ACO Attribution Method Matters for the Seriously Ill Presented by - - PowerPoint PPT Presentation

ACO Attribution Method Matters for the Seriously Ill

Presented by Brystana G. Kaufman, PhD, MSPH June 4, 2019

@Brystana @DukeMargolis

William K. Bleser1, PhD, MSPH Robert Saunders1, PhD Lia Winfield2, PhD Mark Japinga1, MPAff Nathan Smith2, PhD Hannah L. Crook1

1 Robert J. Margolis, MD, Center for Health Policy, Duke University, Washington, DC 2Leavitt Partners, Salt Lake City, UT 3

Department of Population Health Sciences, Duke University School of Medicine, Durham, NC 4Leavitt Partners, Washington, DC;

Funded by the Gordon and Betty Moore Foundation

5/31/2019

Acknowledgements and Disclosures

Brystana G. Kaufman1, PhD, MSPH David Anderson1, MSPPM; Courtney Van Houtven3, PhD David B. Muhlestein4, PhD, JD Mark B. McClellan1, MD, PhD

Serious Illness Care in ACOs

Any Inpatient claim* AND Any Skilled Nursing, Home Health or DME Claim* AND a serious illness OR 3+

• ther comorbidities:

*Calendar year prior to ACO PY; Medicare fee-for-service Beneficiary Summary File, Chronic Conditions Warehouse, Cost &Use

3

10%

• f

ACO

75% of readmissions 47% of Medicare Costs

>$50K PBPY

Medicare Shared Savings Program

Retrospective Attribution

• Most MSSP Contracts
• After Performance Year

Prospective Attribution

• Physician Group Practice

Demonstration

• Before Performance Year

5/31/2019 4

Prospective Cohort Retrospective Cohort Patients in Both Cohorts

ACO Attribution Conceptual Model

Decedents Prospective ONLY Retrospective ONLY

Retrospective Cohort

• N=1,306,722*
• 13% died
• $37,400 pbpy (actual)
• $33,000 pbpy (truncated)

Prospective Cohort

• N=1,336,562*
• 16% died
• $47,300 pbpy (actual)
• $35,600 pbpy (truncated)

Retrospective Only

• N=264,067
• 14% died
• $47,200 pbpy (actual)
• $40,500 pbpy (truncated)

Prospective Only

• N=293,907
• 28% died
• $91,400 pbpy (actual)
• $51,500 pbpy (truncated)

5/31/2019 6

Prospective ACO cohorts include more decedents

*Overlapping; 100% National Medicare FFS Claims Data Notes: 2015-2016 PY; PBPY=Per Beneficiary Per Year Medicare Expenditures; Costs are inflated, standardized, annualized, and truncated

7

0.45 0.66 0.70 0.77

0.00 0.20 0.40 0.60 0.80 1.00

<90 Days (N=69,485) 90-179 Days (N=50,976) 180-269 Days (N=42,941) 270+ Days (1,173,160)

Beneficiary Survival

Retrospective Attribution Rates: Lowest for Q1 Decedents

Con Condit itional on

• n

Pros

• spectiv

ive Assig ssignment (P (PY Y 2015 2015-2016) 2016) Predicted Probability

Among FFS Top 1% of FFS truncated Among Seriously Ill 5% were truncated Among Decedents 25% were truncated Among Truncated 80% were decedents

8

MSSP Stop Loss Rule: Differentially impacts decedents’ annualized costs

Generalized Linear Model with exchangeable correlation structure; ACO and Year Fixed Effects;

9

$24,973 $28,266 $27,061 $29,520 Retrospective Cohort Prospective Cohort Retrospective Cohort Prospective Cohort ACOs with Shared Savings (N=253) ACOs without Shared Savings (N=571)

Per Beneficiary Per Year Costs: Higher in Prospective than Retrospective Cohorts

ACOs and Serious Illness Care

Increase accountability for high risk patients

• 1. Prospective

cohorts include more decedents.

10

Consider alternatives to annualizing costs

• 2. Truncation

differentially impacts decedents

Risk adjustment and unintended consequences

• 3. Prospective

cohorts are more expensive

@Brystana @DukeMargolis

References

• Kelley, A. S. (2013). Epidemiology of care for patients with serious illness. J Palliat Med, 16(7), 730-733. Retrieved

from http://dx.doi.org/10.1089/jpm.2013.9498. doi:10.1089/jpm.2013.9498

• Kelley, A. S. (2014). Defining "serious illness". J Palliat Med, 17(9), 985. Retrieved

from http://dx.doi.org/10.1089/jpm.2014.0164. doi:10.1089/jpm.2014.0164

• Kelley, A. S., & Bollens-Lund, E. (2018). Identifying the Population with Serious Illness: The "Denominator" Challenge. J Palliat Med,

21(S2), S7-s16. Retrieved from http://dx.doi.org/10.1089/jpm.2017.0548. doi:10.1089/jpm.2017.0548

• Kelley, A. S., Covinsky, K. E., Gorges, R. J., McKendrick, K., Bollens-Lund, E., Morrison, R. S., & Ritchie, C. S. (2017). Identifying Older

Adults with Serious Illness: A Critical Step toward Improving the Value of Health Care. Health Serv Res, 52(1), 113-131. Retrieved from http://dx.doi.org/10.1111/1475-6773.12479. doi:10.1111/1475-6773.12479

• Kelley, A. S., Morrison, R. S., Wenger, N. S., Ettner, S. L., & Sarkisian, C. A. (2010). Determinants of treatment intensity for patients

with serious illness: a new conceptual framework. J Palliat Med, 13(7), 807-813. Retrieved from http://dx.doi.org/10.1089/jpm.2010.0007. doi:10.1089/jpm.2010.0007

5/31/2019 11

Serious Illness Criteria

• A. 3+ of the Following:
• Myocardial Infarction
• Atrial Fibrillation
• Ischemic Heart Disease
• Depression
• Rheumatoid/ Osteoarthritis
• Osteoporosis
• Transient Ischemic Attack
• Cancer: Breast or Prostate
• Asthma
• Chronic Kidney Disease
• Diabetes
• B. Any 1 of the following:
• Advanced liver disease or cirrhosis
• Cancer: Lung, colorectal or endometrial
• Chronic obstructive pulmonary disease
• Congestive heart failure
• Dementia including Alzheimer’s
• Diabetes with ischemic heart disease or

peripheral vascular disease

• Hip fracture
• Renal failure as indicated by Dialysis

1) Any Inpatient claim* AND 2) Any Skilled Nursing, Home Health or DME Claim* AND 3) A OR B:

*Within a calendar year using Medicare fee for service claims (5% sample and MBSF); lagged for attribution analysis;

12

Medicare Expenditures

• Defined using annual cost and use beneficiary summary files
• annualized to adjust for survival<1yr, as done in calculation of MSSP per capita spending
• adjusted for inflation using Consumer Price Index
• Standardized across counties using CMS geographic payment adjustment factors.
• Annualized costs were truncated at the 99th percentile in the full fee-for-service Medicare population by

eligibility code group, as done in calculation of MSSP per capita spending

• Attribution Models are adjusted for beneