(B)(iii) of the Act. The proposed CY 2018 national, standardized 60-day episode payment rate is calculated in Table 9.

The proposed CY 2018 national, standardized 60-day episode payment rate for an HHA that does not submit the required quality data is updated by the proposed CY 2018 home health payment update of 1 percent minus 2 percentage points and is shown in Table 10.

c. Proposed CY 2018 National Per-Visit Rates

The national per-visit rates are used to pay LUPAs (episodes with four or fewer visits) and are also used to compute imputed costs in outlier calculations. The per-visit rates are paid by type of visit or HH discipline. The six HH disciplines are as follows:

• Home health aide (HH aide);
• Medical Social Services (MSS);
• Occupational therapy (OT);
• Physical therapy (PT);
• Skilled nursing (SN); and
• Speech-language pathology (SLP).

To calculate the proposed CY 2018 national per-visit rates, we start with the CY 2017 national per-visit rates. We then apply a wage index budget neutrality factor to ensure budget neutrality for LUPA per-visit payments. We calculate the wage index budget neutrality factor by simulating total payments for LUPA episodes using the proposed CY 2018 wage index and comparing it to simulated total payments for LUPA episodes using the CY 2017 wage index. By dividing the total payments for LUPA episodes using the proposed CY 2018 wage index by the total payments for LUPA episodes using the CY 2017 wage index, we obtain a wage index budget neutrality factor of 1.0005. We would apply the wage index budget neutrality factor of 1.0005 in order to calculate the CY 2018 national per-visit rates.

The LUPA per-visit rates are not calculated using case-mix weights. Therefore, there is no case-mix weights budget neutrality factor needed to ensure budget neutrality for LUPA payments. Lastly, the per-visit rates for each discipline are updated by the proposed CY 2018 home health payment update percentage of 1 percent. The national per-visit rates are adjusted by the wage index based on the site of service of the beneficiary. The per-visit payments for LUPAs are separate from the LUPA add-on payment amount, which is paid for episodes that occur as the only episode or initial episode in a sequence of adjacent episodes. The proposed CY 2018 national per-visit rates are shown in Tables 11 and 12.

The proposed CY 2018 per-visit payment rates for HHAs that do not submit the required quality data are updated by the proposed CY 2018 HH payment update percentage of 1 percent minus 2 percentage points and are shown in Table 12.

d. Low-Utilization Payment Adjustment (LUPA) Add-On Factors

LUPA episodes that occur as the only episode or as an initial episode in a sequence of adjacent episodes are adjusted by applying an additional amount to the LUPA payment before adjusting for area wage differences. In the CY 2014 HH PPS final rule, we changed the methodology for calculating the LUPA add-on amount by finalizing the use of three LUPA add-on factors: 1.8451 for SN; 1.6700 for PT; and 1.6266 for SLP (78 FR 72306). We multiply the per-visit payment amount for the first SN, PT, or SLP visit in LUPA episodes that occur as the only episode or an initial episode in a sequence of adjacent episodes by the appropriate factor to determine the LUPA add-on payment amount. For example, in the case of HHAs that do submit the required quality data, for LUPA episodes that occur as the only episode or an initial episode in a sequence of adjacent episodes, if the first skilled visit is SN, the payment for that visit would be $264.46 (1.8451 multiplied by $143.33), subject to area wage adjustment.

e. Proposed CY 2018 Non-Routine Medical Supply (NRS) Payment Rates

Payments for NRS are computed by multiplying the relative weight for a particular severity level by the NRS conversion factor. To determine the proposed CY 2018 NRS conversion factor, we update the CY 2017 NRS conversion factor ($52.50) by the proposed CY 2018 home health payment update percentage of 1 percent. We do not apply a standardization factor as the NRS payment amount calculated from the conversion factor is not wage or case-mix adjusted when the final claim payment amount is computed. The proposed NRS conversion factor for CY 2018 is shown in Table 13.

Using the CY 2018 NRS conversion factor, the payment amounts for the six severity levels are shown in Table 14.

For HHAs that do not submit the required quality data, we update the CY 2017 NRS conversion factor ($52.50) by the proposed CY 2018 home health payment update percentage of 1 percent minus 2 percentage points. The proposed CY 2018 NRS conversion factor for HHAs that do not submit quality data is shown in Table 15.

The payment amounts for the various severity levels based on the updated conversion factor for HHAs that do not submit quality data are calculated in Table 16.

f. Rural Add-On

Section 421(a) of the MMA required, for HH services furnished in a rural areas (as defined in section 1886(d)(2)(D) of the Act), for episodes or visits ending on or after April 1, 2004, and before April 1, 2005, that the Secretary increase the payment amount that otherwise would have been made under section 1895 of the Act for the services by 5 percent.

Section 5201 of the DRA amended section 421(a) of the MMA. The amended section 421(a) of the MMA required, for HH services furnished in a rural area (as defined in section 1886(d)(2)(D) of the Act), on or after January 1, 2006, and before January 1, 2007, that the Secretary increase the payment amount otherwise made under section 1895 of the Act for those services by 5 percent.

Section 3131(c) of the Affordable Care Act amended section 421(a) of the MMA to provide an increase of 3 percent of the payment amount otherwise made under section 1895 of the Act for HH services furnished in a rural area (as defined in section 1886(d)(2)(D) of the Act), for episodes and visits ending on or after April 1, 2010, and before January 1, 2016.

Section 210 of the MACRA amended section 421(a) of the MMA to extend the rural add-on by providing an increase of 3 percent of the payment amount otherwise made under section 1895 of the Act for HH services provided in a rural area (as defined in section 1886(d)(2)(D) of the Act), for episodes and visits ending before January 1, 2018. Therefore, for episodes and visits that end on or after January 1, 2018, a rural add-on payment will not apply.

D. Payments for High-Cost Outliers Under the HH PPS

1. Background

Section 1895(b)(5) of the Act allows for the provision of an addition or adjustment to the home health payment amount in the case of outliers because of unusual variations in the type or amount of medically necessary care. Prior to the enactment of the Affordable Care Act, section 1895(b)(5) of the Act stipulated that projected total outlier payments could not exceed 5 percent of total projected or estimated HH payments in a given year. In the July 3, 2000 Medicare Program; Prospective Payment System for Home Health Agencies final rule (65 FR 41188 through 41190), we described the method for determining outlier payments. Under this system, outlier payments are made for episodes whose estimated costs exceed a threshold amount for each Home Health Resource Group (HHRG). The episode's estimated cost was established as the sum of the national wage-adjusted per-visit payment amounts delivered during the episode. The outlier threshold for each case-mix group or Partial Episode Payment (PEP) adjustment is defined as the 60-day episode payment or PEP adjustment for that group plus a fixed-dollar loss (FDL) amount. The outlier payment is defined to be a proportion of the wage-adjusted estimated cost beyond the wage-adjusted threshold. The threshold amount is the sum of the wage and case-mix adjusted PPS episode amount and wage-adjusted FDL amount. The proportion of additional costs over the outlier threshold amount paid as outlier payments is referred to as the loss-sharing ratio.

In the CY 2010 HH PPS proposed rule (74 FR 40948, 40957), we stated that outlier payments increased as a percentage of total payments from 4.1 percent in CY 2005, to 5.0 percent in CY 2006, to 6.4 percent in CY 2007 and that this excessive growth in outlier payments was primarily the result of unusually high outlier payments in a few areas of the country. In that discussion, we noted that despite program integrity efforts associated with excessive outlier payments in targeted areas of the country, we discovered that outlier expenditures still exceeded the 5 percent target in CY 2007 and, in the absence of corrective measures, would continue do to so. Consequently, we assessed the appropriateness of taking action to curb outlier abuse. As described in the CY 2010 HH PPS final rule (74 FR 58080 through 58087), to mitigate possible billing vulnerabilities associated with excessive outlier payments and adhere to our statutory limit on outlier payments, we finalized an outlier policy that included a 10 percent agency-level cap on outlier payments. This cap was implemented in concert with a reduced FDL ratio of 0.67. These policies resulted in a projected target outlier pool of approximately 2.5 percent. (The previous outlier pool was 5 percent of total home health expenditures). For CY 2010, we first returned the 5 percent held for the previous target outlier pool to the national, standardized 60-day episode rates, the national per-visit rates, the LUPA add-on payment amount, and the NRS conversion factor. Then, we reduced the CY 2010 rates by 2.5 percent to account for the new outlier pool of 2.5 percent. This outlier policy was adopted for CY 2010 only.

As we noted in the CY 2011 HH PPS final rule (75 FR 70397 through 70399), section 3131(b)(1) of the Affordable Care Act amended section 1895(b)(3)(C) of the Act, and required the Secretary to reduce the HH PPS payment rates such that aggregate HH PPS payments were reduced by 5 percent. In addition, section 3131(b)(2) of the Affordable Care Act amended section 1895(b)(5) of the Act by redesignating the existing language as section 1895(b)(5)(A) of the Act, and revising the language to state that the total amount of the additional payments or payment adjustments for outlier episodes may not exceed 2.5 percent of the estimated total HH PPS payments for that year. Section 3131(b)(2)(C) of the Affordable Care Act also added section 1895(b)(5)(B) of the Act which capped outlier payments as a percent of total payments for each HHA at 10 percent.

As such, beginning in CY 2011, our HH PPS outlier policy is that we reduce payment rates by 5 percent and target up to 2.5 percent of total estimated HH PPS payments to be paid as outliers. To do so, we first returned the 2.5 percent held for the target CY 2010 outlier pool to the national, standardized 60-day episode rates, the national per visit rates, the LUPA add-on payment amount, and the NRS conversion factor for CY 2010. We then reduced the rates by 5 percent as required by section 1895(b)(3)(C) of the Act, as amended by section 3131(b)(1) of the Affordable Care Act. For CY 2011 and subsequent calendar years we target up to 2.5 percent of estimated total payments to be paid as outlier payments, and apply a 10 percent agency-level outlier cap.

In the CY 2017 HH PPS proposed and final rules (81 FR 43737 through 43742 and 81 FR 76702), we described our concerns regarding patterns observed in home health outlier episodes. Specifically, we noted that the methodology for calculating home health outlier payments may have created a financial incentive for providers to increase the number of visits during an episode of care to surpass the outlier threshold and simultaneously created a disincentive for providers to treat medically complex beneficiaries who require fewer but longer visits. Given these concerns, in the CY 2017 HH PPS final rule (81 FR 76702), we finalized changes to the methodology used to calculate outlier payments, using a cost-per-unit approach rather than a cost-per-visit approach. This change in methodology allows for more accurate payment for outlier episodes, accounting for both the number of visits during an episode of care and also the length of the visits provided. Using this approach, we now convert the national per-visit rates into per 15-minute unit rates. These per 15-minute unit rates are used to calculate the estimated cost of an episode to determine whether the claim will receive an outlier payment and the amount of payment for an episode of care. In conjunction with our finalized policy to change to a cost-per-unit approach to estimate episode costs and determine whether an outlier episode should receive outlier payments, in the CY 2017 HH PPS final rule we also finalized the implementation of a cap on the amount of time per day that would be counted toward the estimation of an episode's costs for outlier calculation purposes (81 FR 76725). Specifically, we limit the amount of time per day (summed across the six disciplines of care) to 8 hours (32 units) per day when estimating the cost of an episode for outlier calculation purposes.

2. Fixed Dollar Loss (FDL) Ratio

For a given level of outlier payments, there is a trade-off between the values selected for the FDL ratio and the loss-sharing ratio. A high FDL ratio reduces the number of episodes that can receive outlier payments, but makes it possible to select a higher loss-sharing ratio, and therefore, increase outlier payments for qualifying outlier episodes. Alternatively, a lower FDL ratio means that more episodes can qualify for outlier payments, but outlier payments per episode must then be lower.

The FDL ratio and the loss-sharing ratio must be selected so that the estimated total outlier payments do not exceed the 2.5 percent aggregate level (as required by section 1895(b)(5)(A) of the Act). Historically, we have used a value of 0.80 for the loss-sharing ratio which, we believe, preserves incentives for agencies to attempt to provide care efficiently for outlier cases. With a loss-sharing ratio of 0.80, Medicare pays 80 percent of the additional estimated costs above the outlier threshold amount.

Simulations based on CY 2015 claims data (as of June 30, 2016) completed for the CY 2017 HH PPS final rule showed that outlier payments were estimated to represent approximately 2.84 percent of total HH PPS payments in CY 2017, and as such, we raised the FDL ratio from 0.45 to 0.55. We stated that raising the FDL ratio to 0.55, while maintaining a loss-sharing ratio of 0.80, struck an effective balance of compensating for high-cost episodes while still meeting the statutory requirement to target up to, but no more than, 2.5 percent of total payments as outlier payments (81 FR 76726). The national, standardized 60-day episode payment amount is multiplied by the FDL ratio. That amount is wage-adjusted to derive the wage-adjusted FDL amount, which is added to the case-mix and wage-adjusted 60-day episode payment amount to determine the outlier threshold amount that costs have to exceed before Medicare would pay 80 percent of the additional estimated costs.

For this proposed rule, using preliminary CY 2016 claims data (as of March 17, 2017) and the proposed CY 2018 payment rates presented in section III.C of this proposed rule, we estimate that outlier payments would constitute approximately 2.47 percent of total HH PPS payments in CY 2018 under the current outlier methodology. Given the statutory requirement to target up to, but no more than, 2.5 percent of total payments as outlier payments, we are not proposing a change to the FDL ratio for CY 2018 as we believe that maintaining an FDL ratio of 0.55 with a loss-sharing ratio of 0.80 is still appropriate given the percentage of outlier payments projected for CY 2018. Likewise, we are not proposing a change to the loss-sharing ratio (0.80) for the HH PPS to remain consistent with payment for high-cost outliers in other Medicare payment systems (for example, IRF PPS, IPPS, etc.). While we are not proposing to change the FDL ratio of 0.55 for CY 2018, we note that in the final rule, we will update our estimate of outlier payments as a percent of total HH PPS payments using the most current and complete year of HH PPS data (CY 2016 claims data as of June 30, 2017 or later). This may result in changes to the FDL ratio in the final rule.

E. Proposed Implementation of the Home Health Groupings Model (HHGM) for CY 2019

1. Overview, Data, and File Construction

Under the home health prospective payment system (HH PPS), Medicare pays for home health services provided during a 60-day episode of care. Episodes are case-mix adjusted based on the timing of the episode within a sequence of episodes, the patient's clinical status and functional status as determined using information from the Outcome and Assessment Information Set (OASIS), and the amount of therapy service provided during the episode. Therapy service use is measured by the number of therapy visits provided during the episode and can be categorized into nine visit level categories (or thresholds): 0-5; 6; 7-9; 10; 11-13; 14-15; 16-17; 18-19; and 20 or more visits. The combinations of episode timing, clinical and functional levels, and therapy service use categories result in 153 home health resource groups (HHRGs) into which home health episodes are categorized. Each HHRG is assigned a relative weight reflecting the average resource use of patients in that group compared with average resource use across all Medicare home health patients; this weight is then used to case mix adjust the episode's payment (with an additional adjustment for geographic variation in wages). Additional payment adjustments are made for very resource intensive (outlier) episodes, episodes with very few visits, transfers to other HHAs or to hospitals with a return to home health during the episode, and the expected use of non-routine medical supplies (NRS).

As discussed in section II.D of this proposed rule, the Report to Congress, required by section 3131(d) of the Affordable Care Act, found that payment accuracy could be improved under the current payment system, particularly for patients with certain clinical characteristics. Findings from the report suggest that the current home health payment system may discourage HHAs from serving patients with clinically complex and/or poorly controlled chronic conditions who do not need therapy services, but require skilled nursing care. In addition, MedPAC believes that the Medicare home health benefit is ill-defined and the current reliance on therapy service thresholds for determining payment is counter to the goals of a prospective payment system. Under the current payment system, HHAs receive higher payments for providing more therapy visits, which may incentivize unnecessary utilization. MedPAC reitereated their recommendation in the March 2017 Report to Congress that CMS eliminate the use of the number of therapy vists as a payment factor in the home health PPS beginning in 2019.

To better align payment with patient care needs and better ensure that clinically complex and ill beneficiaries have adequate access to home health care, we are proposing for CY 2019 case-mix methodology refinements through the implementation of the Home Health Groupings Model (HHGM). We propose to implement the HHGM for home health periods of care beginning on or after January 1, 2019. The implementation of the HHGM will require provider education and training, updating and revising relevant manuals, and changing claims processing systems. Implementation starting in CY 2019 would provide an opportunity for CMS, its contractors, and the agencies themselves to prepare. This patient-centered model groups periods of care in a manner consistent with how clinicians differentiate between patients and the primary reason for needing home health care. The HHGM uses 30-day periods rather than the 60-day episode used in the current payment system, eliminates the use of the number of therapy visits provided to determine payment, and relies more heavily on clinical characteristics and other patient information (for example, diagnosis, functional level, comorbid conditions, admission source) to place patients into clinically meaningful payment categories. In total, there are 144 different payment groups in the HHGM.

Costs during an episode/period of care are estimated based on the concept of resource use, which measures the costs associated with visits performed during a home health episode/period. For the current HH PPS case-mix weights, we use Wage Weighted Minutes of Care (WWMC), which uses data from the Bureau of Labor Statistics (BLS) reflecting the Home Health Care Service Industry. For the HHGM, we propose shifting to a Cost-Per-Minute plus Non-Routine Supplies (CPM + NRS) approach, which uses information from the Medicare Cost Report. The CPM + NRS approach incorporates a wider variety of costs (such as transportation) compared to the BLS estimates and the costs are available for individual HHA providers while the BLS costs are aggregated for the Home Health Care Service industry. The proposed methodology used to calculate the cost of an episode/period of care is discussed in detail in section III.E.2. of this proposed rule.

We propose using the 30-day periods rather than the 60-day episodes in the current payment system. Episodes have more visits, on average, during the first 30 days compared to the last 30 days. Costs are much higher earlier in the episode and lesser later on, therefore we believe that dividing a single 60-day episode into two 30-day periods more accurately apportions payments. Overall, we found that the average length of an episode of care was 47 days, but roughly a quarter of all 60 days episodes lasted 30 days or less. The proposed change from 60-day billing to 30-day billing under the HHGM is discussed in detail in section III.E.3. of this proposed rule.

Similar to the current payment system, 30-day periods under the HHGM would be classified as “early” or “late” depending on when they occur within a sequence of 30-day periods. Under the current HH PPS, the first two 60-day episodes of a sequence of adjacent 60-day episodes are considered early, while the third 60-day episode of that sequence and any subsequent episodes are considered late. Under the HHGM, the first 30-day period is classified as early. All subsequent 30-day periods in the sequence (second or later) are classified as late. We propose to adopt this episode timing classification for 30-day periods with the implementation of the HHGM. Similar to the current payment system, we propose that a 30-day period could not be considered early unless there was a gap of more than 60 days between the end of one period and the start of another. The comprehensive assessment would still be completed within 5 days of the start of care date and completed no less frequently than during the last 5 days of every 60 days beginning with the start of care date, as currently required by § 484.55, Condition of participation: Comprehensive assessment of patients. The proposed episode timing classification is discussed in detail in section III.E.4. of this proposed rule.

Under the HHGM, each period would be classified into one of two admission source categories—community or institutional—depending on what healthcare setting was utilized in the 14 days prior to home health. The 30-day period would be categorized as institutional if an acute or post-acute care stay occurred in the prior 14 days to the start of the 30-day period of care. The 30-day period would be categorized as community if there was no acute or post-acute care stay in the 14 days prior to the start of the 30-day period of care. We propose to adopt this categorization by admission source with the implementation of the HHGM. The proposed admission classification source is discussed in detail in section III.E.5. of this proposed rule.

The HHGM would group 30-day periods into categories based on a variety of patient characteristics. Within the HHGM, one of the steps in case-mix adjusting the 30-day payment amount would include grouping periods into one of six clinical groups based on the principal diagnosis listed on the home health claim. We propose grouping periods into one of six clinical groups based on the principal diagnosis with the implementation of the HHGM. The principal diagnosis reported would provide information to describe the primary reason for which patients are receiving home health services under the Medicare home health benefit. The proposed six clinical groups, which are discussed in detail in section III.E.6. of this proposed rule, are as follows:

• Musculoskeletal Rehabilitation.
• Neuro/Stroke Rehabilitation.
• Wounds—Post-Op Wound Aftercare and Skin/Non-Surgical Wound Care.
• Complex Nursing Interventions.
• Behavioral Health Care.
• Medication Management, Teaching and Assessment (MMTA).

Under the HHGM, each 30-day period would be placed into one of three functional levels. The level would indicate if, on average, given its responses on certain functional OASIS items, a 30-day period is predicted to have higher costs or lower costs. We propose classifying 30-day periods according to functional level. For each of the six clinical groups, we propose that periods would be further classified into one of three functional levels with roughly 33 percent of periods in each level. The creation of this functional level is very similar to how the functional level is created in the current payment system. The proposed functional levels and corresponding OASIS items are discussed in detail in section III.E.7. of this proposed rule.

Exploratory analyses determined that comorbidities—that is, secondary diagnoses—provide additional information that can further explain resource use differences across 30-day periods of care even after controlling for the primary diagnosis. Comorbidities are tied to poorer health outcomes, more complex medical need and management, and higher costs. The HHGM would include a comorbidity adjustment category based on the presence of secondary diagnoses. We propose that 30-day periods would receive a comorbidity adjustment if any diagnosis codes listed on the home health claim are included on a list of comorbidities that occurred in at least 0.1 percent of 30-day periods and associated with increased average resource use. The proposed comorbidity adjustment is discussed in detail in section III.E.8. of this proposed rule.

Currently, if an HHA provides four visits or less in an episode, they will be paid a standardized per visit payment instead of an episode payment for a 60-day episode of care. These payment adjustments are called Low-Utilization Payment Adjustments (LUPAs). While the HHGM would still include LUPAs, the approach to calculating the LUPA thresholds would need to change in the HHGM because of the switch to 30-day periods from 60-day episodes. Whereas there is a single LUPA threshold of 4 visits for all episodes under the current payment system, we propose the LUPA threshold would vary for a 30-day period under the HHGM depending on the HHGM payment group to which it was assigned. To create LUPA thresholds, 30-day periods (including those that were LUPAs in the current payment system) were grouped into the 144 different HHGM payment groups. For each payment group, we propose to use the 10th percentile value of visits to create a payment group specific LUPA threshold with a minimum threshold of at least 2 for each group. The proposed LUPA thresholds are discussed in more detail in section III.E.9. of this proposed rule.

Figure 5 represents how each 30-day period of care would be placed into one of 144 home health resource groups (HHRGs) under the proposed HHGM.

While the proposed HHGM would reflect a change in the case-mix adjustment methodology, the conditions for payment would remain the same for Medicare home health services, meaning all requirements would still need to be met in accordance with § 424.22. This includes physician certification that: (1) The individual is in need or needed intermittent skilled nursing care, or physical therapy or speech-language pathology services, and is confined to the home; (2) a plan of care has been established and will be periodically reviewed by a physician who is a doctor of medicine, osteopathy, or podiatric medicine; (3) the individual was under the care of a physician who is a doctor of medicine, osteopathy, or podiatric medicine; and, (4) a face-to-face patient encounter, which is related to the primary reason the patient requires home health services, occurred no more than 90 days prior to the home health start of care date or within 30 days of the start of the home health care and was performed by a physician or allowed non-physician practitioner. Likewise, under the HHGM, the Medicare beneficiary would retain all rights that currently exist under the current HH PPS, including those related to beneficiary liability for services or any reduction or termination of services. These would include the issuance of the Advanced Beneficiary Notice (ABN) and the Home Health Change of Care Notice (HHCCN), when appropriate. Medicare home health agencies are required to issue an ABN when a HHA believes Medicare will not pay for some or all of the patient's Medicare home health care. In these circumstances, if the beneficiary chooses to receive the items/services in question and Medicare does not cover the home health care, HHAs may use the ABN to shift liability for the non-covered home health care to the beneficiary. The HHCCN is a written notice that the HHA provides a beneficiary when his/her home health plan of care is changing because the home health agency makes a business decision to reduce or stop providing the patient some or all of the home health services or supplies OR the beneficiary's physician changed orders which may reduce or stop certain Medicare covered home health services or supplies.

To create the HHGM proposed model and related analyses, a data file based on home health episodes of care as reported in Medicare home health claims was utilized. The claims data provide episode-level data (for example, episode From and Through Dates, total number of visits, HHRG, diagnoses), as well as visit-level data (visit date, visit length in 15-minute units, discipline of the staff, etc.). The claims also provide data on whether NRS was provided during the episode and total charges for NRS.

The core file for most of the analyses for this proposed rule includes 100 percent of home health episode claims with Through Dates in Calendar Year (CY) 2016, processed by March 17, 2017, accessed via the Chronic Conditions Data Warehouse (CCW). Original or adjustment claims processed after March 17, 2017, would not be reflected in the core file. The claims-based file was supplemented with additional variables that were obtained from the CCW, such as information regarding other Part A and Part B utilization.

The data were cleaned by processing any remaining adjustments and by excluding duplicates and claims that were Requests for Anticipated Payment (RAP). In addition, visit-level variables needed for the analysis were extracted from the revenue center trailers (that is, the line items that describe the visits) and downloaded as a separate visit-level file, with selected episode-level variables merged onto the records for visits during those episodes. To account for potential data entry errors, the visit-level variables for visit length were top-censored at eight hours.

A set of data cleaning exclusions were applied to the episode-level file, which resulted in the exclusion of the following:

• Episodes with no covered visits.
• Episodes with any missing units or visit data.
• Episodes with zero payments.
• Episodes with no charges.
• Non-LUPA episodes missing an HHRG.

The analysis file also includes data on patient characteristics obtained from the OASIS assessments conducted by HHA staff at the start of each episode. The assessment data are electronically submitted by home health agencies (HHAs) to a central CMS repository. In constructing the core data file, 100 percent of the OASIS assessments submitted October 2015, through December 2016 from the CMS repository were uploaded by CMS to the CCW. A CCW-derived linking key (Bene_ID) was used to match the OASIS data with CY 2016 episodes of care. Episodes that could not be linked with an OASIS assessment were excluded from the analysis file, as they included insufficient patient-level data to create the HHGM.

To construct measures of resource use, a variety of data sources were used (see section III.E.2 of this proposed rule for the proposed methodology used to calculate the cost of care under the HHGM). First, BLS data on average wages and fringe benefits were used to produce one version of the wage-weighted cost per minute for each home health discipline. The wage data are for North American Industry Classification System (NAICS) 621600—Home Health Care Services. The wage data are broken down by the following occupations:

For visits where the service provided—as indicated by the Healthcare Common Procedure Coding System (HCPCS) code—can be provided by only a single standard occupation classification code; for example, establishment or review of a plan of care by a registered nurse (RN; HCPCS = G0162), the wage (and fringe) rate for that standard occupation classification is used to calculate the cost of the minutes for the visit. For visits where the service provided can potentially be provided by different standard occupation classification, such as observation and assessment by an RN or a Licensed Practical Nurse (LPN; HCPCS = G0163), a blended rate is applied, with the rate for each standard occupation classification code weighted by the total home health employment for that standard occupation classification code. The employment data are available from the same BLS table as the wage data.

Home Health Agency Medicare Cost Report (MCR) data were also used to construct a measure of resource use after trimming out HHAs whose costs were outliers. These data are used to provide a representation of the average costs of visits provided by HHAs in the six Medicare home health disciplines: Skilled nursing; physical therapy; occupational therapy; speech-language pathology; medical social services; and home health aide services. Cost report data are publicly available at https://www.cms.gov/Research-Statistics-Data-and-Systems/Downloadable-Public-Use-Files/Cost-Reports/.

The 2016 analytic file included 6,293,442 episodes. Of these, 469,346 (7.5 percent) were excluded because they could not be linked to OASIS assessments or because of the reasons listed above. This yielded an analysis file including 5,824,096 episodes. Those episodes are 60-day episodes under the current payment system, but for the HHGM those 60-day episodes were converted into two 30-day periods. This yielded a final HHGM analytic file that included 10,231,507, 30-day periods. Certain 30-day periods were excluded for the following reasons:

• Periods missing a diagnosis code or where the diagnosis code did not link to a clinical group to case-mix adjust the period's payment (after exclusions, n = 10,177,949).
• Inability to merge to certain OASIS items to create the episode's functional level that is used for risk adjustment. For all the periods in the analytic file, there was a look-back through CY 2015 for a Start of Care or Resumption of Care assessment that preceded the period being analyzed and was in the same sequence of periods. If such an assessment was found, it was used to impute responses for OASIS items that were not included in the follow-up assessment. Periods which did not link to a Start of Care or Resumption of Care assessment were dropped (after exclusions, n = 9,477,856).
• No nursing visits or therapy visits (after exclusions, n = 9,290,340).
• LUPAs were excluded from the analysis. Periods that are identified as LUPAs in the current payment system are excluded in the creation of the functional score. Following the creation of the score (and the corresponding levels), case-mix group specific LUPA thresholds were created and episodes/periods were excluded that were below the new LUPA threshold when computing the case-mix weights. Therefore, the final analytic sample included 8,642,107 30-day periods that were used for the analyses in the HHGM.

As noted in section II.D of this proposed rule, the analyses and the ultimate development of Home Health Groupings Model (HHGM) have been shared with both internal and external stakeholders via technical expert panels, clinical workgroups, special open door forums, and in the CY 2017 HH PPS final rule (81 FR 76702). Technical expert panel and clinical workgroup webinars on the technical report were held in December 2016 and a detailed technical report was posted on the CMS home health agency Web page in December, providing opportunity for stakeholder feedback. We also held a National Provider call in January 2017, to further solicit feedback from the public.

2. Methodology Used To Calculate the Cost of Care

To construct the case-mix weights for the HHGM proposal, the costs of providing care needed to be determined. A Wage-Weighted Minutes of Care (WWMC) approach is used in the current payment system based on data from the BLS. However, we are proposing to adopt a Cost-per-Minute plus Non-Routine Supplies (CPM + NRS) approach, which uses information from Medicare Cost Reports (MCR). We used the following data sources and methodology for calculating these measures of resource use:

• BLS Wage Estimates: For the WWMC method of calculating home health resource use, wage and fringe data was obtained from the BLS by industry code from the NAICS and occupation code from the Standard Operation Classification. These data provide nationwide average wage rates and the average value of fringe benefits per hour of work for specific occupations.
• Home Health Medicare Cost Report Data: All Medicare-certified HHAs must report their own costs through publicly-available home health cost reports maintained by the Healthcare Cost Report Information System (HCRIS). Freestanding HHAs report HHA-specific cost reports while HHAs that are hospital-based report on the HHA component of the hospital cost reports. These cost reports enable estimation of the cost per visit by provider and the estimated NRS cost to charge ratios. To obtain a more robust estimate of cost, a trimming process was applied to remove cost reports with missing or questionable data and extreme values.

• Home Health Claims Data: Medicare home health claims data are used in both the WWMC and CPM+NRS methods to obtain minutes of care by discipline of care.
• Wage-Weighted Minutes of Care (WWMC) Approach: Used in the current payment system, this approach determines resource use for each episode by multiplying utilization (in terms of the number of minutes of direct patient care provided by each discipline) by the corresponding opportunity cost of that care (represented by wage and fringe benefit rates from the BLS). Table 18 shows the occupational titles and corresponding mean hourly wage rates from the BLS. The employer cost per hour worked shown in the fifth column is calculated by adding together the mean hourly wage rates and the fringe benefit rates from the BLS (generally around 37 percent of wages). For home health disciplines that include multiple occupations (such as skilled nursing), the opportunity cost is generated by weighting the employer cost by the proportions of the labor mix. Otherwise, the opportunity cost is the same as the employer cost per hour.

For each home health period of care, the number of minutes of care provided (obtained from the home health claims) is weighted by the corresponding opportunity cost for each discipline providing the minutes. The resulting wage-weighted minutes of care are summed for the 30-day period to obtain total costs. Table 19 shows these costs overall for 30-day periods in CY 2016 (n = 8,642,107). On average, total period costs were $374.52. The distribution ranged from a 5th percentile value of $73.87 to a 95th percentile value of $912.10.

In the current HH PPS, all episodes without a LUPA payment receive payment for NRS, regardless of whether or not the HHA provided NRS during that episode. NRS payment amounts are determined through a payment model separately from the one used to construct the episode's case-mix weight. The current payment system determines NRS payment using the presence of clinical factors associated with NRS provision from the OASIS. Two-thirds of episodes do not include provision of NRS, yet those episodes still receive an NRS payment.

We are proposing to calculate resource use under the HHGM using a Cost-per-Minute plus Non-Routine Supplies (CPM + NRS) approach. It determines resource use using information from Medicare cost reports. Under the proposed HHGM, we would group episodes into their case-mix groups taking into account admission source, timing, clinical group, functional level, and comorbidity adjustment. From there, the average resource use for each case-mix group dictates the group's case-mix weight. Resource use is the estimated cost of visits recorded on the home health claim plus the cost of NRS recorded on the claims. The cost of NRS is generated by taking NRS charges on claims and converting them to costs using a NRS cost to charge ratio that is specific to each HHA. When NRS is factored into the average resource use, NRS costs are reflected in the average resource use that drives the case-mix weights. CMS would return $53.03 to the base rate (that is, the NRS conversion factor). If there is a high amount of NRS cost for all episodes in a particular group (holding all else equal), the resource use will be higher relative to the average and the case-mix weight will correspondingly be higher. Similar to the current system, NRS would still be paid prospectively under the HHGM, but the HHGM eliminates the separate case-mix adjustment model for NRS. Incorporating the NRS cost into the measure of overall resource use (that is, the dependent variable of the payment model) requires adjusting the NRS charges submitted on claims based on the NRS cost-to-charge ratio from cost report data.

The following steps would be used to generate the measure of resource use under this CPM + NRS approach:

(1) From the cost reports, obtain total costs for each of the six home health disciplines for each HHA.

(2) From the cost reports, obtain the number of visits by each of the six home health disciplines for each HHA.

(3) Calculate discipline-specific cost per visit values by dividing total costs [1] by number of visits [2] for each discipline for each HHA. For HHAs that did not have a cost report available (or a cost report that was trimmed from the sample), imputed values were used as follows:

• A state-level mean was used if the HHA was not hospital-based. The state-level mean was computed using all non-hospital based HHAs in each state.
• An urban nationwide mean was used for all hospital-based HHAs located in a Core-based Statistical Area (CBSA). The urban nation-wide mean was computed using all hospital-based HHAs located in any CBSA.
• A rural nationwide mean was used for all hospital-based HHAs not in a CBSA. The rural nation-wide mean was computed using all hospital-based HHAs not in a CBSA.

(4) From the home health claims data, obtain the average number of minutes of care provided by each discipline across all episodes for a HHA.

(5) From the home health claims data, obtain the average number of visits provided by each discipline across all episodes for each HHA.

(6) Calculate a ratio of average visits to average minutes by discipline by dividing average visits provided [5] by average minutes of care [4] by discipline for each HHA.

(7) Calculate costs per minute by multiplying the HHA's cost per visit [3] by the ratio of average visits to average minutes [6] by discipline for each HHA.

(8) Obtain 30-day period costs by multiplying costs per minute [7] by the total number of minutes of care provided during a 30-day period by discipline. Then, sum these costs across the disciplines for each period.

This approach accounts for variation in the length of a visit by discipline. NRS costs are added to the resource use calculated in [8] in the following way:

(9) From the cost reports, determine the NRS cost-to-charge ratio for each HHA. The NRS ratio is trimmed if the value falls in the top or bottom 1 percent of the distribution across all HHAs from the trimmed sample. Imputation for missing or trimmed values is done in the same manner as it was done for cost per visit (see [3] above).

(10) From the home health claims data, obtain NRS charges for each period.

(11) Obtain NRS costs for each period by multiplying charges from the home health claims data [10] by the cost-to-charge ratio from the cost reports [9] for each HHA.

Resource use is then obtained by:

(12) Summing costs from [8] with NRS costs from [11] for each 30-day period.

Table 20 shows these costs overall for 30-day periods in CY 2015 (n = 8,642,107). On average, total 30-day period costs are $1,585.48. The distribution ranges from a 5th percentile value of $300.03 to a 95th percentile value of $3,908.93.

The distributions and magnitude of the estimates of costs for the two methods are very different. The differences arise because the CPM + NRS method incorporates HHA-specific costs that represent the total costs incurred during a 30-day period (including overhead costs), while the WWMC method provides an estimate of only the labor costs (wage + fringe) related to direct patient care from patient visits that are incurred during a 30-day period. Those costs are not HHA-specific and do not account for any non-labor costs (such as transportation costs) or the non-direct patient care labor costs (such as, administration and general labor costs). Because the costs estimated using the two approaches are measuring different items, they cannot be directly compared. However, if the true cost of a 30-day period is correlated with the labor that is provided during visits, the two approaches should be highly correlated. The correlation coefficient between the two approaches to calculating resource use is equal to 0.8016 (n = 8,642,107). Therefore, the relationship in relative costs is similar between the two methods.

Using cost report data to develop case-mix weights more evenly weights skilled nursing services and therapy services than the BLS data. Table 21 shows the ratios between the estimated costs per hour for each of the home health disciplines compared with skilled nursing resulting from the CPM +NRS versus WWMC methods. Under the CPM+NRS methodology, the ratio for physical therapy costs per hour to skilled nursing is 1.14 compared with 1.40 using the WWMC method.

We believe that using cost report data to calculate the cost of home health care better aligns the case-mix weights with the total relative cost for treating various patients. In addition, using cost report data allows us to incorporate NRS into the case-mix system, rather than maintaining a separate payment system. Therefore, we are proposing to calculate the cost of a 30-day period of home health care under the HHGM using the cost per minute plus non-routine supplies (CPM+NRS) approach outlined above. We invite comments on the proposed methodology for calculating the cost of a 30-day period of care under the HHGM.

3. Change From 60-Day Billing to 30-Day Billing Under the HHGM

a. 30-Day Unit of Payment

Currently, HHAs are paid for each 60-day episode of home health care provided. We are proposing 30-day periods of payment for the HHGM. Through examination of the resources used within a 60-day episode of care, we identified differences in resources used between the first 30-day period within a 60-day episode and the second 30-day period within a 60-day episode. Episodes have more visits, on average, during the first 30 days compared to the last 30 days (see Tables 22 and 23). Costs are much higher earlier in the episode and lesser later on, therefore, dividing a single 60-day episode into two 30-day periods more accurately apportions payments. This difference in resource use between the first and second 30-day period within a 60-day episode is one of the main reasons we are proposing 30-day periods of payment for the HHGM. Another reason for proposing to change the unit of payment from 60-days to 30-days is the removal of the therapy visit thresholds from the case-mix adjustment methodology under the HHGM (the current system accounts for therapy visit variation through the use of these thresholds). Without thresholds being used to account for resource use variation, a shorter period of care is needed to reduce the variation and improve the accuracy of the case-mix weights generated under the HHGM. The HHGM's goodness of fit statistics (for example, R-squared) improve due to reduced resource use variation when a shorter, more constrained time period is examined. Therefore, the case-mix weights and proposed move to a 30-day period under the HHGM better approximate relative resource use. Furthermore, by switching to a 30-day period, the billing cycle for Medicare home health services would be the same as for other Medicare health care settings, such as hospices and SNFs, which currently bill on a monthly basis.

Using two segments of the current 60-day episodes, 30-day periods were constructed as follows for the development of the HHGM:

• A 30-day period comprising days 1-30 of a current 60-day episode where “day 1” is the current 60-day episode's From Date.
• A second period comprising days 31 and above of a current 60-day episode. This period would be 30-days in length if the current episode was 60-days (from the From Date of the episode to the Through Date of the episode) and some lesser length if the current episode were fewer than 60-days.

A typical 60-day episode was broken down into two portions: A first 30-day period; and a second 30-day period consisting of the remaining days. For example, if the current episode was 58 days, then the first period was 30-days, and the second period was comprised of the remaining 28 days. Resource utilization was calculated for each 30-day period based on the discipline visits that occur within each respective 30-day time span. The OASIS information that is applied to the two 30-day periods (for example, OASIS information) is established by the same OASIS that is linked to the current 60-day episode.

Table 22 shows the average number of visits by discipline and resource use estimates during 15-day periods in a 60-day episode, and shows that visit patterns differ over the course of a 60-day episode. Across all labor categories there is a decline in visits as the episode proceeds; in total there are 6.8 visits on average in days 1-15 and 2.6 visits on average in days 46-60 which is a 61.8 percent decline from the first 15 days of care in a 60-day episode to the last 15 days of care in a 60-day episode.

Table 23 shows the average number of visits and resource use estimates by discipline during 15-day periods in a 60-day episode, but for only those episodes that are first in a sequence of episodes and last a full 60-days. A sequence of episodes contains episodes where no more than 60-days elapse from the end of one episode to the start of the next. Therefore, first episodes are those where the beneficiary has not had home health in the 60-days prior to the start of the first episode. Even among this subset of episodes, there is a decline in average visits by quarter as the episode proceeds.

These results show that there is variation in average resource use across 60-day episodes. By moving to two 30-day periods within a 60-day episode (or a single 30-day period if the 60-day episode contains 30 or fewer days), the HH PPS case mix weights better align with the resource use patterns across the current 60-day episode. Though the analyses are based on two 30-day periods in a 60-day episode, we are not proposing a change in the requirements for completing the comprehensive assessment. Under the HHGM, the comprehensive assessment would still be required, as outlined in § 484.55 roughly every 60-days as is required under the current HH PPS. While we examined resource use in 15-day periods in a 60-day episode of care, as outlined in Tables 22 and 23, in order to strike an appropriate balance between increasing payment accuracy and being cognizant of increasing burden for the home health industry, we are not proposing to adjust payments every 15 days. We expect that billing on a 30-day basis should not be completely unfamiliar to HHAs as HHAs billed as such prior to the implementation of the HH PPS.

Overall, approximately 25 percent of episodes are 30 days or less in length, and therefore, would produce no second 30-day period under the HHGM. These episodes (with 30 days or fewer) would convert to only one 30-day period each; any 60-day episode that is 31 days or more would produce two 30-day periods: A first period comprising 30 days in length and then a second period with the remaining days in the 60-day episode.

Overall, after conversion from the 5,110,629 60-day episodes, there were 8,642,107 30-day periods:

• There were 1,197,740 30-day periods that could potentially be one-to-one conversions from 60-day episodes that were 30-days or fewer in length.
• Additionally, there were 3,912,889 60-day episodes that were between 31 and 60-days in length in which two 30-day periods could be produced. That is, those 60-day episodes could produce up to 7,825,778 30-day periods.
• However, from the above episodes (which were used to create the 30-day periods), there were 381,411 periods that had no visits included or were considered a LUPA under the HHGM and therefore were excluded. This is shown in Table 24.

Tables 25 and 26 show the frequency of episode length in days and estimates of resource use among the original, 60-day episodes and the corresponding distribution of episode length and resource use estimates among the simulated 30-day periods. Again, these results show differences by the length of care. By shortening the unit of time that CMS pays for within the HH PPS (from 60-day episodes to 30-day periods), payment would more accurately relate to the variation in costs seen across episodes and periods of care.

The 60-day episode unit of payment was originally implemented on October 1, 2000, because most episodes in the HHA per-episode PPS demonstration, which was used to inform the development of the HH PPS, ended in 60 days or less, the OASIS data would be captured on a 60-day cycle, and Medicare plan of care/certification requirements continue to be bimonthly (64 FR 58143). In the FY 2001 HH PPS proposed rule, we noted that about 60 percent of episodes paid under the HH PPS were completed within one 60-day episode and 73 percent within two 60-day episodes. In the FY 2001 HH PPS final rule, we noted that we would continue to monitor the appropriateness of the 60-day unit of payment, and would consider modifying our approach to the episode definition in subsequent years of PPS, if warranted (65 FR 41136).

In CY 2016, 73 percent of episodes were completed within one 60-day episode and 86 percent within two 60-day episodes. We currently observe wide variation in the length of care in the current HH PPS. Overall, the average length of home health care was approximately 46 days, but roughly a quarter of all 60-day episodes lasted 30 days or less. For example, those episodes that had a hospital stay in the seven days prior to the start of the episode where the Diagnosis Related Group (DRG) was either 469 or 470 (major joint replacement or reattachment of lower extremity) had an average length equal to 23.7 days. As noted above, there is a decline in visits as the episode proceeds with a 61.8 percent decline from the first 15 days of care in a 60-day episode to the last 15 days of care in a 60-day episode.

The wide variation in resource use and trends toward shorter episodes of care, the difference in resources between the first and second 30-day period within a 60-day episode, and the removal of the therapy visit thresholds from the case-mix adjustment methodology (which currently account for variation in resource use, but create adverse incentives as outlined in section II.D of this proposed rule) result in less accurate case-mix weights. When a shorter, more constrained time period is used for payment, the HHGM's goodness of fit statistics (for example, R-squared) improve due to reduced resource use variation. Accordingly, the case-mix weights under the HHGM better approximate relative resource use. Therefore, we are proposing to change the unit of payment under section 1895(b)(2) of the Act from a 60-day episode of care to 30-day periods of care. Section 1895(b)(2) of the Act requires the Secretary to consider potential changes in the mix of services provided within that unit and their cost. Our analysis shows evidence of a change in the mix of services under a 60-day episode of care, as outlined above and in section II.D of this proposed rule. Therefore, to better account for changes in the mix of services over time; to ensure that the unit of payment reflects an appropriate number, type, and duration of visits provided within a unit of payment; and to provide continued access to quality services, we are proposing to change the unit of payment from a 60-day episode of care to a 30-day period of care and to implement case-mix adjustment methodology refinements, outlined in sections III.E.1 through III.E.12 of this proposed rule.

b. National, Standardized 30-Day Payment Amount

We note that we propose to implement the HHGM for 30-day periods of care beginning on or after January 1, 2019. As a result, we would calculate a proposed national, standardized 30-day payment amount in the CY 2019 HH PPS proposed rule. In calculating a national, standardized 30-day payment amount for CY 2019, we propose to start with the CY 2019 national, standardized 60-day episode payment amount reflecting the HHA market basket update as specified in section 1895(b)(3)(B) of the Act, add back in the CY 2019 non-routine medical supply (NRS) conversion factor amount reflecting the HHA market basket update as specified in section 1895(b)(3)(B) of the Act, and then divide the sum by two.

If we had proposed to implement the HHGM in CY 2018, we would have calculated a proposed 30-day payment amount for CY 2018 by starting with the CY 2018 proposed national, standardized 60-day episode payment amount of $3,038.43, adding back in the CY 2018 proposed NRS conversion factor amount of $53.03, and dividing the sum by two to produce a 30-day payment amount of $1,545.73. However, we reiterate that we propose to implement the HHGM for 30-day periods of care beginning on or after January 1, 2019; so we propose to calculate a national, standardized 30-day payment amount for CY 2019 using the CY 2019 60-day episode payment amount, adding back in the CY 2019 NRS conversion factor and dividing the sum by two to produce a 30-day payment amount. Finally, we note that the calculation proposed above would only be used to calculate a national, standardized 30-day payment amount for CY 2019. To calculate a national, standardized 30-day payment amount for CY 2020 and subsequent years, we would update the national, standardized 30-day payment amount from the immediate preceding year by the home health payment update percentage required by the statute, as described in section III.C.1 of this rule.

In determining the 30-day payment amount, we evaluated whether starting with the national, standardized 60-day episode payment amount, adding back in the NRS conversion factor amount and dividing the sum by two was an appropriate estimate of the cost of a 30-day period of care. Section 1895(b)(3) of the Act provides a methodology for determining an initial payment amount for the PPS and for calculating annual increases. As noted in this proposed rule, the Act at section 1895(b)(2) gives the Secretary the discretion to determine the “unit of payment” (also referred to in the statute as a “unit of service”) on which a standard prospective payment amount would be based. Since we are proposing to change the unit of payment, we believe it is necessary to calculate a 30-day payment amount that would accurately reflect what a 30-day payment would be had we chosen to use a 30-day rather than a 60-day unit of payment when we first implemented the PPS.

To do this, we calculated an estimated 30-day payment amount by taking the average number of visits per discipline per 30-day period of care in CY 2016 multiplied by the FY 2001 per-visit amounts (including average NRS costs per visit) initially established under the HH PPS based on the most recent audited cost report data available to the Secretary in accordance with section 1895(b)(3)(A)(I) of the Act, as adjusted for inflation and productivity. The FY 2001 per-visit amounts were adjusted for inflation by the actual HHA market basket updates (reflecting historical data from FY 2002 to CY 2016), the regulatory HHA market basket updates for CY 2017 (which is based on the CY 2017 forecasted data at the time of CY 2017 rulemaking (81 FR 76714)) and CY 2018 (which is based on the CY 2018 forecasted data in this CY 2018 proposed rule), and for productivity (using Economy-wide Multifactor Productivity as specified in section 1895(b)(3)(B)(vi) to the Act and described in section 1886(b)(3)(B)(xi)(II) of the Act) beginning in 2015, as reflected in Table 26B.

As shown in Table 28, using the FY 2001 per-visit amounts initially established under the HH PPS results in an estimated 30-day payment amount of $1,494.64. This value is less than, but similar to half the sum of the proposed CY 2018 national, standardized 60-day episode payment amount and proposed CY 2018 NRS conversion factor amount ($1,545.73).

We also calculated an estimated 30-day payment amount by taking the average number of visits per discipline per 30-day period of care in CY 2016 multiplied by the FY 2015 costs-per-visit, per discipline, based on the most recent cost report data available at the time of CY 2018 HH PPS rulemaking (as outlined in Table 2 in section III.A of this proposed rule) and further adjusted to include average NRS costs per visit, for outliers in accordance with section 1895(b)(3)(C) of the Act, and for inflation and productivity. As shown in Table 29, using 2015 costs-per-visit, per discipline, based on the most recent cost report data available at the time of CY 2018 HH PPS rulemaking, results in an estimated 30-day payment amount of $1,485.11. This value is also less than, but similar to half the sum of the proposed CY 2018 national, standardized 60-day episode payment amount and proposed CY 2018 NRS conversion factor amount ($1,545.73).

We believe our proposal to start with the national, standardized 60-day episode payment amount, add back in NRS conversion factor amount, and then divide the sum by two is a reasonable estimate of the cost of a 30-day period of care. We propose to implement the change in the unit of payment from 60-day episodes of care to 30-day periods of care in a non-budget neutral manner. We note that in its March 2017 Report to Congress, MedPAC highlighted that home health payments have consistently and substantially exceeded costs because agencies are able to reduce the number of visits provided and cost growth is generally lower than the annual payment updates for home health care. MedPAC recommended a 5 percent reduction in the base rate for 2018 and a 2-year rebasing beginning in 2019. We invite comments on the proposed calculations for determining the 30-day payment amount, including our rationale for proposing to implement the HHGM in a non-budget neutral manner.

We are further proposing to implement the HHGM in a fully non-budget neutral manner beginning in CY 2019 or alternatively to use a phased approach to implementation. We acknowledge that implementing the HHGM in a partially budget-neutral manner could lessen the economic impact for HHAs in transitioning to the HHGM. Therefore, we considered potential alternative implementation approaches for the HHGM, including, but not limited to, a partially budget-neutral approach with a phase-out period. Specifically, for the phased approach, we propose to apply a HHGM partial budget neutrality adjustment factor in CY 2019 that would reduce the estimated impact of the HHGM from an estimated −4.3 percent to −2.2 percent in the initial year of implementation with the removal of the HHGM partial budget neutrality adjustment factor in CY 2020. We invite comments on whether to implement the HHGM in a fully non-budget neutral manner beginning in CY 2019; whether to alternatively implement the HHGM in CY 2019 with a HHGM partial budget neutrality adjustment factor applied and then subsequently removed in CY 2020; or whether a HHGM partial budget neutrality adjustment factor should be applied and then phased-out over a longer period of time.

c. Split Percentage Payment Approach for 30-Day Periods of Care

In the current HH PPS there is a split percentage payment approach to the 60-day episode. The first bill, a Request for Anticipated Payment (RAP), is submitted at the beginning of the episode. The second, final bill is submitted at the end of the 60-day episode of care. An initial percentage payment of 60 percent of the anticipated final claim payment amount is paid at the beginning of the episode and a final percent payment of 40 percent is paid at the end of the episode. For all subsequent episodes for beneficiaries who receive continuous home health care, the episodes are paid at a 50/50 percentage payment split. A new initial and final bill must be submitted for each 60-day episode period. HHAs are encouraged to submit the RAP as soon as possible after care begins to assure being established as the primary HHA for the beneficiary and so that the claims processing system is alerted that a beneficiary is under a HH episode of care to enforce the consolidating billing edits required by law.

We are not proposing a change to the split percentage payment approach in conjunction with proposing to change the unit of payment from a 60-day episode to a 30-day period of care. Under the proposed HHGM, we propose that the initial payment for initial 30-day periods would be paid at 60 percent of the case-mix and wage-adjusted 30-day payment rate. The residual final payment for initial 30-day periods would be paid at 40 percent of the case-mix and wage-adjusted 30-day payment rate. We propose the initial payment for subsequent 30-day periods would be paid at 50 percent of the case-mix and wage-adjusted 30-day payment rate. The residual final payment for subsequent 30-day periods would be paid at 50 percent of the case-mix and wage-adjusted 30-day payment rate.

However, we note the length of time HHAs currently take to submit the RAP indicates that the RAP payment might not be necessary for the majority of HHAs to maintain an adequate cash flow (see Table 30). Approximately 5 percent of RAPs (95th percentile) are not submitted until the end of an episode of care and the median length of days for RAP submission is 12 days from the start of the episode. In addition, eliminating RAP payments would address existing program integrity vulnerabilities. For example, $1.8 billion in RAP payments (July 1, 2015 through July 31, 2016) were auto-cancelled, and of that amount, a final claim was never submitted for $321 million worth of RAP payments.

We are soliciting comments as to whether the split payment approach would still be needed for HHAs to maintain adequate cash flow if the unit of payment changes from 60-day episodes to 30-day periods of care under our proposal. In addition, we are soliciting comments on ways to phase-out the split percentage payment approach in the future if the proposed HHGM is finalized with the split percentage payment approach being initially maintained. Specifically, we are soliciting comments on reducing the percentage of the upfront payment over a period of time. We believe that payment based on 30-day periods would reduce, if not eliminate, the need for these partial, up-front payments that occur in the current payment system. Home health agencies would bill on a monthly basis, similar to hospices and SNFs, and thus receive final payment sooner.

If in the future the split percentage approach was eliminated, we are also soliciting comments on the need for HHAs to submit a notice of admission within 5 days of the start of care to assure being established as the primary HHA for the beneficiary and so that the claims processing system is alerted that a beneficiary is under a HH period of care to enforce the consolidating billing edits required by law.

We invite comments on the proposed change in the unit of payment from a 60-day episode of care to a 30-day period of care under the HHGM; the calculation of the national, standardized 30-day payment amount, initially maintaining the split percentage payment approach and applying such policy to 30-day periods of care; and the associated regulations text changes outlined in section III.E.13. of this proposed rule. We are also soliciting comments on ways the split percentage payment approach could be phased-out and whether to implement a notice of admission process if the split percentage payment approach is eliminated in the future.

4. Episode Timing Categories

To advance the goals of better aligning payment with patient needs, as well as addressing payment incentives and vulnerabilities within the current system, we investigated the impact of episode timing on home health resource use. In the current payment system, 60-day episodes are classified as “early” if they are the first or second in a sequence of episodes and “late” if they are the third or later in the sequence. Episodes are defined as being in the same sequence if there are no more than 60 days between the end of one episode and the start of the next. In the development of the proposed HHGM, we sought to evaluate whether payments to providers appropriately reflect the varying resource needs of home health beneficiaries during various portions of the home health stay, accounting for contrasting patient characteristics.

We endeavored to evaluate whether beneficiaries in their first 30-day period of care have different needs and patterns of resource use than those in later 30-day periods, thus possibly resulting in the potential need for differentiated payment amounts. We reviewed related research, held technical and clinical expert panels, and performed our own investigative analyses. In particular, we were interested in whether home health patients utilize more resources at the beginning of home health than in later periods of the home health stay, and, if so, does the current payment structure sufficiently account for this elevated need. In a review of research related to episode timing, studies show that more frequent skilled visits in the first few weeks of a home health stay can prove beneficial for certain diagnoses by reducing the likelihood of readmission to an institutional setting and easing the transition from hospital to home, which can be challenging for patients.

The Visiting Nurse Associations of America defines “frontloading” as the practice of providing an increase in intensity of visits during the first two to three weeks of the home health care episode for patients that have been determined to be at high risk for hospitalization. A 2014 literature review titled “Frontloading and Intensity of Skilled Home Health Visits: A State of the Science” found that Medicare patients benefited from an intensified level of care through a “frontloading” approach, which reduced the need for re-hospitalization among skilled home health patients, and especially for those with heart failure. For the purposes of this particular study, frontloading was defined as providing 60 percent of planned visits within the first 2 weeks of the home health episode of care. Furthermore, frontloading was also found by the Briggs® National Quality Improvement/Hospitalization Reduction Study, to be one of 15 best practices routinely employed by 64 percent of the HHAs who were most successful at reducing hospitalizations. Similarly, in an article titled “The Effect of Frontloading Visits on Patient Outcomes,” the authors assessed the impact of frontloading on patients with insulin-dependent diabetes and with heart failure. In their research, the authors found that frontloading was effective for patients with heart failure, decreasing re-hospitalization by more than half (39.4 percent vs. 16 percent), with fewer visits overall (15.5 vs. 9.5) and equal clinical outcomes and patient satisfaction. These improvements in overall outcomes were presumably due to the timing of the services, where more visits were provided in the beginning portion of the episode, even when fewer visits were provided overall. However, we note that there was no significant impact for those patients with diabetes. No specific effect for patients with mental health or behavioral health conditions was noted. Given the potential positive outcomes of the practice of frontloading, specifically for those beneficiaries with heart disease, we expect that HHAs would provide more frequent skilled services in the beginning portion of a home health stay to educate patients in medication management, coordinate the instruction of both the patient and family, and support patients in navigating their clinical situation, especially in cases of heart disease. The first and fourth reported top primary reasons for home health care in CY 2016 were hypertension and heart failure, respectively, and we therefore believe an opportunity exists for HHAs to improve the outcomes for these high-volume home health beneficiaries by providing more resources in the early period of a home health stay.

For many patients admitted to home health, the transition from hospital or other institutional settings back to the home environment can be very challenging and lead to adverse effects for the beneficiary, such as medication errors, harmful drug events, and additional complications. The provision of intensified home health services early in a home health stay can potentially help to mitigate any negative events that could result from this time of transition from the institutional setting to the home. As such, we would expect that beneficiaries would require more resources, particularly from skilled disciplines providing teaching and medication management, during the first 30 days of a home health admission.

As described in section III.E.3 of this proposed rule, analysis of home health data demonstrates that HHAs provide more services in the first 30-day period of home health than in later periods of care. The differences in the resource utilization during home health episodes are presented in Table 22, which shows the average resource use of home health episodes divided into 15-day segments. The first two 15-day periods in a home health episode have significantly higher average resource use at $261.97 and $162.44, respectively, as compared with the third and fourth 15-day segments in a 60-day period, at $107.49 and $88.67, respectively. Additionally, the average number of visits by the six disciplines is also significantly higher in the first two 15-day segments, at 6.8 and 4.9 visits per segment, respectively as compared to the third and fourth 15-day segments of a 60-day episode, at 3.3 and 2.6, respectively.

Further analysis of home health data demonstrates that under the current payment system, when analyzed by 30-day periods, HHAs provide more resources in the first 30-day period of home health (“early”) than in later periods of care. The differences in the average resource use during early and late home health episodes when divided into 30-day periods are presented in Table 28, and shows the first 30-day periods in a home health sequence have significantly higher average resource use at $2,102.29 as compared with subsequent 30-day periods. Specifically, the later 30-day periods showed an average resource use of $1,348.18, a difference of more than $700 or a 36 percent decrease. Table 31 also shows a significant difference between the early and late episode median values of resource use. The median for the first 30-day period is $1,848.12, while the median for subsequent 30-day periods is $987.54, a difference of more than $850 or an approximately 47 percent decrease.

There is significant difference in the resource utilization between early and late 30-day periods as demonstrated in Table 31. Moreover, the predictive power of the HHGM in terms of estimating resource utilization improved when separating episodes into 30-day periods rather than 60-day periods (that is, the first and second 30-day periods). We believe that an HHGM that accounts for the demonstrated increase in resource utilization in the first 30-day period better captures the variations in resource utilization and further promotes the goal of payment accuracy within the HH PPS. We are proposing to classify the 30-day periods under the proposed HHGM as “early” or “late” depending on when they occur within a sequence of 30-day periods. For the purposes of defining “early” and “late” periods for the proposed HHGM, we are proposing that only the first 30-day period in a sequence of periods be defined as “early” and all other subsequent 30-day periods would be considered “late”. Additionally, we are proposing that the definition of a “home health sequence” (as currently described in § 484.230) will remain unchanged relative to the current system, that is, 30-day periods are considered to be in the same sequence as long as no more than 60 days pass between the end of one period and the start of the next, which is consistent with the definition of a “home health spell of illness” described at section 1861(tt)(2) of the Act. We note that because section 1861(tt)(2) of the Act is a definition related to eligibility for home health services as described at section 1812(a)(3) of the Act, it does not affect or restrict our ability to propose a 30-day prospective payment period.

To identify the first 30-day period within a sequence, the Medicare claims processing system would verify that the claim “From date” and “Admission date” match. If this condition were to be met, our systems would send the “early” indicator to the HH Grouper for the 30-day period of care. When the claim is received by CMS's Common Working File, the system would look back 60 days to ensure there is not a prior, related episode. If not, the claim would continue to be paid as “early.” If another related episode were to be identified, that is an earlier 30-day period in the sequence, the claim would be returned to the shared systems for subsequent regrouping and re-pricing. Those periods that are not the first 30-day period in a sequence of adjacent periods, separated by no more than a 60 day gap, would be categorized as “late” periods and placed in corresponding HHGM categories.

We invite public comments on the timing categories in the proposed HHGM and the associated regulations text changes outlined in section III.E.13 of this proposed rule.

5. Admission Source Category

In accordance with the statute, as amended by the BBA, we published a final rule in the July 3, 2000 Federal Register (65 FR 41128) implementing the HH PPS. In that final rule, we discussed and finalized the use of a methodology that included variables identifying pre-admission location (that is, whether certain inpatient and other stays occurred in the 14-day period immediately preceding the home health episode) as part of our case-mix adjustment methodology. We stated that not only were pre-admission inpatient stays a traditional indication of need in clinical practice, but also that such variables were useful correlates of resource cost in our evaluation of the home health case-mix data (65 FR 41146). This pre-admission information was submitted by HHAs via OASIS assessments.

In the CY 2008 HH PPS final rule, we removed elements from the case-mix adjustment methodology that were based upon the source of admission (72 FR 49766). In the CY 2008 HH PPS proposed and final rules, we assessed variables for policy and payment appropriateness and ultimately decided to remove the variable that had been used to identify the patient's pre-admission location from the case-mix adjustment methodology (72 FR 25361 and 72 FR 49766, respectively). This decision was based, in part, upon concerns that some agencies were encountering challenges in obtaining concrete information regarding the patient's preadmission location while performing the initial home health assessment and thus the OASIS item used to indicate the preadmission location of the patient was not always reliable. Moreover, the pre-admission information did not perform well in terms of the four-equation model used for payment estimation and also had a small impact in terms of payment accuracy within the model. In the CY 2008 HH PPS final rule, we further noted that the item's results across the four equation model created difficulties in terms of interpretation and the explanatory power (for example, its contribution to the R-squared value) was minimal (72 FR 49766).

For the purposes of constructing the HHGM, which would not use a 4-equation model or otherwise adjust payments based on therapy visit thresholds; we reexamined the impact of beneficiary admission source, either from the community or from an institutional setting, on home health resource use. In our review of related scholarly research, we found that beneficiaries admitted directly or recently from an institutional setting (acute or post-acute care (PAC)) tend to have different care needs and higher resource use than those admitted from the community, thus indicating the need for differentiated payment amounts. For instance, a literature review of 25 research studies published between 2002 and 2011, titled “Hospitalization Among Medicare-Reimbursed Skilled Home Health Recipients,” found that Medicare beneficiaries discharged from PAC and acute facilities differ significantly in resource need when compared to community-admitted beneficiaries. Patients discharged from acute and PAC settings tend to be sicker upon admission and are being discharged rapidly back to the community. Additionally, they are more likely to be re-hospitalized after discharge due to the acute nature of their illness. One study discussed in this literature review determined that patients being discharged from an inpatient hospitalization typically present with multiple comorbidities, suggesting that initially-hospitalized patients subsequently transferred to home care were more likely to have four or more secondary diagnoses, as well as a pressure or stasis ulcer, urinary incontinence, a urinary catheter, depression, or dyspnea. They generally had more than five medications than their non-hospitalized counterparts and required assistance with medication management. As such, patients referred to home health after an institutional stay tend to be more infirm, requiring significant resources upon admission to home health. Additionally, the same literature review also highlighted a study titled “Unplanned hospital readmissions: A home care perspective” that demonstrated that patients referred from acute and PAC settings are at a high risk of hospitalization within 14 to 21 days of admission to home health. Given that the first few weeks after an institutional stay represent a critical window in terms of providing beneficiaries with appropriately intensive supports and services, as well as preventing re-hospitalization, we would expect that providing care for those beneficiaries admitted from institutional settings would require more resource use compared to patients admitted to home health from the community. Comprehensive and deliberate interventions in this timeframe could also potentially reduce re-hospitalization rates.

Research studies also demonstrate that patients admitted to home health from institutional settings are more vulnerable to adverse effects and injury because of the functional decline that occurs as a result of their institutional stay, indicating that this patient population requires more concentrated resources and supports to account for and mitigate this functional decline. In the article titled “The Incidence and Severity of Adverse Events Affecting Patients after Discharge from the Hospital,” Alan J. Forster, MD noted that beneficiaries are susceptible to harm post-hospitalization: “Patients may be especially vulnerable to injuries during this [post-discharge] period because they may still have functional impairments and because discontinuities may occur at the interface of acute and ambulatory care.” The author also notes that the current health care environment encourages potentially expedited discharges from hospital stays, “in which patients are leaving the hospital ‘quicker and sicker.’ ” Patients may be leaving the hospital environment in a tenuous and fragile state, leaving them vulnerable to further harm once returned to the home environment. Additionally, the change from constant monitoring in the inpatient facility to less frequent monitoring in the home environment can potentially cause gaps in care and consequently increased risk for adverse events for the newly-admitted home health beneficiary. The article notes that many of the negative impacts of the transition can be reduced by an appropriate increase in care for the beneficiary in the home setting, notably with more frequent assessment of their condition and ongoing monitoring. Therefore, we believe that an opportunity may exist for the HHGM to account for this increased need and accordingly provide a differentiated payment to facilitate the provision of more frequent assessments and monitoring for beneficiaries admitted to home health from acute and PAC settings, which could in turn help prevent re-hospitalizations and adverse events. We expect that HHAs would provide more resource-intensive services after discharge from an institutional setting to educate patients in new medication management, facilitate discharge education for the patient and family, and provide support in the recovery from the illness that caused the originating hospitalization or institutional stay.

In the guidebook “Patient Safety and Quality: An Evidence-based Handbook for Nurses,” authors Ruth M. Kleinpell, Kathy Fletcher, of and Bonnie M. Jennings note in chapter 11 that deconditioning, a status characterized by a “decrease in muscle mass and the other physiologic changes related to bed rest, contributes to overall weakness,” has become commonplace in the post-institutional beneficiary population. This physiological weakening of the institutionalized beneficiary can then, in turn, lead to significant functional decline, resulting in reduction in ability to perform Activities of Daily Living (ADLs), and ultimately in increased home health resource utilization. The article notes that hospitalization of the elderly is usually marked by decreased levels of mobility and increased levels of bed rest, with deterioration from their baseline levels as soon as day two of the hospitalization. Hence, a hospitalization itself leads to declines in mobility, which consequently yields reduced functionality in patients relative to their status before their inpatient stay. This decline in functional ability likewise merits appropriate skilled services to support the patient's increased needs after a hospital stay.

In the article “Determinants of health after hospital discharge: Rationale and design of the Vanderbilt Inpatient Cohort Study (VICS),” the authors describe the period after a hospitalization as a “vulnerable time” for patients. This vulnerability is due to a number of factors, including the need to manage new health care issues, major modifications to medication interventions, and the coordination of follow-up appointments, all while a beneficiary strives to recuperate after a hospital stay for an acute medical event. Of particular concern are the risks for adverse drug events, for errors in a beneficiary's medication regimen, and for the need to readmit to the hospital due to deterioration of the patient's condition. Given the risks during this intense, challenging, and potentially costly period after discharge, we would expect that beneficiaries would require more visits from skilled disciplines, particularly for the purpose of teaching and medication management. This increased utilization of resources would, in turn, warrant a differentiated, potentially higher payment for such services, and the proposed HHGM payment system refinement could account for this difference with varying payment amounts based upon admission source. We note that we do not expect the source of the patient's admission would lead to an HHA furnishing home health services that would replace any orders made by the referring physician regarding the type or frequency of services the patient might need during the home health stay. The admission source variable in the proposed HHGM is meant to serve as a meaningful indicator of resource utilization, which determines Medicare payment. The HHA, in consultation with the physician and ordered by the physician, will continue to articulate, in the plan of care, what services are required to meet the needs of the patient, as well as the frequency of such services.

With regard to beneficiaries admitted to home health from the community, research related to home health admission source demonstrates that community-admitted beneficiaries tend to receive care from the less-costly disciplines. In its 2016 Report to Congress, MedPAC noted that, in their analysis of CY 2013 HH claims, beneficiaries admitted from the community tend to receive more visits from home health aides than their non-community counterparts, stating that “aide services were the majority of services provided in 14 percent of the episodes for community-admitted users compared with 5 percent for PAC users.” However, these same community entrants averaged 2.6, 60-day episodes, while the institutional admits averaged only 1.4, 60-day episodes, demonstrating longer lengths of stay for the community-admitted beneficiaries than those entering from institutional settings. These findings suggest that beneficiaries admitted to home health from the community typically require less resources but for longer periods of time when compared to the beneficiaries admitted from an institutional stay. Additionally, a 2001 Department of Health and Human Services Office of Inspector General study found Medicare home health referrals coming from the community (in this case defined as a referral for a beneficiary who had not been admitted to an overnight stay in a hospital or skilled nursing facility for 15 days prior to beginning a home health care episode) were more likely to have chronic conditions than those referred from hospitals, and therefore, were more likely to require ongoing but less resource-intensive care.

In addition to our review of related research, we also evaluated home health utilization and patient assessment data as described in section III.E.1 of this proposed rule, and our findings demonstrate that those beneficiaries admitted from PAC, as well as acute care settings demonstrate higher resource utilization than their community-admitted counterparts.

The differences in care needs during home health based on admission source are illustrated in the resource utilization figures presented in Table 32, which shows the distribution of admission sources as well as average resource use for 30-day periods by admission source. Institutional admissions have significantly higher average resource use at $2,165.06 compared with community admissions at $1,393.10, a difference of $771.96. Median values of resource use also show a significant difference between sources of admission, with institutional resource use at $1,899.41 while community resource use is at $1,060.51, a difference of nearly $840. The pattern of higher resource use for institutional admissions as compared to community admissions continues for the 25th and 75th percentiles, with a difference of approximately $700 and $900, respectively.

For all of these reasons, we are proposing to establish two admission source categories for grouping 30-day periods of care under the HHGM—institutional and community—as determined by the healthcare setting utilized in the 14 days prior to home health admission. We are proposing the institutional category would include 30-day periods of care for patients admitted from either acute care or PAC settings. Thirty-day periods for beneficiaries with any inpatient acute care hospitalizations, skilled nursing facility stays, inpatient rehabilitation facility stays, or long term care hospital stays within the 14 days prior to a home health admission would be designated as institutional admissions. Similarly, we are proposing that the institutional admission source category would also include patients that had an acute care hospital stay during a previous 30-day period of care and within 14 days prior to the subsequent, contiguous 30-day period of care and for which the patient was not discharged from home health and readmitted (that is, the admission date and from date for the subsequent 30-day period of care do not match) as we acknowledge that HHAs have discretion as to whether they discharge the patient due to a hospitalization and then readmit the patient after hospital discharge. However, we would not categorize post-acute care stays that occur during a previous 30-day period and within 14 days of a subsequent, contiguous 30-day period of care (that is, the admission date and from date for the subsequent 30-day period of care do not match) as institutional as we would expect the HHA to discharge the patient if the patient requires post-acute care in a different setting (for example, a SNF or IRF) and then readmit the patient, if necessary, after discharge from such setting. If the patient is discharged and then readmitted to home health, the admission date and from date on the 30-day claim will match and the claims processing system will look for an acute or a post-acute care stay within 14 days of the home health admission date. This admission source designation process would be applicable to institutional stays paid by Medicare or any other payer. All other 30-day periods would be designated as community admissions.

We initially investigated maintaining two separate institutional categories, one for PAC and another for acute care settings, to identify any meaningful differences in resource use. However, we observed similar resource use in those cases where the patient was admitted from both PAC and acute care settings. Furthermore, in our analysis of the data from CY 2013, we found that the volume of home health cases with an admission from PAC settings across all 30-day periods of care was a low value at 736,112 cases (approximately 8 percent) out of a total of 8,539,996 cases as compared with cases admitted from acute settings at 1,376,567 cases (approximately 16 percent). The number of cases admitted from acute settings was approximately double the number of cases admitted from PAC settings. Moreover, in the creation of case-mix groups that differentiated between community, acute, and PAC admission sources, there were some case-mix groups with a very low number of 30-day periods of care, which in turn can result in substantial variability in the average resource use from year- to- year. We were concerned that this variability could introduce unnecessary instability in the case-mix weights under the proposed HHGM. As such, we are proposing to group 30-day periods of care for patients admitted from acute care and PAC settings together as “institutional” admissions.

We also considered the employment of a “look-back” period for determining the admission source that was longer than 14 days and thus examined data for a longer 30-day “look-back” period to assess the resource utilization for patients admitted to home health from institutional and community settings; however, our findings indicated that there is only a slight difference in resource use, as well as volume of beneficiaries utilizing PAC or acute services before home health between the two timeframes. Table 33 shows the distribution of 30-day periods and average resource utilization with admission source categories now defined by service use for beneficiaries in the 30 days prior instead of 14 days prior. In general, results are similar to those for the 14-day look-back period when compared to the 30-day “look-back” window. Average resource use under a 14-day “look-back” period for institutional entrants is at $2,165.06 while the 30-day entrants show an average resource use of $2,140.40. The same similarity holds true for community entrants, who show an average resource use of $1,393.10 for the 14-day “look-back” period versus $1,382.38 under the 30-day window. We note that the 30-day “look-back” period only produces a slightly higher proportion of institutional periods of care, at 2,315,557 periods as compared with the 14-day period value of 2,153,712, a difference of approximately 10 percent.

We believe that a 14-day “look-back” period is more likely to be directly related to the patients' need for home health care than a 30-day “look-back” period. This would also be more intuitive for HHAs, as the OASIS item M1000 specifically assesses whether a beneficiary was discharged from an institutional setting within the past 14 days. Thus, we ultimately are proposing to use the 14-day “look-back” period as we believe it will better categorize those beneficiaries with a relatively short transition between institutional care and home health care. Given that beneficiary admission source has previously been utilized for the purposes of Medicare home health payment, HHAs will be familiar with this concept. Moreover, the proposed 14-day “look-back” period simplifies the structure of the proposed model and limits burden on claims systems and related processing. Additionally, a “look-back” period of 14 days is consistent with section 1861(tt)(1) of the Act, which defines the term “post-institutional home health services”.

To differentiate between an institutional and community admission source, we would establish an evaluation process whereby the Medicare claims processing system would check for the presence of an acute/post-acute Medicare claim occurring within 14 days of the home health admission on an ongoing basis. In the past, HHAs stated that they had encountered challenges in terms of identifying the source of admission for home health beneficiaries, and we believe that an automated systems approach where Medicare systems evaluate for the presence of an institutional claim within the 14-day “look-back” window will serve to overcome this earlier challenge. Under this approach, the Medicare systems would only evaluate for whether an acute/post-acute Medicare claim occurring within 14 days of the home health admission was processed by Medicare, not whether it was paid.

Moreover, we propose that newly-created occurrence codes would also be established that would allow HHAs to manually indicate on Medicare home health claims an institutional admission source prior to an acute/post-acute Medicare claim, if any, being processed by Medicare systems. We note that the use of these occurrence codes would not be limited to home health beneficiaries for whom the acute/post-acute claims were paid by Medicare. HHAs would also use the occurrence codes for beneficiaries with acute/post-acute care stays paid by other payers, such as the Veterans Administration. Although a home health claim with a non-Medicare institutional admission source can be categorized by the HHA as an institutional admission and paid accordingly, we may conduct medical review as discussed below. We expect home health agencies would utilize discharge summaries from institutional providers to inform the usage of these occurrence codes. We note that these discharge documents should already be part of the beneficiary's home health medical record used to support the certification of patient eligibility as outlined in § 424.22(c).

If an occurrence code is submitted on the home health claim, the home health claim would be categorized as an institutional admission. However, if a home health claim is submitted without an institutional admission occurrence code, thereby categorizing it with a community admission source, and later an acute/post-acute Medicare claim for an institutional stay occurring within 14 days of the home health admission is submitted within the timely filing deadline and processed by the Medicare systems, the home health claim would be automatically adjusted and re-categorized as an institutional admission and appropriate payment modifications would be made. Our systems would adjust community-admitted home health claims on a claim-by-claim, flow basis if an acute/post-acute Medicare claim for an institutional stay occurring within 14 days of the home health admission is received. Given that our systems can only evaluate for the presence of a Medicare acute/post-acute claim, if there was a non-Medicare institutional stay occurring within 14 days of the home health admission but the HHA was not aware of such a stay, upon learning of the institutional stay, the HHA would be able to resubmit a home health claim that included an occurrence code, subject to the timely filing deadline, and payment adjustments would be made accordingly.

Conversely, if an occurrence code is submitted on the home health claim along with dates of the institutional stay, and an acute/post-acute Medicare claim for an institutional stay occurring within 14 days of the home health admission is not subsequently submitted within the timely filing deadline and processed by the Medicare systems, or an acute/post-acute Medicare claim for an institutional stay occurring within 14 days of the home health admission was submitted but later denied for payment, we may conduct post-payment medical review of the home health claim to determine whether the admission was in fact preceded by an institutional stay occurring within 14 days of the home health admission. If upon medical review a determination is made that the admission was not from an institutional setting, we would take appropriate administrative action, including correcting any improper payments and potentially referring the provider to another CMS review contractor for further review or investigation. In summary, we believe that allowing HHAs to submit a claim with an institutional admission occurrence code for a beneficiary with either a Medicare or non-Medicare institutional admission source would enable HHAs to receive appropriate payment for the home health services, while also allowing us the opportunity and flexibility to verify the source of the admission and correct any improper payments as deemed appropriate.

For the purposes of a RAP, we would only adjust the final home health claim submitted for source of admission. For example, if a RAP for a community admission was submitted and paid, and then an acute/post-acute Medicare claim was submitted for that patient before the final home health claim was submitted, we would not adjust the RAP and would only adjust the final home health claim so that it reflected an institutional admission. Additionally, HHAs would only indicate admission source occurrence codes on the final claim and not on any RAPs submitted.

We invite public comments on the admission source component of the proposed HHGM payment system.

6. Proposed Clinical Groupings

a. Background

As discussed in section II.D of this proposed rule, the Home Health Study Report to Congress found that the current payment system may encourage HHAs to select certain types of patients over others, as some clinical sub-groups within the current case mix system are associated with lower margins. These sub-groups include patients with a higher severity of illness, including those receiving a greater level of skilled nursing care; for example, patients with wounds, with ostomies, or who are receiving total parenteral nutrition or mechanical ventilation. Additionally, the Medicare Payment Advisory Commission (MedPAC) has expressed concerns that the HH PPS disincentivizes care for patients needing skilled nursing visits, thereby limiting access of care to the most clinically vulnerable patient populations.

Although the clinical domain of the current case-mix system accounts for whether or not the patient has one or more certain clinical conditions, there could be improvements in clarity regarding patient needs to clearly explain resource use and cost. Given that payment should be predicated on resource use, providing additional clinical groups in the case-mix system and adjusting payment based on identified clinical characteristics and associated services, along with other patient variables, should better align payment with resource use. As such, under the HHGM, we propose grouping 30-day periods of care into six clinical groups designed to capture the most common types of care that HHAs provide. The proposed groups mirror how clinicians differentiate between patients as to what types of care they are receiving. To inform the development of the clinical groups, Abt Associates and CMS conducted an extensive review of diagnosis codes to identify the primary reasons for home health services under the Medicare home health benefit. The workgroup developed six clinical groups reflecting the reported principal diagnosis, clinical relevance, and coding guidelines and conventions, see Table 34.

The 30-day periods of care were assigned to one of the six clinical groups based on the reported principal diagnosis. However, roughly 19 percent of 30-day periods could not be assigned to a clinical group based on principal diagnosis alone. Reasons for the inability to group 30-day periods based on primary diagnoses included codes that were too vague, meaning the code did not provide adequate information to support the need for home health services (for example, T14.90 Injury, unspecified); codes that would not be Medicare covered services in other settings (for example, dental codes); codes that would be unlikely to require skilled home health services (for example, R68.89 Other general symptoms and signs); codes that indicate death as the outcome (for example, G93.82, Brain death); manifestation codes, where coding guidelines require an etiology code to be reported as a principal diagnosis (for example, I39 Endocarditis and heart valve disorders in diseases classified elsewhere); or code first, meaning the diagnosis is subject to sequencing conventions under ICD-10-CM, where the underlying condition must be sequenced first (for example, dementia in Parkinson's disease, in which Parkinson's disease must be sequenced first). In these instances, 30-day periods were considered “questionable encounters” and secondary diagnosis codes were examined to group the 30-day period of care. An ICD-10-CM list with all of the codes that would assign 30-day periods into the six clinical groupings can be found on CMS's HHA Center Web page at https://www.cms.gov/center/provider-Type/home-Health-Agency-HHA-Center.html. More information on the analysis and development of the groupings can be found in the HHGM technical report, also available on the HHA Center Web page. Table 35 shows the distribution of episodes and associated resource use across the six clinical groups.

Table 35 illustrates the differences in average resource use between 30-day periods with similar care needs. Under the HHGM, we propose that each 30-day period would be assigned to a clinical group according to the primary reason the patient was receiving home health, which would be derived from the principal diagnosis code reported on the home health claims. If a 30-day period of care could not be grouped based on the home health reported principal diagnosis due to the reasons listed above, we propose that the claim for that 30-day period would remain a questionable encounter and be returned to the provider for more accurate or definitive coding. Upon publication of this proposed rule, we will post a complete list of ICD-10 codes and their assigned clinical groupings on the CMS HHA Center Web page (https://www.cms.gov/center/provider-Type/home-Health-Agency-HHA-Center.html) to allow ample time for HHAs to understand those codes which would be considered a “questionable encounter.” We believe this will help to minimize any returned claims for more definitive coding. Each code should be reported to the level of certainty and specificity known for the home health admission. Under our proposal, secondary diagnosis codes would not be used to assign the clinical group, as the intent of the HHGM is to increase clarity by classifying the 30-day period based on the primary reason for home health services. Although the principal diagnosis code is the basis for the home health period, secondary diagnosis codes would then be used to case-mix adjust the period further through additional elements of the model, such as the comorbidity adjustment. Using principal diagnoses as the core of the model would create a clinically intuitive payment system that more clearly identifies the types of patients that are treated in home health. Diagnosis codes would also provide clarity and transparency since they are clearly described and reported on claims and other care tools. Additionally, they would support medical necessity for services furnished, and provide information for establishing the home health plan of care. Ultimately, developing clinically similar groups based on the reported principal diagnosis as part of the larger structure of the model would allow for more meaningful analysis of home health resource use, ensure that patients are receiving care commiserate with their level of need, and more accurately align payment with cost.

b. Musculoskeletal and Neuro/Stroke Rehabilitation

Rehabilitation is an integral part of recovery following an illness, injury, or surgical procedure, whether due to a neurological or a musculoskeletal condition. Given that different care goals and expected outcomes of neuro-rehabilitation and musculoskeletal rehabilitation affect resource use, the clinical groups in the HHGM would differentiate between the two. Patient characteristics between the two groups determine whether resources are directed towards preventing the loss of function or slowing the rate of loss of function; improvement or restoration of function; compensation for lost function; and maintenance of current function. Musculoskeletal rehabilitation focuses on individuals with impairments or disabilities due to disease, disorders, or trauma to the muscles or bones, whereas neurological rehabilitation is designed for individuals with disease, trauma, or disorders of the nervous system. Rehabilitation following a stroke, for instance, is primarily initiated early and intensively with the most recovery of function occurring within the first 3 months; however, reacquiring the skills to perform ADLs may be an on-going process depending on the extent and area of injury. However, if improvement or recovery are not expected or achieved, the focus of therapy may shift to maintenance to prevent further decline. Therefore, the VA Clinical Practice Guidelines for Management of Stroke Rehabilitation “strongly recommend that rehabilitation therapy should start as early as possible, once medical stability is reached” and “recommend that the patient receive as much therapy as needed and tolerated to adapt, recover, and/or reestablish the premorbid or optimal level of functional independence.” Neuro-rehabilitation resource use can encompass evaluation and treatment of impairments in cognitive and spatial functioning, swallowing, communication, and psychological or emotional deficit; whereas musculoskeletal rehabilitation generally focuses on evaluation and treatment of the impaired muscle, bone, or joint. Musculoskeletal rehabilitation is more targeted toward proprioception, strength, imbalances, orthopedic surgeries, and abnormal functional movement patterns, and generally streamlines resources following a surgery or injury. Because of these clinical differences and associated resource use differences based on variables in length and intensity of rehabilitation, the HHGM would adjust payment between musculoskeletal and neuro/stroke rehabilitation accordingly.

c. Wounds

Wound care is provided in a variety of settings, including in the home. Advances in wound care treatments have increasingly allowed for a wide range of wound therapies to be provided in the home. According to the article “Wound Care Outcomes and Associated Cost Among Patients Treated in US Outpatient Wound Centers: Data From the US Wound Registry,” a “rough population prevalence rate for chronic non-healing wounds in the United States is 2 percent of the general population,” with an estimated cost of caring for these wounds exceeding $50 billion a year. Non-healing, chronic wounds are often found in home health patients considering “prolonged and non-healing connective tissue injuries are often associated with common diseases, such as metabolic disorders, obesity, hypertension, arteriosclerosis, neuropathy, and diabetes mellitus,” which are among the top home health diagnoses.

Surgical wound care is essential at preventing post-operative complications such as surgical site infections (SSIs) and dehiscence. Research has shown that post-discharge SSIs occur in 3 to 5 percent of all surgical patients, and up to 33 percent of patients undergoing abdominal surgery, and that “more than half of patients who develop post-discharge SSIs are readmitted to the hospital, making SSIs the overall costliest healthcare-associated infection.” Home care management of burns requires a variety of resources as “burn patients are unique, representing the most severe model of trauma.” The management of burn injury involves a multidisciplinary approach which may include nurses, occupational and physical therapists, dieticians, and psychosocial experts. Pressure ulcers are associated with an increased risk of morbidity and mortality and have a variety of intrinsic and external factors affecting their incidence and treatment. The incidence of pressure ulcers in home health is projected to rise due to the aging population, increasingly fragmented care, and nursing shortage. Ultimately, wound care depends on a multitude of characteristics driving resource utilization. By highlighting them as a clinical group, the HHGM would recognize the variety of resources and skills that necessitate careful treatment and healing of different types of wounds, and more accurately ascribe resource use to payment.

d. Behavioral Health Care

The World Health Organization (WHO) defines health as “a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity.” As such, behavioral and mental home health is an important clinical group of the HHGM. If all eligibility and coverage criteria are met according to § 409.42, then a patient may receive skilled nursing services for the assessment, treatment, and evaluation of psychiatric conditions. The Home Health Benefit Policy Manual states that “the evaluation, psychotherapy, and teaching needed by a patient suffering from a diagnosed psychiatric disorder that requires active treatment by a psychiatrically trained nurse, and the costs of the psychiatric nurse's services may be covered as a skilled nursing service.” However, the psychiatric care must be furnished by an agency that does not primarily provide care and treatment of mental diseases. Older adults may be more susceptible to psychiatric and behavioral health issues due to limited mobility, bereavement, loss of ability to live independently, or drop in socioeconomic status due to retirement. Although psychiatric and behavioral conditions have different signs, symptoms, and treatment options than physical illness, mental health can have major consequences on physical health. Behavioral health research suggests that “a model of care including solely hospital based provision (usually inpatient and outpatient care) will be insufficient to provide access for people facing barriers to care.” Additionally, the length of stay among Medicare beneficiaries who have been hospitalized for mental illness has declined over the last decade, with patients being discharged to home health rather than extending a hospitalization. For these reasons, behavioral home health remains a crucial aspect of keeping beneficiaries out of the hospital. Distinguishing it as a clinical group delineates the resources associated with the unique care needs of these patients and would more accurately assign payment based on patient characteristics.

e. Complex Nursing Interventions

Understandably, the growing trend toward providing more healthcare services in the community shifts an increasing number of complex nursing interventions to home health. Providing complex nursing interventions in the home reflects a patient population with “more complex health care needs who require more intensive medical services coordinated across multiple providers, as well as a wide range of social supports to maintain health and functioning.” Because of the range and intensity of services needed, these patients tend to generate high resource utilization and associated costs due to the need for a higher level of knowledge and expertise. Additionally, readmission rates can be high in this vulnerable population as patients adjust to their home with therapies generally administered in the hospital or post-acute environment.

For instance, the introduction of home mechanical ventilation is a technological advancement that not only keeps healthcare costs down but also allows patients, whose condition would otherwise necessitate an institutional environment, a maximum quality of life. For example, the results from one study found that long-term mechanical ventilation on average costs $14,500 less per patient, per month when administered at home rather than in an acute or post-acute facility. However, it does not come without challenges. Caregiver competency, evolving technology, changes in patient medical status, and safety of home environment can lead to higher home health resource utilization. Likewise, management of ostomies and vascular access devices (VADs) are associated with higher resource use in the home. The impact on patients living with VADs and ostomies is significant, with research identifying physical, psychological, and social effects. Ostomy and VAD specific challenges or complications may occur initially and persist and change daily as patients learn to troubleshoot and manage life with an ostomy or VAD. Care often requires resources aimed at education and support in addition to physical care. This can be made more challenging by the social and psychological effects that many new patients experience. Under the HHGM, ICD-10-CM codes on the home health claim that identify complex nursing interventions as the principal reason for home health would generate higher payment to account for these inherent challenges requiring additional resource utilization.

f. Medication Management, Teaching, and Assessment (MMTA)

Based on our analysis, the majority of 30-day periods of care in the HHGM would likely be classified under the MMTA clinical group. These 30-day periods would be characterized by codes that identify direct services related to the management and evaluation of the care plan, observation and assessment of the patient's condition, and training and/or education of a patient or family member that are not classified into one of the other clinical groups. The numerous and diverse conditions found in home health, and their associated medications and interventions, influence the principal diagnosis that would classify a 30-day period as under the MMTA clinical group.

Research on home health patient characteristics, home health nursing interventions, and outcomes of care show that there are four broad categories of interventions most frequently provided in the home:

(1) Health teaching, guidance and counseling;

(2) Treatments and procedures;

(3) Case management; and,

(4) Surveillance

Of these interventions, surveillance is the most frequently occurring intervention, closely followed by health teaching, guidance and counseling. Specific patient problems most frequently identified in the home health setting are related to medication regimens, especially with polypharmacy, and health-related behaviors. The majority of home health care patients routinely take more than five prescription drugs, and many likely deviate from their prescribed medication regimen. This increases the potential for medication errors or adverse effects in home health, highlighting the substantial need for education and medication management regardless of whether the patient needs wound care, rehabilitation, or complex nursing interventions.

Additionally, patients with comorbidities tend to be high users of home health, making education and assessment of disease diagnosis, medication interactions, lifestyle changes, and avoidance of adverse events a considerable portion of home health care. In an elderly patient population, the number of chronic conditions increases with age. Medications used to treat or prevent blood clots (anticoagulants), diabetes (insulin), and pain (opioid analgesics) are some of the most commonly implicated drugs in emergency room visits and emergent hospitalizations for adverse drug events in older adults. These adverse events can potentially be reduced by improving dosing and monitoring of these drugs in high risk populations and settings like older adults in home health programs. Anticoagulants are challenging to manage in home health settings and have been identified as targets for improvements in monitoring and care coordination by HHS. Also, as the number of medications being taken increases, so does the risk of adverse drug reactions, and the risk of drug reaction related emergency room visits and hospital admissions, especially in patients who are in poor health. Elderly patients are especially at risk for adverse drug reactions as the organs that metabolize drugs have reduced functional ability which can lead to increased toxicity. Similarly, roughly 31 percent of younger Medicare beneficiaries with disabilities report having five or more chronic conditions. Polypharmacy can lead to reduced compliance with medication regimens, thus putting the patient at risk for adverse events resulting from poorly managed conditions. In the home healthcare setting, management of polypharmacy is a primary focus of nursing interventions. These interventions include assessment of the patient's chronic conditions and medications used to treat those conditions; assessment of the patient's understanding of and compliance with his or her medication regimen; and teaching and reinforcing treatment and medication regimens. The medication review by the home health nurse can help reduce duplicate medications, medications that are contraindicated for older adults, and provide ways to ensure patients are being appropriately monitored and understand why they are taking the medications as well as how to take them correctly.

Other studies show that primary functions of home health care skilled nursing interventions include providing disease-specific and general health information; helping patients to practice and refine disease management skills; assessing efficacy of treatment; and, advocating for any needed changes to established treatment and drug regimens. The interventions encompassed under the MMTA clinical group are shown extensively in research literature to be the most prevalent services provided by home health clinicians. Analysis of home health episodes for the HHGM suggests that the MMTA services would be the most frequent home health service being provided to Medicare home health beneficiaries.

We believe that the proposed clinical groupings add a needed level of clarity in identifying and meeting the needs of home health patients; particularly the patient populations addressed in the Home Health Study Report to Congress as outlined in section II.D. of this proposed rule. Recognizing that all 30-day periods of home health care cannot be defined by the principal diagnosis alone, the clinical groupings would only be one step in the case-mix adjustment under the HHGM. We invite comments on the proposed clinical groups, which are designed to capture the most common types of care that HHAs provide.

7. Functional Levels and Corresponding OASIS Items

Research has shown a relationship exists between functional status, rates of hospital readmission, and the overall costs of health care services. Functional status is defined in a number of ways, but generally, functional status reflects an individual's ability to carry out activities of daily living (ADLs) and to participate in various life situations and in society. The assessment of functional status is often called “the sixth vital sign”, which reflects its clinical relevance in the plan of care. CMS requires the collection of data on functional status in home health through a standardized assessment instrument: The Outcome and Assessment Information Set (OASIS). Under the current HH PPS, functional status is assessed through the following OASIS items:

• M1810: Dressing Upper Body.
• M1820: Dressing Lower Body.
• M1830: Bathing.
• M1840: Toileting.
• M1850: Transferring.
• M1860: Ambulation/Locomotion.

For each of these OASIS items, the clinician or therapist conducting the assessment selects a numbered checkbox that best describes the patient's functional status in terms of ability to perform certain tasks. These numbered checkboxes typically range from zero, meaning independent with the task or no functional deficits, to higher numbers, meaning decreasing independence and/or increasing deficits. Responses to these OASIS items result in “points” to calculate an overall functional score which conveys the functional status of the patient. This means that patients with a higher functional score (that is, reduced functional status) have, on average, higher resource use compared to patients with a lower functional score (that is, higher functional status). As such, the functional status of the patient is a useful case-mix adjuster. Including functional status in the case-mix adjustment methodology allows for higher payment for those patients with higher service needs. As functional status is commonly used for risk adjustment in various payment systems, including in the current HH PPS, the proposed HHGM would also adjust payments to account for differences in resource use associated with functional status.

During the development of the HHGM, each OASIS-C item was evaluated using clinical review and analytical methods. Because the current case-mix adjustment methodology already utilizes OASIS items associated with functional status to adjust the home health payment, utilizing these OASIS items for inclusion in the HHGM was a primary focus. All OASIS items, including items not used in the current case-mix adjustment methodology, were evaluated for potential inclusion in the HHGM. OASIS items were eliminated for inclusion based on statistical factors (for example, the relationship of the item with resource use), clinical factors (for example, clinical appropriateness of using the item for payment purposes) and incentive factors (for example, potential for unintended consequences such as overutilization solely for increased reimbursement).

We presented our analysis of the OASIS items to a clinical workgroup that included physicians, nurses, and therapists with substantial home health clinical expertise, to obtain input regarding which OASIS items to include in the HHGM. Based on the clinical workgroup feedback and additional analyses by the research team, the following decisions were made regarding the narrowed list of OASIS items being considered for a functional status payment adjustment under the HHGM:

• M066, M0110: Age, Episode timing—Both age and episode timing were determined to be appropriate for the HHGM, but both items can be accurately obtained directly from the home health claims data, rather than the OASIS. As such, responses on these OASIS items would not be used for this functional status adjustment under the HHGM.
• M1018, M1030: Selected prior conditions and types of therapies a patient receives—These OASIS items would not be used for functional status adjustment in the HHGM because the clinical groups, specifically Complex Nursing Interventions, (described in section III.E.6. of this proposed rule) account for most of the conditions described in these OASIS items (for example, IV therapy, TPN) so using these OASIS items would be duplicative.
• M1200: Vision—The clinical workgroup believed this OASIS item to be clinically significant. However, while this item is used in the current HH PPS, there are no longer “points” associated with this item for the clinical domain because there is no additional resource use related to this item beyond the average across all periods of care. Additionally, analysis of this vision impairment OASIS item showed decreased resource use in the HHGM and; therefore, was determined to have a counterintuitive relationship. As a result, this OASIS item would not be used for functional status adjustment in the HHGM. Analysis of this item is found in the “Overview of the Home Health Groupings Model” technical report found on the HHA Center Web page.

• M1220, M1230: Understanding of verbal content, speech and oral—These items were determined to be subjective in nature and may not provide information that is an accurate reflection of the patient's cognitive status. As with other OASIS items in this analysis, these items showed that there was decreased resource costs associated with worsening status. As a result, these OASIS items would not be used for functional status adjustment in the HHGM.
• M1242: Pain—While this item is used in the current HH PPS, this is shown to have only a minimal relationship with resource use in the current payment model. Although the clinical workgroup believed this item to be clinically significant, CMS clinicians agreed this one item alone may not be robust enough to fully capture the pain presentation of the patient and its impact on resource utilization. Therefore, this OASIS item would not be used for functional status adjustment in the HHGM.
• M1302, M1308, M1320, M1322, M1324, M1332, M1334, and M1340: Ulcers and wounds—These OASIS items would not be used for functional status adjustment in the HHGM because the Wound clinical group (described in section III.E.6.of this proposed rule) already adjusts the period payment for these conditions and using these OASIS items would be duplicative.
• M1400: Shortness of breath—Although the clinical workgroup believed this item to be clinically significant, this OASIS item would not be used for functional status adjustment in the HHGM because the analysis showed decreased resource costs with worsening dyspnea which appears to be clinically counterintuitive.

• M1700—M1750: Cognitive items—These items were initially determined to be clinically appropriate for inclusion in the HHGM but were later removed due to analysis that showed a counterintuitive relationship, meaning costs decreased as cognitive status worsened. This negative relationship with resource use was consistent with most of the OASIS cognitive items. This analysis is discussed more in depth in this section below and the full analysis of all of the cognitive items is found in the technical report.
• M1800—M1890: Functional items—These OASIS items include both ADLs and Instrumental Activities of Daily Living (IADLs). ADLs are routine activities that people tend to do every day without needing assistance. There are six basic ADLs: Eating, bathing, dressing, toileting, transferring (walking) and continence. IADLs are activities related to independent living and include preparing meals, managing money, shopping for groceries or personal items, performing light or heavy housework, doing laundry, and using a telephone. While most of these items were determined to be clinically appropriate for inclusion in the HHGM, M1870-M1890 (IADLs) would not be used for functional status adjustment in the HHGM due to responses having a negative relationship with resource use (for example, worsening status in performing IADLs was associated with decreased resource use).
• M2030: Management of injectable medications—This OASIS item would not be used for functional status adjustment in the HHGM because most of the responses associated with this item reflected less resource use when the patient increasingly had issues with preparing and taking injectable medications. We believe that clinically counterintuitive relationships resulting from responses to OASIS items, where the expectation would be to see increased resource costs associated with decreased function or ability, should not be included in the case mix adjustment.

In addition to the OASIS items listed above, the clinical workgroup also discussed M2100 (types and sources of assistance-specifically non-HHA caregiver assistance). Workgroup members agreed that the availability of non-agency caregiver assistance can be an important determinant of home health care needs. Caregiver availability and assistance was a focus in the Report to Congress “Medicare Home Health Study: An Investigation on Access to Care and Payment for Vulnerable Patient Populations”. Vulnerable patient populations examined in this study included those patients with minimal or no caregiver support. Results from this study revealed that HHAs and physicians stated that family or caregiver issues are an important contributing factor in the inability to admit or place patients in home health. However, the survey results suggest that much of the variation in access to Medicare home health services is associated with social and personal conditions, and therefore, CMS' ability to improve access for certain vulnerable patient populations through payment policy alone may be limited. OASIS-C item M2100 identifies the ability and willingness of the caregiver(s) (other than home health agency staff) to provide categories of assistance needed by the patient, including ADL/IADL assistance, medication administration, and management of equipment. This particular OASIS item is multi-faceted, meaning this items requires one of six responses for seven different types of caregiver assistance. Because the responses to this item generally are not based on direct observation by the clinician conducting the assessment, this presents a limitation for use in a case mix adjustment as the accuracy of the responses cannot be easily validated. Patients or caregivers may overestimate or underestimate their ability or willingness to assist in the patient's care. Analysis of the resource use associated with this item showed ambiguous results where the same response (“assistance needed, but no caregiver(s) available”) would be associated with increased resource costs for certain types of assistance but decreased resource costs for other types of assistance. We believe this is clinically counterintuitive as it would be expected that if a need for caregiver assistance exists but there are no available caregivers, then the result would be an increased need for resources for all of the types of caregiver assistance listed on this OASIS item. Analysis of OASIS-C item M2110, frequency of ADL/IADL assistance, which identifies the frequency of assistance provided by non-agency caregiver(s), also showed a counterintuitive and contradicting relationship with M2100. Therefore, these OASIS items would not be included as part of the functional status payment adjustment under the HHGM.

During the analysis of functional case mix adjustment under the HHGM, a review of the literature revealed growing evidence suggesting that cognitive dysfunction is an important risk factor in the development of functional disability and loss of independence. The research team analyzed the responses to the OASIS items associated with cognitive status, but found there was decreased resource use associated with worsening cognitive status. We decided to further evaluate OASIS cognitive items (M1700-1750) in addition to functional items (M1800-1860), as well as other possible OASIS items that may contribute to overall function status. The following OASIS items were determined to be indicators of cognitive and functional status that potentially could be used as case mix adjusters:

• M066: Age.
• M1032: Risk of Hospitalization.
• M1220: Understanding of Verbal Content.
• M1230: Speech and Oral (Verbal) Expression of Language.
• M1700: Cognitive functioning.
• M1710: Confusion indicator.
• M1720: Anxiety indicator.
• M1740: Cognitive, behavioral, and psychiatric symptoms.
• M1745: Frequency of disruptive behavior symptoms.
• M1750: Receipt of psychiatric nursing services.
• M1800: Grooming.
• M1810: Current ability to dress upper body safely.
• M1820: Current ability to dress lower body safely.
• M1830: Bathing.
• M1840: Toilet transferring.
• M1845: Toilet hygiene.
• M1850: Transferring.
• M1860: Ambulation/locomotion.

One difficulty in using certain OASIS items (for example, M1700) to examine relationships with resource use is that they are only questioned on the Start of Care and Resumption of Care assessments, and not on follow-up assessments. Therefore, for this analysis, as outlined in the technical report, we looked back for the most recent period in the same sequence of periods that was linked to a Start of Care or Resumption of Care assessment, and carried forward the information from that assessment to the subsequent periods of care linked to follow-up (recertification) assessments. Analysis of these items, including looking at interactions between certain items, continued to show decreased resource use associated with worsening severity. The research team believed that clinically counterintuitive relationships to resource use may have the unintended consequence of discouraging HHAs to provide the appropriate amount of care to the patients who are clinically complex and need home health services the most.

For several of the OASIS items listed above, particularly the functional items, worsening status is associated with higher resource use, indicating that these items may be useful as adjustors to construct case-mix weights for the HHGM. However, several responses within other individual OASIS items had very similar average resource use. Due to the lack of variation in resource use across certain responses and because certain responses were infrequently chosen, some responses were combined into larger response categories to better capture the relationship between worsening status and resource use. Responses on these OASIS items were combined using the following methodology:

• Responses that corresponded to a small number of periods were combined with responses that corresponded to a larger number of periods and;
• Responses that had similar average resource use were combined together.

The resulting combinations of responses for these OASIS items are found at Exhibit 7-2 in the HHGM technical report.

After making these combinations, the newly combined OASIS items and resource use were analyzed again to determine if those OASIS items could be used to help case-mix adjust periods within the HHGM. Results showed that decreasing functional status, increasing age, and increasing risk of hospitalization tended to be associated with higher resource use, while worsening cognitive status tended to be associated with lower resource use. The relationship between worsening cognitive status but lower resource use is counterintuitive to existing research regarding cognitive status and health care costs. To further explore the relationship between the functional and cognitive OASIS items and resource use, additional analyses were conducted where the coefficients (that is, resource costs) associated with the functional and cognitive items were converted into a table of points to calculate the functional score for home health periods of care. However, even after controlling for each OASIS variable (as well as other components of the HHGM), the general trends between the cognitive and functional items from the other analyses remained the same. That is, worsening cognitive status was generally associated with less resource use; worsening functional status was generally associated with increased resource use; increased risk of hospitalization was associated with increased resource use; and age was not associated with either increased or decreased resource use. The summary statistics of these analyses are found at Exhibit 7-3 of the technical report, “Overview of the Home Health Groupings Model”. Therefore, we decided not to include those OASIS items with these types of inverse relationships to resource costs as part of the adjustment to the HHGM period payment. However, given the research support and clinical input from home health clinicians, we will continue to analyze the inclusion of cognitive items into the HHGM case mix adjustment. The analyses of the complete list of all OASIS items analyzed can be found in the Appendix Exhibits A7-1 and A7-2 of the technical report mentioned above.

On the basis of input from the clinical workgroup and these analytic results, all cognitive items, functional items with a negative relationship with resource use, and age were removed and the model was re-estimated. Each OASIS item included in the final model has a positive relationship with resource use, meaning as functional status declines (as measured by a higher response category), periods have more resource use on average. Additionally, periods with a higher risk of hospitalization (meaning four or more items checked on M1033) are associated with higher resource use compared with periods with a lower risk of hospitalization. This indicates that these items could be used to help risk adjust a period's payment and help determine case-mix weights for the HHGM. As such, we are proposing that the following OASIS items be included as part of the functional payment adjustment under the proposed HHGM:

• M1800: Grooming.
• M1810: Current Ability to Dress Upper Body.
• M1820: Current Ability to Dress Lower Body.
• M1830: Bathing.
• M1840: Toilet Transferring.
• M1850: Transferring.
• M1860: Ambulation/Locomotion.
• M1032 (M1033 in OASIS-C1): Risk of Hospitalization.

While the original analyses of these OASIS functional items were conducted using CY 2013 data from the OASIS-C version (as presented in the technical report), the updated analyses for CY 2016 reported in Tables 36, 37, and 38 are based on data obtained from OASIS C-1. While the OASIS item number for “Risk of Hospitalization” changed from M1032 (in OASIS C) to M1033 (in OASIS C-1), the remaining OASIS items (and item numbers) used for this functional adjustment analysis are the same. As discussed earlier in this section, to facilitate the interpretation of this analysis of the functional items used to construct the case mix weights, the results of this analysis were converted into a table of points that can be used to calculate the functional score for a home health period. Table 36 shows the points for 2013 and 2016 for those items associated with increased resource use using a reduced set of OASIS C-1 items:

Similar to the current case-mix adjustment methodology, the points generated in Table 36 were then used to create a functional score for each home health period of care in the HHGM. That is, a home health period of care receives points based on each of the responses associated with the OASIS items listed above. The sum of all of these points results in a functional score which is used in the HHGM to group home health periods into a functional level. As part of the HHGM case-mix adjustment, we are proposing to assign points for each of the responses to the proposed OASIS functional items and to sum up the points to create a functional score for the period of care. Whereas the results presented in the technical report showed that the number of functional levels varied by clinical group, continued analysis ultimately established three functional levels for each of the clinical groups—low, medium and high, with approximately one third of home health periods from each of the clinical groups within each level. This means home health periods in the low level have responses for the above OASIS items that are associated with the lowest resource use on average. Home health periods in the high level have responses on the above OASIS items that are associated with the highest resource use on average. We are proposing to use the three functional levels of low, medium, and high, based on the CY 2016 data for each of the clinical groups. Table 37 shows the functional thresholds for each functional level by clinical group for CYs 2013 and 2016.

Table 38 shows the average resource use by clinical group and functional level for CY 2016:

Like the annual recalibration of the case-mix weights under the current HH PPS, we expect that annual recalibrations would also be made to the HHGM case-mix weights. If the HHGM is finalized, we will continue to analyze all of the components of the case-mix adjustment, including adjustment for functional status, and would make refinements as necessary to ensure that payment for home health periods are in alignment with costs. We invite comments on the proposed OASIS items and the associated points and thresholds used to group patients into three functional levels under the HHGM, as outlined above.

8. Comorbidity Adjustment

The HHGM groups home health periods based on the primary reason for home health care (principal diagnosis), functional level, admission source, and timing. To further account for differences in resource use based on patient characteristics in the development of the HHGM, we analyzed the presence of comorbidities as another factor that could impact resource utilization and costs. We conducted a comprehensive literature review examining published, peer-reviewed research regarding the relationship between comorbidity and resource use. This review also included findings on those conditions that impact health care resource utilization. Based on this review and findings, we propose a comorbidity adjustment to account for higher costs associated with comorbidities.

A comorbidity is most often defined as two or more coexisting medical conditions or disease processes that are in addition to an initial diagnosis. Typically, a comorbidity is a condition(s) in which there is no direct correlation in the treatment of the principal diagnosis, but the presence of that condition(s) may impact the home health plan of care in terms of resource utilization and costs. With aging, the presence of comorbidity increases markedly because the frequency of individual conditions arises with age. While the elderly are far more likely to have multiple comorbidities, comorbidities also are prevalent in Medicare beneficiaries under the age of 65 who have intellectual and physical disabilities. Research has repeatedly shown that comorbidity is associated with high health care utilization and expenditures. Additionally, comorbidity is tied to worse health outcomes and the need for more complex treatment and disease management, which in turn results in higher health care costs. Patients with comorbidities tend to be high users of home health visits and overall Medicare spending increases with the number of chronic conditions.

In the home health setting, information regarding the patient's health conditions for which home health services are provided are assessed and documented by skilled clinicians on the OASIS. These conditions would include secondary diagnoses in addition to the principal diagnosis supporting the need for home health services. As such, exploratory analyses for the HHGM determined that secondary diagnoses (that is, comorbidities) provide additional information that can predict resource use even after controlling for the period's clinical group. We examined multiple approaches for a comorbidity adjustment in the HHGM and the analyses on these approaches is found in the “Overview of the Home Health Groupings Model” technical report found on the HHA Center Web page. Based on the results of these analyses, we moved towards the development of a home health specific comorbidity list for the HHGM comorbidity adjustment.

For the analysis of a comorbidity adjustment in the HHGM, some diagnosis exclusions were made. Under the HHGM, certain reported principal diagnosis codes, including some ICD-10-CM “R-codes” (R00-R99) which identify symptoms and abnormal clinical findings, would be considered a “questionable encounter”, meaning these codes may be too vague to group the home health period, subject to sequencing or other ICD-10-CM coding conventions, not a Medicare-covered diagnosis, or a condition unlikely to require home health services. For these “questionable encounters”, more information was needed to assign the period to a clinical group. This meant, for analysis purposes only, we looked at the secondary diagnoses to assign the home health period to one of the six clinical groups. As such, those periods with a principal diagnosis that was determined to be a “questionable encounter” code were excluded from our comorbidity adjustment analysis. However, if the HHGM is finalized, we are proposing that claims submitted with principal reported diagnosis codes that are considered “questionable encounters” would be returned to the provider for more definitive coding. Once the claim is resubmitted without a principal diagnosis that is considered a “questionable encounter”, the home health period would be grouped into one of the six clinical groups. The secondary diagnoses on those resubmitted claims would then be eligible for the comorbidity adjustment.

Another exclusion from this comorbidity analysis included those secondary diagnoses that had the same three character ICD-10-CM code as the diagnosis used to assign a case to a particular clinical group (that is, musculoskeletal rehab, neuro/stroke rehab, wounds, behavioral health, complex nursing interventions, and MMTA). An additional exclusion was added that applied to diagnoses that identify an unspecified site or side (meaning the code is defined by laterality or site specificity). There are ICD-10-CM codes that are specific to site, laterality, and proximal versus distal parts of the body. For example, L89.004, Pressure ulcer of unspecified elbow, stage 4, can be coded to identify whether the pressure ulcer is on the left or right elbow. ICD-10 CM coding guidelines state to report diagnoses to the greatest level of specificity. The home health clinician should be able to identify the specific side or body part involved through either direct assessment or of a query of the certifying physician.

Finally, an exclusion was added for some secondary diagnoses that would not be considered a comorbidity if reported with certain Z codes. For example, if Z96.651, presence of right artificial knee joint, is reported as secondary, it would not be considered a comorbidity if Z47.1, aftercare following joint replacement surgery, was reported as the principal diagnosis. The secondary diagnosis in this scenario is not a comorbidity because this secondary diagnosis explains the reason for the aftercare. We are utilizing this approach to minimize the unintended consequence of providers reporting comorbidities that are duplicative of the principal diagnosis, or are a further description of the principal diagnosis, which could potentially overestimate the actual resources needed for a home health period and could result in inaccurate payment.

Using the research from the comprehensive literature review, we identified common chronic comorbid conditions frequently cited as drivers of increased health care resource utilization, including coronary artery disease, congestive heart failure, diabetes, COPD, asthma, chronic wounds, and depression. In addition to chronic comorbid conditions, other acute comorbid conditions have been shown to affect overall resource utilization as well. These conditions include pneumonia, Clostridium difficile (c-diff), and Methicillin-resistant Staphylococcus aureus (MRSA). After compiling a list of both acute and chronic comorbid diagnoses that could affect home health resource utilization, we conducted initial analyses looking at controlling for the presence of the individual diagnoses. However, these analyses showed some counterintuitive relationships with resource use, meaning the presence of certain comorbidities showed that there would be less resource use than if the comorbidity was not present.

Because the core of the HHGM is a clinical one, CMS clinicians utilized the principles of patient assessment by body systems and their associated diseases, conditions, and injuries as a way to examine potential clinically relevant relationships. Next, we combined those individual diagnoses into larger categories utilizing the body systems as a clinically intuitive way to consider what diagnoses potentially could impact the home health plan of care and resource utilization. When combining the individual diagnoses into larger comorbidity categories, the counterintuitive relationships decreased. These broad body system categories include conditions, diseases, and injuries that affect each of the individual body systems (for example, heart disease). Neoplasms and infectious diseases were given their own discrete categories because of their potential to affect more than one body system. The broad categories used to group comorbidities within the HHGM were further refined by grouping similar diagnoses within the broad categories into subcategories. The subcategories allowed for additional refinement of diagnoses to include as part of the home health specific list. Subcategories were distinguished primarily (but not exclusively) by the first three characters of the ICD-10-CM diagnosis code to represent related conditions within the same body system. For example, subcategory Heart 10 includes diagnoses associated with various cardiac arrhythmias. The home health specific comorbidity list includes 13 broad body system based categories and 116 total subcategories using ICD-10-CM diagnosis codes. The broad categories used to group comorbidities within the HHGM include the following:

• Heart Disease (11 subcategories).
• Respiratory Disease (9 subcategories).
• Circulatory Disease and Blood Disorders (12 subcategories).
• Cerebral Vascular Disease (4 subcategories).
• Gastrointestinal Disease (9 subcategories).
• Neurological Disease and Associated Conditions (11 subcategories).
• Endocrine Disease (6 subcategories).
• Neoplasm (24 subcategories).
• Genitourinary and Renal Disease (5 subcategories).
• Skin Disease (5 subcategories).
• Musculoskeletal Disease or Injury (5 subcategories).
• Behavioral Health (11 subcategories).
• Infectious Disease (4 subcategories).

The secondary diagnoses listed on the OASIS that are attributed to any one of the listed subcategories were used to identify whether a period fell into one or more comorbidity categories and subcategories.

For the purpose of evaluating these identified comorbidities for inclusion in the HHGM, we assigned the CY 2016 home health periods that reported a secondary diagnosis included on this home health specific list to a comorbidity subcategory and subsequently dropped any subcategories that were in less than 0.1 percent of periods. This was done because low volume leads to instability in our estimates of how resource use is related to the comorbidity. A regression model was used to determine the relationship between the remaining subcategories and resource use. After this analysis, we dropped comorbidity subcategories that were not statistically significant with regards to their relationship to resource use (a coefficient with a p-value greater than 0.05). After these exclusions, we kept the subcategories associated with increased resource use that was at least as high as the median resource use, as they indicated a direct relationship between the comorbidity subcategories and resource utilization. These remaining subcategories would receive a comorbidity adjustment. As such, there are 15 subcategories that meet the current criteria for the comorbidity adjustment in the HHGM. This is a decreased number of subcategories that were presented in the technical report where 29 subcategories met the criteria to qualify for the comorbidity adjustment. The comorbidity analysis presented in the technical report was based on CY 2013 data and used ICD-9-CM diagnosis codes. There are several potential reasons for this decrease including that the analysis exclusions for the 2016 analysis were slightly different than were used in the technical report. Another potential reason for the decrease in subcategories may be due to diagnosis exclusions based on changes from ICD-9-CM to ICD-10-CM with regards to specificity. Some of this decrease could be related to the changes in case-mix weights from 2013 to 2016 where secondary conditions that received clinical points in 2013 may not have had any associated points in 2016 and hence, there would be no incentive to report those conditions. The analysis on the CY 2013 and CY 2016 data, including all of the diagnoses and their assigned subcategories is posted on the HHA Center Web page. The 15 subcategories included in the comorbidity adjustment in the HHGM are as follows:

• Heart Disease 1: Includes hypertensive heart disease.
• Cerebral Vascular Disease 4: Includes sequelae of cerebrovascular disease.
• Circulatory Disease and Blood Disorders 9: Includes venous embolisms and thrombosis.
• Circulatory Disease and Blood Disorders 10: Includes varicose veins of lower extremities with ulcers and inflammation, and esophageal varices.
• Circulatory Disease and Blood Disorders 11: Includes lymphedema.
• Endocrine Disease 2: Includes diabetes with complications due to an underlying condition.
• Neoplasm 18: Includes secondary malignant neoplasms.
• Neurological Disease and Associated Conditions 5: Includes secondary parkinsonism.
• Neurological Disease and Associated Conditions 7: Includes encephalitis, myelitis, encephalomyelitis, and hemiplegia, paraplegia, and quadriplegia.
• Neurological Disease and Associated Conditions 10: Includes diabetes with neurological complications.
• Respiratory Disease 7: Includes pneumonia, pneumonitis, and pulmonary edema.
• Skin Disease 1: Includes cutaneous abscesses, and cellulitis.
• Skin Disease 2: Includes stage one pressure ulcers.
• Skin Disease 3: Includes atherosclerosis with gangrene.
• Skin Disease 4: Includes unstageable and stages two through four pressure ulcers.

We propose that if a period had at least one secondary diagnosis reported on the home health claim that fell into one of the 15 subcategories, that period would receive a comorbidity adjustment to account for higher costs associated with the comorbidity. The comorbidity adjustment amount would be the same across all of the subcategories. A period would receive only one comorbidity adjustment regardless of the number of secondary diagnoses reported on the home health claim that fell into one of the 15 subcategories. Table 39 shows information on resource use for periods with and without the comorbidity adjustment.

The HHGM payment adjustment for comorbidities is predicated on the presence of one of the identified diagnoses within the subcategories associated with increased resource use at or above the median. If there is no reported diagnosis that meets the comorbidity adjustment criteria, the period would not qualify for the payment adjustment. We consider this comorbidity adjustment component of the proposed HHGM to be fluid, where OASIS-reported secondary diagnoses may be removed from, or added to the home health specific comorbidity list dependent upon the relationship between the comorbidity and resource costs. If the HHGM is finalized and implemented, we anticipate there may be behavioral shifts in secondary diagnosis reporting and the proposed comorbidity list and its associated subcategories may change to capture resource utilization associated with these or other conditions. We invite comments on the proposed comorbidity diagnoses, including additions or subtractions to the proposed home health specific list, and this comorbidity adjustment approach under the HHGM.

9. Change in the Low-Utilization Payment Adjustment (LUPA) Threshold

An episode with four or fewer visits is paid the national per visit amount by discipline, adjusted by the appropriate wage index based on the site of service of the beneficiary, instead of the full episode amount. Such payment adjustments are called Low Utilization Payment Adjustments (LUPAs). While the proposed HHGM system would still include LUPA payments, we are proposing that the approach to calculating the LUPA thresholds would change in the HHGM because of the proposed change in the unit of payment to 30-day periods from 60-day episodes. Whereas LUPAS are paid for all episodes consisting of four or fewer visits under the current payment system, in order to receive full episode amount under the HHGM (rather than receive a LUPA where the episode would be paid the national per visit amount by discipline) we propose to vary the LUPA threshold for a 30-day period under the HHGM depending on the HHGM payment group to which it is assigned. The 30-day periods have substantially more instances of four or fewer visits than 60-day episodes. To create LUPA thresholds, 30-day periods (including those that were LUPAs in the current payment system) were grouped into the 144 different HHGM payment groups. For each payment group, we propose to set the LUPA threshold at the 10th percentile value of visits or 2 visits, whichever is higher. In the current payment system approximately 8 percent of episodes are LUPAs. Under the HHGM, we propose the 10th percentile value of visits or 2 visits, whichever is higher, to target approximately the same percentage of LUPAs (approximately 7 percent of 30-day periods would be LUPAs (assuming no behavior change)).

For example, for 30-day periods of care in the payment group corresponding to “MMTA- Functional Level Medium—Early Timing—Institutional Admission—No Comorbidity Adjustment”, the threshold is four visits. If 30-day periods assigned to that particular payment group had three or fewer visits they would be paid using the national per-visit rates in section III.C.3 of this proposed rule instead of the case-mix adjusted 30-day payment amount. We propose that the LUPA thresholds for each HHGM payment group would be re-evaluated every year based on the most current, complete utilization data available. The LUPA thresholds, based on the most current utilization data available (CY 2016 data as of March 17, 2017), for each corresponding HIPPS code, are listed in Table 40. We would propose updated LUPA thresholds using the most current, complete utilization data available at the time of rulemaking.

We invite public comments on the LUPA threshold methodology proposed for the HHGM and the associated regulations text changes in section VIII. of this proposed rule.

10. HH PPS Case-Mix Weights Under the HHGM

Section 1895(b)(4)(B) of the Act requires the Secretary to establish appropriate case mix adjustment factors for home health services in a manner that explains a significant amount of the variation in cost among different units of services. We are proposing the HHGM case-mix adjustment methodology, which sorts 30-day periods of care into different payment groups based on five categories (admission source, timing, clinical group, functional level, and comorbidity group), for 30-day periods of care that begin on or after January 1, 2019. In combination, this would yield a total of 144 HHGM payment groups, which we would still refer to as Home Health Resource Groups (HHRGs) under the HHGM. To generate HHGM case-mix weights, we utilized a data file based on home health episodes of care, as reported in Medicare home health claims, as well as OASIS assessment data. The claims data provide episode-level data, as well as visit-level data. The claims also provide data on whether NRS was provided during the episode and the total charges for NRS. We determined the case-mix weight for each of the different HHGM payment groups by regressing resource use on a series of indicator variables for each of the five categories listed above using a fixed effects model. The regression measures resource use with the proposed Cost per Minute (CPS) + NRS approach outlined in section III.E.2 of this proposed rule.

To normalize the results from the fixed effects regression model, we divided the predicted resource use for each 30-day period by the overall average resource use for all 30-day periods used to estimate the model to calculate the case mix weight of all 30-day periods within a particular payment group, where each payment group is defined as the unique combination of the subgroups within the five HHGM categories (admission source, timing of the episode, clinical grouping, functional level, and comorbidity adjustment). The case-mix weight is then used to adjust the 30-day payment rate to determine each 30-day period payment. Table 41 shows the coefficients of the payment regression used to generate the weights, and the coefficients divided by average resource use. Information can be found in section III.E.6 of this proposed rule for the clinical groups, section III.E.7 of this proposed rule for the functional levels, section III.E.5 of this proposed rule for admission source, section III.E.4 of this proposed rule for episode timing, and section III.E.8 of this proposed rule for the comorbidity adjustment.

Table 42 presents the case-mix weight for each HHRG in the regression model (from Table 46's coefficients). LUPA episodes, outlier episodes, and episodes with PEP adjustments were excluded. These are the case-mix weights for the HHGM based on the most current, complete data available (CY 2016 data as of March 17, 2017). We would propose updated case-mix weights using the latest CY 2017 data in the CY 2019 HH PPS proposed rule. LUPA information can be found in section III.E.9 of this proposed rule. Weights are determined by first calculating the predicted resource use for episodes with a particular combination of admission source, episode timing, clinical grouping, functional level, and comorbidity adjustment. This combination specific calculation is then divided by the average resource use of all the episodes that were used to estimate, which is $1,585.48. The resulting ratio represents the case-mix weight for that particular combination of a HHRG payment group. The adjusted R-squared value for this model is 0.2704. The adjusted R-squared value provides a measure of how well observed outcomes are replicated by the model, based on the proportion of total variation of outcomes explained by the model. In this instance, the fixed effects regression model used to generate the case-mix weight under the HHGM predicts about 27 percent of the variation in resource use in a given 30-day period of home health care.

As noted above, there are 144 different HHRG payment groups under the HHGM. There are 9 HHRG payment groups that represent roughly 50.5 percent of the total episodes. There are 33 HHRG payment groups that represent roughly 1.0 percent of total episodes. The HHRG payment group with the smallest weight has a weight of 0.5034 (community, late, behavioral health, low functional level, with no comorbidity adjustment). The HHRG payment group with the largest weight has a weight of 1.9533 (institutional admission, early, wound, high functional level, with comorbidity adjustment).

We invite comments on the proposed case-mix weight methodology under the HHGM.

11. Low-Utilization Payment Adjustment (LUPA) Add-On Payments and Partial Payment Adjustments Under the HHGM

LUPA episodes that occur as the only episode or as an initial episode in a sequence of adjacent episodes are adjusted by applying an additional amount to the LUPA payment before adjusting for area wage differences. Under the HHGM, we propose the LUPA add-on factors will remain the same as the current payment system, described in section III.C.3. of this proposed rule. We propose to multiply the per-visit payment amount for the first SN, PT, or SLP visit in LUPA episodes that occur as the only episode or an initial episode in a sequence of adjacent episodes by the appropriate factor (1.8451 for SN, 1.6700 for PT, and 1.6266 for SLP) to determine the LUPA add-on payment amount. For example, for LUPA episodes that occur as the only episode or an initial episode in a sequence of adjacent episodes in CY 2019, if the first skilled visit is SN, the payment for that visit would be the CY 2019 per-visit rate for SN, multiplied by 1.8451, subject to area wage adjustment.

The current partial episode payment (PEP) adjustment is a proportion of the episode payment and is based on the span of days including the start-of-care date or first billable service date through and including the last billable service date under the original plan of care before the intervening event in a home health beneficiary's care defined as:

• A beneficiary elected transfer, or
• A discharge and return to home health that would warrant, for purposes of payment, a new OASIS assessment, physician certification of eligibility, and a new plan of care.

For 30-day periods of care, we propose the process for partial payment adjustments would remain the same as the existing policies pertaining to partial episode payments. When a new 30-day period begins due to the intervening event of the beneficiary elected transfer or there was a discharge and return to home health during the 30-day period, we propose the original 30-day period would be proportionally adjusted to reflect the length of time the beneficiary remained under the agency's care prior to the intervening event. The proportional payment is the partial payment adjustment. The partial payment adjustment is calculated by using the span of days (first billable service date through and including the last billable service date) under the original plan of care as a proportion of 30. The proportion is multiplied by the original case-mix and wage index to produce the 30-day payment.

12. Payments for High-Cost Outliers Under the HHGM

As described in section III.D. of this proposed rule, section 1895(b)(5) of the Act allows for the provision of an addition or adjustment to the home health payment amount in the case of outliers because of unusual variations in the type or amount of medically necessary care. The history of and current methodology for payment of high-cost outliers under the HH PPS is described in detail in section III.D. of this proposed rule. We are proposing to maintain the current methodology for payment of high-cost outliers upon implementation of the HHGM in CY 2019 and we would calculate payment for high-cost outliers on 30-day periods of care.

Simulating payments using preliminary CY 2016 claims data and the CY 2018 payment rates, we estimate that outlier payments under the proposed HHGM with 30-day periods of care would comprise approximately 4.50 percent of total HH PPS payments in CY 2018. Given the statutory requirement to target up to, but no more than, 2.5 percent of total payments as outlier payments, we currently estimate that the FDL ratio under the HHGM would need to change from 0.55 to 0.93. However, given the proposed implementation of the HHGM for 30-day periods of care beginning on or after January 1, 2019, we will update our estimate of outlier payments as a percent of total HH PPS payments using the most current and complete utilization data available at the time of CY 2019 rate-setting. We would propose a change in the FDL ratio for CY 2019, if needed.

We invite public comments on maintaining the current outlier payment methodology outlined in section III.D. of this proposed rule for the proposed HHGM and the associated changes in the regulations text as described in section III.E.13 of this proposed rule.

13. Conforming Regulations Text Revisions for the Implementation of the HHGM in CY 2019

We are proposing to make a number of revisions to the regulations to implement the HHGM for periods beginning on or after January 1, 2019, as outlined in sections III.E.1. through III.E.12. of this proposed rule. We propose to make conforming changes in § 409.43 and part 484 subpart E to revise the unit of service from a 60-day episode to a 30-day period. In addition, we are proposing to restructure § 484.205. These revisions would be effective on January 1, 2019. We are not proposing any revisions to the regulations for CY 2018. These revisions and others are discussed below. Specifically, we propose to:

• Revise § 409.43, which outlines plan of care requirements. We propose to revise several paragraphs to phase out the unit of service from a 60-day episode for episodes beginning on or before December 31, 2018, and to implement a 30-day period as the new unit of service for periods beginning on or after January 1, 2019 under the HHGM.
• Revise the definitions of rural area and urban area in § 484.202 to remove “with respect to home health episodes ending on or after January 1, 2006” from each definition, as this verbiage is no longer necessary.
• Restructure § 484.205 to provide more logical organization. Specifically, we propose to add paragraphs to paragraph (b) to define the unit of payment. We propose to move language which addresses the requirement for OASIS submission from § 484.210 and insert it into § 484.205 as new paragraph (c). We also propose to add paragraph (f) to discuss split percentage payments under the current model and the proposed HHGM. In addition, we propose to revise § 484.205 to remove references to “60-day episode” and to refer more generally to the “national, standardized prospective payment”. While we are proposing to revise § 484.205 to account for the change in the unit of payment under the HH PPS for CY 2019, we are not proposing to change the requirements or policies relating to durable medical equipment or furnishing negative pressure wound therapy using a disposable device.
• Remove § 484.210 which discusses data used for the calculation of the national prospective 60-day episode payment as we believe that this information is incorporated in other sections of part 484 subpart E, such as § 484.205(c), § 484.215(a) and (b), § 484.220 and § 484.215.
• Revise the section heading of § 484.215 from “Initial establishment of the calculation of the national 60-day episode payment” to “Initial establishment of the calculation of the national, standardized prospective 60-day episode payment and 30-day payment rates.” Also, we propose to add paragraph (f) to this section to describe how the national, standardized prospective 60-day episode payment rate is converted into a national, standardized prospective 30-day period payment and when it applies.
• Revise the section heading of § 484.220 from “Calculation of the adjusted national prospective 60-day episode payment rate for case-mix and area wage levels” to “Calculation of the case-mix and wage area adjusted prospective payment rates.” We propose to remove the reference to “national 60-day episode payment rate” and replace it with “national, standardized prospective payment”.
• Revise the section heading in § 484.225 from “Annual update of the unadjusted national prospective 60-day episode payment rate” to “Annual update of the unadjusted national, standardized prospective 60-day episode and 30-day payment rates”. Also, we propose to revise § 484.225 to remove references to “60-day episode” and to refer more generally to the “national, standardized prospective payment”. In addition, we propose to add paragraph (d) to describe the annual update for CY 2019.
• Revise the section heading of § 484.230 from “Methodology used for the calculation of low-utilization payment adjustment” to “Low utilization payment adjustment”. Also, we propose to designate the current text to paragraph (a) and insert language such that proposed paragraph (a) applies to episodes beginning on or before December 31, 2018, using the current payment system. We propose to add paragraph (b) to describe how low utilization payment adjustments are determined for periods beginning on or after January 1, 2019, using the proposed HHGM.
• Revise the section heading of § 484.235 from “Methodology used for the calculation of partial episode payment adjustments” to “Partial payment adjustments”. We propose to remove paragraphs (a), (b), and (c). We propose to remove paragraphs (1), (2), and (3) which describe partial payment adjustments from paragraph (d) in § 484.205 and incorporate them into § 484.235. We propose to add paragraph (a) to describe partial payment adjustments under the current system, that is, for episodes beginning on or before December 31, 2018, and paragraph (b) to describe partial payment adjustments under the proposed HHGM, that is, for periods beginning on or after January 1, 2019.
• Revise the section heading for § 484.240 from “Methodology used for the calculation of the outlier payment” to “Outlier payments.” In addition, we propose to remove language at paragraph (b) and append it to paragraph (a). We propose to add language to proposed revised paragraph (a) such that paragraph (a) will apply to payments under the current system, that is, for episodes beginning on or before December 31, 2018. We propose to revise paragraph (b) to describe payments under the proposed HHGM, that is, for periods beginning on or after January 1, 2019. In paragraph (c), we propose to replace the “estimated” cost with “imputed” cost. Lastly, we propose to revise paragraph (d) to reflect the per-15 minute unit approach to imputing the cost for each claim.

We are soliciting comments on the proposed HHGM as outlined in sections III.E.1. through III.E.12. and the associated regulations text changes described above and in the regulations text of this proposed rule.

IV. Proposed Provisions of the Home Health Value-Based Purchasing (HHVBP) Model

A. Background

As authorized by section 1115A of the Act and finalized in the CY 2016 HH PPS final rule (80 FR 68624), we began testing the HHVBP Model on January 1, 2016. The HHVBP Model has an overall purpose of improving the quality and delivery of home health care services to Medicare beneficiaries. The specific goals of the Model are to: (1) Provide incentives for better quality care with greater efficiency; (2) study new potential quality and efficiency measures for appropriateness in the home health setting; and (3) enhance the current public reporting process.

Using the randomized selection methodology finalized in the CY 2016 HH PPS final rule, nine states were selected for inclusion in the HHVBP Model, representing each geographic area across the nation. All Medicare-certified HHAs providing services in Arizona, Florida, Iowa, Maryland, Massachusetts, Nebraska, North Carolina, Tennessee, and Washington (competing HHAs) are required to compete in the Model. Requiring all Medicare-certified HHAs providing services in the selected states to participate in the Model ensures that: (1) there is no selection bias; (2) participating HHAs are representative of HHAs nationally; and, (3) there is sufficient participation to generate meaningful results.

As finalized in the CY 2016 HH PPS final rule, the HHVBP Model will utilize the waiver authority under section 1115A(d)(1) of the Act to adjust Medicare payment rates under section 1895(b) of the Act beginning in CY 2018 based on performance on applicable measures. Payment adjustments will be increased incrementally over the course of the HHVBP Model in the following manner: (1) A maximum payment adjustment of 3 percent (upward or downward) in CY 2018; (2) a maximum payment adjustment of 5 percent (upward or downward) in CY 2019; (3) a maximum payment adjustment of 6 percent (upward or downward) in CY 2020; (4) a maximum payment adjustment of 7 percent (upward or downward) in CY 2021; and (5) a maximum payment adjustment of 8 percent (upward or downward) in CY 2022. Payment adjustments will be based on each HHA's Total Performance Score (TPS) in a given performance year (PY) on (1) a set of measures already reported via OASIS and HHCAHPS for all patients serviced by the HHA and select claims data elements, and (2) three New Measures where points are achieved for reporting data.

As finalized in the CY 2017 HH PPS final rule (81 FR 76741 through 76752), in addition to providing an update on the progress towards developing public reporting of performance under the HHVBP Model, we finalized the following changes related to the HHVBP Model:

• Calculating benchmarks and achievement thresholds at the state level rather than the level of the size-cohort and revising the definition for benchmark to state that benchmark refers to the mean of the top decile of Medicare-certified HHA performance on the specified quality measure during the baseline period, calculated for each state;
• Requiring a minimum of eight HHAs in a size-cohort;
• Increasing the timeframe for submitting New Measure data from seven calendar days to 15 calendar days following the end of each reporting period to account for weekends and holidays;
• Removing four measures (Care Management: Types and Sources of Assistance, Prior Functioning Activities of Daily Living (ADL)/Instrumental ADL (IADL), Influenza Vaccine Data Collection Period, and Reason Pneumococcal Vaccine Not Received) from the set of applicable measures;
• Adjusting the reporting period and submission date for the Influenza Vaccination Coverage for Home Health Personnel measure from a quarterly submission to an annual submission; and
• Allowing for an appeals process that includes the recalculation process finalized in the CY 2016 HH PPS final rule (80 FR 68688 through 68689), as modified, and adds a reconsideration process.

B. Quality Measures

1. Proposed Adjustment to the Minimum Number of Completed Home Health Care Consumer Assessment of Healthcare Providers and System (HHCAHPS) Surveys

The HHCAHPS survey presents home health patients with a set of standardized questions about their home health care providers and about the quality of their home health care. The survey is designed to measure the experiences of people receiving home health care from Medicare-certified home health care agencies and meet the following three broad goals to: (1) Produce comparable data on the patient's perspective that allows objective and meaningful comparisons between home health agencies on domains that are important to consumers; (2) create incentives through public reporting of survey results for agencies to improve their quality of care; and (3) enhance public accountability in health care by increasing the transparency of the quality of care provided in return for public investment through public reporting.

As finalized in the CY 2016 HH PPS final rule (80 FR 68685 through 68686), if a HHA does not have a minimum of 20 episodes of care during a performance year to generate a performance score on at least five measures, that HHA would not be included in the Linear Exchange Function (LEF) and would not have a payment adjustment percentage calculated. The LEF is used to translate an HHA's Total Performance Score (TPS) into a percentage of the value-based payment adjustment earned by each HHA under the HHVBP Model. For the HHCAHPS measures, a minimum of 20 HHCAHPS completed surveys would be necessary in order for scores to be generated for the HHCAHPS quality measures that can be included in the calculation of the TPS.

We believe, however, that using a minimum of 40 completed HHCAHPS surveys, rather than a minimum of 20 completed HHCAHPS surveys, would better align the Model with HHCAHPS policy for the Patient Survey Star Ratings on Home Health Compare. The decision to use a minimum of 40 completed surveys for these star ratings was a result of balancing two competing goals. One goal was to provide star ratings that were meaningful and minimized random variations. This goal was best served by calculating star ratings for large numbers of cases by having a larger minimum of completed HHCAHPS surveys (for example, 50 or 100 completed HHCAHPS surveys). At the same time, we also wanted to be able to provide star ratings for as many HHAs as possible. This goal was best served by using a lower minimum of completed HHCAHPS surveys (for example, 20 completed HHCAHPS surveys). We chose to balance these opposing and necessary goals by using 40 completed HHCAHPS surveys for the Patient Survey Star Ratings. Because we believe that aligning the Patient Survey Star Ratings system and the HHVBP model provides uniformity, consistency, and standard transformability for different healthcare platforms, we therefore propose using a minimum of 40 instead of 20 completed HHCAHPS surveys under the HHVBP.

We note that we received a comment in response to the CY 2016 HH PPS proposed rule in support of using a higher minimum threshold for HHCAHPS completed surveys for the Patient Survey Star Ratings if the data are going to be used in HHVBP or any other quality assessment program (80 FR 68709). We also note that we received public comment in response to the CY 2017 HH PPS proposed rule in support of using a higher minimum threshold for HHCAHPS completed surveys in the HHVBP Model, including a recommendation to use a minimum of 100 HHCAHPS rather than a sample size of 20 surveys (81 FR 76747). We believe that proposing a minimum of 40 completed HHCAHPS surveys for the Model would be more appropriate than the higher minimums previously recommended by some commenters because it represents a balance between providing meaningful data and having sufficient numbers of HHAs with performance scores for at least 5 measures in the cohorts. Moreover, as we noted, it aligns with the Patient Survey Star Ratings on Home Health Compare.

To understand the possible impact of our proposal to use a minimum of 40 HHCAHPS completed surveys, we note that HHAs may refer to the Interim Performance Reports (IPRs) issued in October 2016, January 2017 and April 2017, which analyzed 40 or more completed HHCAHPS surveys across both small and large cohorts in determining each HHA's HHCAHPS quality measure scores. As a point of comparison to the minimum of 40 HHCAHPS completed surveys, we note that these IPRs will be reissued using 20 or more completed HHCAHPS surveys and include quality measure scores, for these same time periods, calculated with HHAs that qualify for the LEF by having sufficient data for at least five measures. HHAs will have the opportunity to submit a request for recalculation of the revised interim performance scores.

HHAs have an opportunity to evaluate these IPRs in light of our proposal to change to a minimum of 40 HHCAHPS completed surveys, as well as seek clarification on the difference in their reports. The participating HHAs will receive concurrent IPRs in July 2017 and concurrent Annual Total Performance Score and Payment Adjustment Reports, which we plan to make available in the last week of August 2017. The concurrent reports will show one report with HHCAHPS quality measure scores calculated based on a minimum of 40 completed surveys and one report with HHCAHPS quality measure scores calculated based on a minimum of 20 completed surveys. Because this proposed rule will not be finalized before the timeline for submission of recalculation and reconsideration requests, HHAs will have the opportunity to submit recalculation requests for the interim performance scores based on both a minimum of 40 and 20 completed surveys, and recalculation and reconsideration requests, as applicable, for the annual total performance scores included in these reports for these thresholds in accordance with the appeals process set forth at § 484.335, which was finalized in the CY 2017 HH PPS final rule.

We analyzed the effects on participating HHAs of using the proposed 40 or more completed HHCAHPS surveys as compared to using 20 or more completed HHCAHPS surveys by examining OASIS measures submitted from January 1, 2015 through December 31, 2016, claims measures submitted from September 1, 2015 through September 30, 2016, and 12 months ending June 30, 2016 for HHCAHPS-based measures. We also found that achievement thresholds, which are calculated as the median of all HHAs' performance on the specified quality measures during the 2015 baseline year for each state, would not change by more than ±1.1 percent, with the largest changes occurring in the statewide achievement thresholds for the HHCAHPS Willingness to Recommend the Agency measure in Arizona (+1.1 percent) and Nebraska (−1.1 percent). Benchmarks (the mean of the top decile of Medicare-certified HHA performance on the specified quality measures during the 2015 baseline year, calculated for each state) had greater potential for change, ranging down to −3.2 percent. For instance, we found that when calculated using a minimum of 40 surveys rather than a minimum of 20 surveys, there was a −2.0 percent reduction in the benchmark for the HHCAHPS Willingness to Recommend the Agency measure for Arizona and a −1.7 percent reduction in the benchmark for Nebraska. We also found that when calculated using a minimum of 40 surveys rather than a minimum of 20 surveys, there was a −1.7 percent reduction in the benchmark for the HHCAHPS Communications between Providers and Patients measure for Arizona, a −1.7 percent reduction in the benchmark for Florida, and a −3.2 percent reduction in the benchmark for Nebraska.

Overall, the proposed change in the HHCAHPS minimum of 40 completed surveys is estimated to result in a limited percent change in the average statewide TPS for larger-volume HHAs, ranging from −0.4 through +2.2 percent. Because the underlying data does not cover the full 2016 calendar year, the data limitation may impact the final total performance scores and corresponding payment adjustment percentages. We provide estimates of the expected payment adjustment distribution based on the proposed minimum of 40 completed HHCAHPS surveys in the impact analysis of this proposed rule.

We are inviting public comments on our proposal to use 40 or more completed HHCAHPS surveys as the minimum to generate a quality measure score on the HHCAHPS measures, as is currently used in Home Health Compare and the Patient Survey Star Ratings. Therefore, we propose to revise the definition of “applicable measure” at § 484.305 to reflect this proposal, from a measure for which the competing HHA has provided 20 home health episodes of care per year to a measure for which a competing HHA has provided a minimum of 20 home health episodes of care per year for the OASIS-based measures, 20 home health episodes of care per year for the claims-based measures, or 40 completed surveys for the HHCAHPS measures. This proposal, if finalized, would apply to the calculation of the benchmark and achievement thresholds and the calculation of performance scores for all Model years, beginning with Performance Year (PY) One.

2. Proposal To Remove One OASIS-Based Measure Beginning With Performance Year 3

In the CY 2016 HH PPS final rule, we finalized a set of quality measures in Figure 4a: Final PY1 Measures and Figure 4b: Final PY1 New Measures (80 FR 68671 through 68673) for the HHVBP Model to be used in the first performance year (PY1), referred to as the starter set.

The measures were selected for the Model using the following guiding principles: (1) Use a broad measure set that captures the complexity of the services HHAs provide; (2) Incorporate the flexibility for future inclusion of the Improving Medicare Post-Acute Care Transformation Act of 2014 (IMPACT) measures that cut across post-acute care settings; (3) Develop `second generation' (of the HHVBP Model) measures of patient outcomes, health and functional status, shared decision making, and patient activation; (4) Include a balance of process, outcome and patient experience measures; (5) Advance the ability to measure cost and value; (6) Add measures for appropriateness or overuse; and (7) Promote infrastructure investments. This set of quality measures encompasses the multiple National Quality Strategy (NQS) domains (80 FR 68668). The NQS domains include six priority areas identified in the CY 2016 HH PPS final rule (80 FR 68668) as the CMS Framework for Quality Measurement Mapping. These areas are: (1) Clinical quality of care; (2) Care coordination; (3) Population & community health; (4) Person- and Caregiver-centered experience and outcomes; (5) Safety; and (6) Efficiency and cost reduction. Figures 4a and 4b of the CY 2016 HH PPS final rule identified 15 outcome measures (five from the HHCAHPS, eight from Outcome and Assessment Information Set (OASIS), and two from the Chronic Care Warehouse (claims)), and nine process measures (six from OASIS, and three New Measures, which were not previously reported in the home health setting).

In the CY 2017 HH PPS final rule, we removed the following four measures from the measure set for PY 1 and subsequent performance years: (1) Care Management: Types and Sources of Assistance; (2) Prior Functioning ADL/IADL; (3) Influenza Vaccine Data Collection Period: Does this episode of care include any dates on or between October 1 and March 31?; and (4) Reason Pneumococcal Vaccine Not Received, for the reasons discussed in that final rule (81 FR 76743 through 76747).

For Performance Year 3 (PY 3), we are proposing to remove one OASIS-based measure, Drug Education on All Medications Provided to Patient/Caregiver during All Episodes of Care, from the set of applicable measures. As part of our ongoing monitoring efforts, we found that based on the standard metrics of measure performance, many providers have achieved full performance on the Drug Education measure. For example, for the January 2017 IPRs (which covered the 12-month period of October 1, 2015 through September 30, 2016), the average value for this measure across all participating HHAs was 95.69 percent from October 2015 through September 2016. When looking at just September 2016, the mean value on this measure across all participating HHAs had increased to 97.8 percent. Also, there are few HHAs with poor performance on the measure. Based on the January 2017 IPRs, across all participating HHAs, the 10th percentile was 89 percent and the 5th percentile was 81.8 percent, but only 1.8 percent of HHAs had a value below 70 percent on the measure. We believe that removing this measure would be consistent with our policy, as noted in the CY 2017 HH PPS final rule (81 FR 76746), that when a measure has achieved full performance, we may propose the removal of the measure in future rulemaking. In addition, our contractor's Technical Expert Panel (TEP), which consists of 11 panelists with expertise in home health care and quality measures, expressed concern that the Drug Education measure does not capture whether the education provided by the HHA was meaningful.

The revised set of applicable measures, if our proposal to remove the OASIS-based measure, Drug Education on All Medications Provided to Patient/Caregiver during All Episodes of Care, is finalized, is presented in Table 43. This measure set would be applicable to PY3 and each subsequent performance year until such time that another set of applicable measures, or changes to this measure set, are proposed and finalized in future rulemaking.

We invite public comment on the proposal to remove one OASIS-based measure, Drug Education on All Medications Provided to Patient/Caregiver during All Episodes of Care, from the set of applicable measures for PY3 and subsequent performance years and Table 43.

C. Quality Measures for Future Consideration

The CY 2016 HH PPS final rule discusses the HHVBP Model design, the guiding principles to select measures, and the six priority areas of the National Quality Strategy (NQS) we considered for the Model (80 FR 68656 through 68678). Under the HHVBP Model, any measures we determine to be good indicators of quality will be considered for use in the HHVBP Model in future years, and may be added or removed through the rulemaking process. To further our commitment to objectively assess HHVBP quality measures, we are utilizing an implementation contractor that invited a group of measure experts to provide advice on the adjustment of the current measure set for consideration. The contractor convened a technical expert panel (TEP) consisting of 11 panelists with expertise in home health care and quality measures that met on September 7, 2016, in Baltimore, Maryland and via conference call on December 2, 2016. The TEP discussed developing a composite total change in ADL/IADL measure; a composite functional decline measure; a measure to capture when an HHA correctly identifies the patient's need for mental and behavioral health supervision; and a measure to identify if a caregiver is able to provide the patient's mental or behavioral health supervision, to align with § 409.45(b)(3)(iii) and the Medicare Benefit Policy Manual (Pub. 100-02), Chapter 7, Section 20.2. We discuss each of these potential measures in further detail in this section of the proposed rule. While any new measures would be proposed for use in future rulemaking, we are inviting comment on these potential measures now to inform measure development and selection.

As noted in the CY 2017 HH PPS final rule (81 FR 76747), we received several comments expressing concern that the measures under the Model do not reflect the patient population served under the Medicare Home Health benefit as the outcome measures focus on a patient's clinical improvement and do not address patients with chronic illnesses; deteriorating neurological, pulmonary, cardiac, and other conditions; and some with terminal illness. These commenters opined that the value of including stabilization measures in the HHVBP Model is readily apparent as it aligns the Model with the Medicare Home Health benefit. Commenters also expressed concerns that improvement is not always the goal for each patient and that stabilization is a reasonable clinical goal for some patients. Commenters suggested the addition of stabilization or maintenance measures be considered for the HHVBP Model. Many commenters objected to the use of improvement measures in the HHVBP Model. We did not receive any specific measures for future consideration. In the subsections that follow, we are identifying measures that we are considering for possible inclusion under the Model in future rulemaking and are seeking input from the public on the measures mentioned, as well as any input about the development or construction of the measures and their features or methodologies.

1. Total Change in ADL/IADL Performance by HHA Patients

The measure set finalized in the CY 2016 HH PPS final rule included Change in Daily Activity Function as Measured by the Activity Measure for Post-Acute Care (AM-PAC) (NQF #0430). However, the measure was removed in the CY 2017 HH PPS final rule and never used in the HHVBP Model because the measure required use of a proprietary data collection instrument in the home health environment. We are considering replacing Change in Daily Activity Function as Measured by AM-PAC (NQF #0430) with a composite total ADL/IADL change performance measure. During the September 2016 TEP meeting, an alternative to the Change in Daily Activity Function measure was presented. The TEP requested that a composite Total ADL/IADL Change measure be investigated empirically. This measure was discussed as part of the follow-up conference call, and the TEP supported continued development of the measure in the HHVBP Model as a way of including a measure that captures all three potential outcomes for home health patients: Stabilization; decline; and improvement. They provided input on the technical specifications of the potential composite measure, including the feasibility of implementing the measure and the overall measure reliability and validity. We have reviewed this suggested alternative and believe this measure would provide actionable and transparent information that would support HHA efforts to improve care and prevent functional decline for all patients across a broad range of patient functional outcomes. The measure would also improve accountability during an episode of care when the patient is directly under the HHA's care.

The name of this potential composite measure could be Total Change in ADL/IADL Performance by HHA Patients. The measure would report the average, normalized, total improved functioning across the 11 ADL/IADL items on the current OASIS-C2 instrument. The measure is calculated by comparing scores from the start-of-care/resumption of care to scores at discharge. For each item the patient's discharge assessed performance score is subtracted from the patient's start of care/resumption of care assessed performance score, and then divided by the maximum improvement value based on the number of response options for that item. These values are summed into a total normalized change score that can range from −11 (that is, for an episode where there is maximum decline on all 11 items used in the measure) to +11 (that is, for an episode where there is the maximum improvement on all 11 items). An HHA's score on the measure is based on its average across all eligible episodes. Patients who are independent on all 11 ADL/IADL items at Start of Care (SOC)/Resumption of Care (ROC) would also be included in the measure. The HHA's observed score on the measure is the average of the normalized total scores for all eligible episodes for its patients during the reporting period.

The following 11 ADLs/IADL-related items from OASIS-C2 items were included in developing a composite measure:

ADL OASIS-C2 items related to Self-Care:

• M1800 (Grooming).
• M1810 (Upper body dressing).
• M1820 (Lower body dressing).
• M1845 (Toileting hygiene).
• M1870 (Eating).

ADL OASIS-C2 items related to Mobility:

• M1840 (Toilet transferring).
• M1840 (Bed transferring).
• M1860 (Ambulation).

Other IADLs OASIS items:

• M1880 (Light meal preparation).
• M1890 (Telephone use).
• M2020 (Oral medication management).

Based on the measures identified above, we would risk-adjust using OASIS-C2 items to account for case-mix variation and other factors that affect functional decline but are beyond the influence of the HHA. The risk-adjustment model uses an ordinary least squares (OLS) regression framework because the outcome measure (normalized change in ADL/IADL performance) is a continuous variable.

The prediction model for this outcome measure was derived using the predicted values from the 11 individual outcomes that are currently used to risk adjust these 11 individual quality measures. Of the 11 values tested, the 8 identified in this proposed rule were found to be statistically related to the Total Change in ADL/IADL Performance by HHA Patients measure at p <0.0001 level and would be used in the prediction model that we are considering proposing to use to risk adjust the HHA's observed value with respect to this potential future measure. The prediction model for this outcome measure uses predicted values from the following individual outcomes (Note: The primary source OASIS item is listed in parenthesis after the name of the quality measure):

• Improvement in Upper Body Dressing (M1810).
• Improvement in Management of Oral Medications (M2020).
• Improvement in Bed Transferring (M1850).
• Improvement in Ambulation/Locomotion (M1860).
• Improvement in Grooming (M1800).
• Improvement in Toileting Hygiene (M1845).
• Discharged to the Community (M2420).
• Improvement in Toileting Transfer (M1840).

Two predictive models, one based on predicted values from CY2014 and one from CY2015, were computed. The correlations at the episode level between observed and predicted values for the target outcome measure Total Change in ADL/IADL Performance by HHA Patients are shown in Table 44.

The results in Table 44 suggest that either model would account for 25 percent or more of the variability in the outcome measure. These models could be considered very strong predictive models for the target outcome measure. Although the analysis supports developing a composite measure, the analysis assumes that the OASIS-C2 items identified to be used in the composite measure do not change; however, we recognize that OASIS-C2 items could be removed or added in any given year. We expect to conduct an additional analysis, in advance of any future proposal, to assess whether changes to OASIS-C2 items that are removed or added could significantly impact a HHA's ability to address several measures to improve its overall score in the composite measure. We are soliciting public comments on whether or not to include a composite total ADL/IADL change performance measure in the set of applicable measures, the name of any such measure, the risk adjustment method, and whether we should conduct an analysis of the impact of removal/addition of OASIS-C2 items.

2. Composite Functional Decline Measure

The second measure we are considering for possible inclusion under the Model in future rulemaking is a Composite Functional Decline Measure that could be the percentage of episodes where there was decline on one or more of the eight ADL items used in the measure. As noted in this proposed rule, we received comments on the CY 2017 HH PPS proposed rule suggesting that we consider the addition of stabilization or maintenance measures. To address this suggestion, we are considering a composite functional decline measure because the existing functional stabilization measures, taken individually, are topped out, with HHA level means of 95 percent or higher. This type of composite functional decline measure is similar to the composite ADL decline measure that is used in the Skilled Nursing Facility (SNF) Quality Reporting program (QRP). The SNF QRP measure is constructed from four ADL items: Bed mobility; transfer; eating; and toileting.

An HHVBP composite functional decline measure could provide actionable and transparent information that could support HHA efforts to improve care and prevent functional decline for all patients, including those for whom improvement in functional status is not a realistic care goal. This concept was discussed during the TEP meeting on September 7, 2016, with a follow-up conference call held on December 2, 2016. The TEP supported the inclusion of measures of stabilization and decline in the HHVBP Model, as well as further development of the composite functional decline measure. They provided input on the technical specifications of the potential composite measure, including the feasibility of implementing the measure and the overall measure reliability and validity.

When calculating the composite functional decline measure, we could use the following 8 existing OASIS-C2 items identified below:

• Ambulation/Locomotion (M1860).
• Bed Transferring (M1840).
• Toilet Transferring (M1840).
• Bathing (M1830).
• Toilet Hygiene (M1845).
• Lower Body Dressing (M1820).
• Upper Body Dressing (M1810).
• Grooming (M1800).

The measure could be defined as 1 if there is decline reported in one or more of these items between the Start of Care and the Discharge assessments and zero if no decline is reported on any of these items. As with other OASIS-based measures, a performance score for the measure would only be calculated for HHAs that have 20 or more episodes of care during a performance year.

The measure could be risk-adjusted using OASIS-C2 items to account for case-mix variation and other factors that affect functional decline but are beyond the influence of the HHA. The risk-adjustment model uses a logistic regression framework. The model includes a large number of patient clinical conditions and other characteristics measured at start of care. A logistic regression model is estimated to predict whether the patient will have length of stay of greater than 60 days. The predicted probability of length of stay of greater than 60 days is used, along with other patient characteristics, to construct a logistic regression model to predict the probability of decline in any of eight ADLs. This model is used to estimate the predicted percent of ADL decline at the HHA level. To calculate case-mix adjusted values, the observed value of the measure is adjusted by the difference between the HHA predicted percent and the national predicted percent. The risk-adjustment model reduces the adjusted difference between HHAs that serve a disproportionate number of longer-stay patients and those that serve patients with more typical lengths of stay of one episode.

Across all participating HHAs in the HHVBP Model, for HHAs that had less than 20 percent of episodes lasting more than 60 days, the average on the functional decline measure was 8.08 percent. This increased to 11.08 percent for HHAs with 20 percent to 40 percent of episodes lasting more than 60 days, 14.23 percent for HHAs with 40 percent to 60 percent of episodes lasting more than 60 days, and 20.59 percent for HHAs with more than 60 percent of episodes lasting more than 60 days. This finding suggests that, in addition to focusing on prevention of functional decline, we should also attempt to better predict a patient's functional trajectory and potentially stratify the population to exclude those on a likely downward trajectory. However, in spite of this finding, the inclusion of a measure that rewards providers for avoiding functional decline has the advantage of diversifying the set of measures for the HHVBP model. We are soliciting public comments on whether or not to include a composite functional decline measure in the set of applicable measures, the name of any such measure, the risk adjustment method, and whether we should conduct an analysis of the impact of removal/addition of OASIS-C2 items.

3. Behavioral Health Measures

Although we did not receive comments or suggestions through the rulemaking process for the HHVBP Model regarding behavioral or mental health measures, we recognize that the Model does not include such measures. The OASIS-C2 collects several items related to behavioral and mental health (M1700 Cognitive Functioning; M1710 Confusion Frequency; M1720 Anxiety; M1730 Depression Screening; M1740 Cognitive, Behavioral, and Psychiatric Symptoms; M1745 Frequency of Disruptive Behavior Symptoms; and M1750 Psychiatric Nursing Services). These items are used to compute both Improvement and Process measures as well as Potentially Avoidable Events. The inclusion of behavioral health measures is important for care transformation and improvement activities as many persons served by the Home Health program may have behavioral health needs.

The TEP made several suggestions during the December 2016 conference call as to whether the focus of a behavioral or mental health measure could be identifying whether a patient needed mental or behavioral health assistance compared to the supervision of the patient or advocacy assistance. The TEP supports the supervision type measure due to its opportunity for potential improvement. In further analyses, we identified two underlying components to outcomes for providing assistance. We developed a method, described below, to identify patients who have or do not have needs for mental or behavioral health supervision. We are considering further refining this method by identifying the involvement of the caregiver in addressing the patient's mental or behavioral health supervision needs as an important outcome measure, and we seek comment on whether this is an appropriate factor or feature that we should consider in developing such a measure in future rulemaking.

a. HHA Correctly Identifies Patient's Need for Mental or Behavioral Health Supervision

We are considering adding a HHA Correctly Identifies Patient's Need for Mental or Behavioral Health Supervision measure to the HHVBP Model in the future to capture a patient's need for mental or behavioral health supervision based on an identifier. This identifier is based on information from existing Neuro/Emotional/Behavioral Status OASIS items, along with other indicators of mental/behavioral health problems to identify a patient in need of supervisory assistance. The outcome measure assesses whether the HHA correctly identifies whether or not the patient needs mental or behavioral health supervision based on the OASIS SOC/ROC assessment item M2102f, Types and Sources of Assistance: Supervision and Safety.

A composite Mental/Behavioral Health measure could be a dichotomous measure that reports the percentage of episodes of care where the HHA correctly identifies: (a) Patients who need mental or behavioral health supervision; and (b) patients who don't need mental or behavioral health supervision. The numerator could be a combination of two values: (1) The number of episodes of care where the HHA correctly identifies patients who need mental or behavioral health supervision; plus (2) the number of episodes of care where the HHA correctly identifies patients who don't need mental or behavioral health supervision. The denominator is all episodes of care.

The composite measure requires that a patient's need for mental or behavioral health supervision be identified. The following algorithm was designed to identify if a patient was in need of mental or behavioral health supervision. If the patient met any of the following conditions, the patient was identified by the algorithm as in need of mental or behavioral health supervision:

• Was discharged from a psychiatric hospital prior to entering home health care (M1000 = 6);
• Is diagnosed as having chronic mental behavioral problems (M1021 and M1023);
• Is diagnosed with a mental illness (M1021 and M1023);
• Is cognitively impaired (M1700 > = 2);
• Is confused (M1710 > = 2);
• Is identified as having a memory deficit (M1740 = 1);
• Is identified as having impaired decision-making (M1740 = 2);
• Is identified as being verbally disruptive (M1740 = 3);
• Is identified as being physically aggressive (M1740 = 4);
• Is identified as exhibiting disruptive, infantile, or inappropriate behaviors (M1740 = 5);
• Is identified as being delusional (M1740 = 6); or
• Has a frequency of disruptive symptoms (M1745 >= 2.

The measure also requires that the HHA identify if the patient is in need of mental or behavioral health supervision. This requirement is based on the SOC/ROC code for M2102f, Types and Sources of Assistance: Supervision and Safety. If the HHA codes a value of 0, then the HHA has identified this patient as not needing mental or behavioral health supervision. If the HHA codes another value for M2102f, Types and Sources of Assistance: Supervision and Safety, then the HHA has identified this patient as needing mental or behavioral health supervision. The outcome measure is defined as the agreement between the algorithm's identification of a patient's need for mental or behavioral health supervision and the HHA's coding of this need. That is, if—

• The algorithm identifies the patient as not in need of mental or behavioral health supervision and the HHA identifies the patient as not in need of mental or behavioral health supervision, or
• The algorithm identifies the patient as in need of mental or behavioral health supervision and the HHA identifies the patient as in need of mental or behavioral health supervision, then
• The outcome is coded as 1, successful.

As with other OASIS-based measures, a performance score for the measure would only be calculated for HHAs that have 20 or more episodes of care during a performance year.

The measure is risk-adjusted using OASIS-C2 items to account for case-mix variation and other factors that affect functional decline but are beyond the influence of the HHA. The risk-adjustment model uses a logistic regression framework. The model includes a large number of patient clinical conditions and other characteristics measured at the start of care. To calculate case-mix adjusted values, the observed value of the measure is adjusted by the difference between the HHA predicted percent and the national predicted percent.

The prediction model for this outcome measure uses 39 risk factors with each risk factor statistically significant at <0.0001. The correlation for the model between observed and predicted values as estimated by Somers' D is 0.427, that yields an estimated coefficient of determination (r2) value based on the Tau-a of 0.201. This suggests that the variability in the model accounts for (predicts) approximately 20 percent of the variability in the outcome measure. The best statistic for evaluating the power of a prediction model that is derived using logistic regression is the c-statistic. This statistic identifies the overall accuracy of prediction by comparing observed and predicted value pairs to the proportion of the time that both predict the outcome in the same direction with 0.500 being a coin-flip. The discussed prediction model has a c-statistic equal to 0.713, which is considered to be strong. Using data from CY 2015, the episode-level mean for the HHA Correctly Identifies Patient's Need for Mental or Behavioral Health Supervision measure is 61.98 percent, nationally, and 62.98 percent for the HHVBP states.

b. Caregiver Can/Does Provide for Patient's Mental or Behavioral Health Supervision Need

We are considering including under the Model in future rulemaking a Caregiver Can/Does Provide for Patient's Mental or Behavioral Health Supervision Need measure that would encourage HHAs to ensure that patients who need mental or behavioral health supervision are receiving such care from the patient's caregivers, and would be a realistic care goal.

When considering how to develop a measure to determine whether or not the caregiver can/does provide the patient's mental or behavioral health supervision, we would create an identifier of a patient's need for mental or behavioral health supervision. This identifier is based on the same algorithm described in the previous section from existing Neuro/Emotional/Behavioral Status OASIS items along with other indicators of mental/behavioral health problems to identify a patient in need of supervisory assistance. The outcome measure is whether the HHA correctly identifies this patient as having the need for mental or behavioral health supervision based on the OASIS SOC/ROC assessment item M2102f, Types and Sources of Assistance: Supervision and Safety.

The measure could be a dichotomous measure that reports the percentage of episodes where patients with identified mental or behavioral health supervision needs have their needs met or could have their needs met by the patient's caregiver with additional training (if needed) and support by the HHA. The numerator is the intersection of: (1) The number of episodes of care where the patient needs mental or behavioral health supervision; and (2) the number of episodes of care where these patients have their needs met or could have their needs met by the patient's caregiver with additional training (if needed) and support by the HHA. By intersection, we mean that, for the numerator to equal one, a patient has to need mental or behavioral health supervision and has to have these needs met by his or her caregiver, or could have their needs met by the caregiver with additional training and/or support by the HHA. The denominator is all episodes of care. The algorithm discussed above for HHA Correctly Identifies Patient's Need for Mental or Behavioral Health Supervision could also be used to first identify if a patient was in need of mental or behavioral health supervision.

To identify whether caregivers are able to provide supervisory care or, with training, could be able to provide supervisory care for these patients, we could use the SOC/ROC code for M2102f, Types and Sources of Assistance: Supervision and Safety. If the HHA codes a value of 1 (Non-agency caregiver(s) currently provide assistance) or 2 (Non-agency caregiver(s) need training/supportive services to provide assistance), then the measure identifies that a caregiver does or could provide supervision to a patient who has been identified as needing mental or behavioral health supervision.

The outcome measure is defined as the agreement between the algorithm's identification of a patient's need for mental or behavioral health supervision and the availability of supervision from the patient's caregiver(s). That is, if—

• The algorithm identifies the patient as in need of mental or behavioral health supervision and there is documentation that the patient's caregiver(s) do or could provide this supervision; then
• The outcome is coded as 1, successful.

As with other OASIS-based measures, a performance score for the measure would only be calculated for HHAs that have 20 or more episodes during a performance year. We would use the same methodology to risk-adjust by using OASIS-C2 items and the prediction model described above. The prediction model for this outcome measure uses 55 risk factors with each risk factor significant at p < 0.0001. The correlation for the model between observed and predicted values as estimated by Somers' D is 0.672, that yields an estimated coefficient of determination (r²) value based on the Tau-a of 0.205. This suggests that the variability in the model accounts for (predicts) approximately 20 percent of the variability in the outcome measure. The best statistic for evaluating the power of a prediction model that is derived using logistic regression is the c-statistic. This statistic identifies the overall accuracy of prediction by comparing observed and predicted value pairs to the proportion of the time that both predict the outcome in the same direction with 0.500 being a coin-flip. The prediction model has a c-statistic equal to 0.836, which is considered to be extremely strong.

We are considering whether the HHA Correctly Identifies Patient's Need for Mental or Behavioral Health Supervision measure or the Caregiver Can/Does Provide for Patient's Mental or Behavioral Health Supervision Need measure would be most meaningful to include in the Model. We are also considering the interactions between the Home Health Grouping Model (HHGM) proposal on quality measures discussed in section III of this proposed rule and the HHVBP Model for the quality measures discussed in section IV.B of this proposed rule. We are soliciting public comments on the methodologies, analyses used to test the quality measure, and issues described in this section for future measure considerations. We will continue to share analyses as they become available with participating HHAs during future webinars.

V. Proposed Updates to the Home Health Care Quality Reporting Program (HH QRP)

A. Background and Statutory Authority

Section 1895(b)(3)(B)(v)(II) of the Act requires that for 2007 and subsequent years, each HHA submit to the Secretary in a form and manner, and at a time, specified by the Secretary, such data that the Secretary determines are appropriate for the measurement of health care quality. To the extent that an HHA does not submit data in accordance with this clause, the Secretary is directed to reduce the home health market basket percentage increase applicable to the HHA for such year by 2 percentage points. As provided at section 1895(b)(3)(B)(vi) of the Act, depending on the market basket percentage increase applicable for a particular year, the reduction of that increase by 2 percentage points for failure to comply with the requirements of the HH QRP, and further reduction of the increase by the productivity adjustment described in section 1886(b)(3)(B)(xi)(II) of the Act, may result in the home health market basket percentage increase being less than 0.0 percent for a year, and may result in payment rates under the Home Health PPS for a year being less than payment rates for the preceding year.

We use the terminology “CY [year] HH QRP” to refer to the calendar year for which the HH QRP requirements applicable to that calendar year must be met in order for an HHA to avoid a 2 percentage point reduction to its market basket percentage increase under section 1895(b)(3)(B)(v)(I) of the Act when calculating the payment rates applicable to it for that calendar year.

The Improving Medicare Post-Acute Care Transformation Act of 2014 (Pub. L. 113-185, enacted on October 6, 2014) (IMPACT Act) amended Title XVIII of the Act, in part, by adding new section 1899B of the Act, entitled “Standardized Post-Acute Care Assessment Data for Quality, Payment, and Discharge Planning,” and by enacting new data reporting requirements for certain post-acute care (PAC) providers, including Home Health Agencies (HHAs). Specifically, new sections 1899B(a)(1)(A)(ii) and (iii) of the Act require HHAs, Inpatient Rehabilitation Facilities (IRFs), Long Term Care Hospitals (LTCHs) and Skilled Nursing Facilities (SNFs), under each of their respective quality reporting program (which, for HHAs, is found at section 1895(b)(3)(B)(v) of the Act), to report data on quality measures specified under section 1899B(c)(1) of the Act for at least five domains, and data on resource use and other measures specified under section 1899B(d)(1) of the Act for at least three domains. Section 1899B(a)(1)(A)(i) of the Act further requires each of these PAC providers to report under their respective quality reporting program standardized patient assessment data in accordance with subsection (b) for at least the quality measures specified under subsection (c)(1) and that is for five specific categories: Functional status; cognitive function and mental status; special services, treatments, and interventions; medical conditions and co-morbidities; and impairments. All of the data that must be reported in accordance with section 1899B(a)(1)(A) of the Act must be standardized and interoperable, so as to allow for the exchange of the information among PAC providers and other providers, as well as for the use of such data to enable access to longitudinal information and to facilitate coordinated care. We refer readers to the CY 2016 HH PPS final rule (80 FR 68690 through 68692) for additional information on the IMPACT Act and its applicability to HHAs.

B. General Considerations Used for the Selection of Quality Measures for the HH QRP

We refer readers to the CY 2016 HH PPS final rule (80 FR 68695 through 68698) for a detailed discussion of the considerations we apply in measure selection for the HH QRP, such as alignment with the CMS Quality Strategy, which incorporates the three broad aims of the National Quality Strategy. As part of our consideration for measures for use in the HH QRP, we review and evaluate measures that have been implemented in other programs and take into account measures that have been endorsed by NQF for provider settings other than the HH setting. We have previously adopted measures with the term “Application of” in the names of those measures. We have received questions pertaining to the term “application” and want to clarify that when we refer to a measure as an “Application of” the measure, we mean that the measure would be used in a setting other than the setting for which it was endorsed by the NQF. For example, in the FY 2016 SNF PPS Rule (80 FR 46440 through 46444 we adopted an Application of Percent of Residents with Experiencing Falls with Major Injury (Long Stay) (NQF #0674), which is endorsed for the Nursing Home setting but not the SNF setting. For such measures, we intend to seek NQF endorsement for the HH setting, and if the NQF endorses one or more of them, we will update the title of the measure to remove the reference to “Application of.”

C. Accounting for Social Risk Factors in the HH QRP

We consider related factors that may affect measures in the HH QRP. We understand that social risk factors such as income, education, race and ethnicity, employment, disability, community resources, and social support (certain factors of which are also sometimes referred to as socioeconomic status (SES) factors or socio-demographic status (SDS) factors) play a major role in health. One of our core objectives is to improve beneficiary outcomes including reducing health disparities, and we want to ensure that all beneficiaries, including those with social risk factors, receive high quality care. In addition, we seek to ensure that the quality of care furnished by providers and suppliers is assessed as fairly as possible under our programs while ensuring that beneficiaries have adequate access to excellent care.

We have been reviewing reports prepared by the Office of the Assistant Secretary for Planning and Evaluation (ASPE ) and the National Academies of Sciences, Engineering, and Medicine on the issue of measuring and accounting for social risk factors in CMS' value-based purchasing and quality reporting programs, and considering options on how to address the issue in these programs. On December 21, 2016, ASPE submitted a Report to Congress on a study it was required to conduct under section 2(d) of the Improving Medicare Post-Acute Care Transformation (IMPACT) Act of 2014. The study analyzed the effects of certain social risk factors of Medicare beneficiaries on quality measures and measures of resource use used in one or more of nine Medicare value-based purchasing programs. The report also included considerations for strategies to account for social risk factors in these programs. In a January 10, 2017 report released by The National Academies of Sciences, Engineering, and Medicine, that body provided various potential methods for measuring and accounting for social risk factors, including stratified public reporting.

As discussed in the CY 2017 HH PPS final rule, the NQF has undertaken a 2-year trial period in which new measures, measures undergoing maintenance review, and measures endorsed with the condition that they enter the trial period can be assessed to determine whether risk adjustment for selected social risk factors is appropriate for these measures. Measures from the HH QRP, Rehospitalization During the First 30 Days of Home Health (NQF #2380), and Emergency Department Use without Hospital Readmission During the First 30 Days of Home Health (NQF #2505) are being addressed in this trial. This trial entails temporarily allowing inclusion of social risk factors in the risk-adjustment approach for these measures. At the conclusion of the trial, NQF will issue recommendations on the future inclusion of social risk factors in risk adjustment for quality measures.

As we continue to consider the analyses and recommendations from these reports and await the results of the NQF trial on risk adjustment for quality measures, we are continuing to work with stakeholders in this process. As we have previously communicated, we are concerned about holding providers to different standards for the outcomes of their patients with social risk factors because we do not want to mask potential disparities or minimize incentives to improve the outcomes for disadvantaged populations. Keeping this concern in mind, while we sought input on this topic previously, we continue to seek public comment on whether we should account for social risk factors in measures in the HH QRP, and if so, what method or combination of methods would be most appropriate for accounting for social risk factors. Examples of methods include: Confidential reporting to providers of measure rates stratified by social risk factors, public reporting of stratified measure rates, and potential risk adjustment of a particular measure as appropriate based on data and evidence.

In addition, we are seeking public comment on which social risk factors might be most appropriate for reporting stratified measure scores and potential risk adjustment of a particular measure. Examples of social risk factors include, but are not limited to, dual eligibility/low-income subsidy, race and ethnicity, and geographic area of residence. We are seeking comments on which of these factors, including current data sources where this information would be available, could be used alone or in combination, and whether other data should be collected to better capture the effects of social risk. We will take commenters' input into consideration as we continue to assess the appropriateness and feasibility of accounting for social risk factors in the HH QRP. We note that any such changes would be proposed through future notice and comment rulemaking.

We look forward to working with stakeholders as we consider the issue of accounting for social risk factors and reducing health disparities in CMS programs. Of note, implementing any of the above methods would be taken into consideration in the context of how this and other CMS programs operate (for example, data submission methods, availability of data, statistical considerations relating to reliability of data calculations, among others), so we also welcome comment on operational considerations. We are committed to ensuring that beneficiaries have access to and receive excellent care, and that the quality of care furnished by providers and suppliers is assessed fairly in CMS programs.

D. Proposed Data Elements for Removal From OASIS

We are proposing to remove 247 data elements from 35 OASIS items collected at specific time points during a home health episode. These data elements are not used in the calculation of quality measures already adopted in the HH QRP, nor are they being used for previously established purposes unrelated to the HH QRP, including payment, survey, the HH VBP Model or care planning. A list of the proposed 35 OASIS items and data elements are listed in Table 45 and also at https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HomeHealthQualityInits/OASIS-Data-Sets.html.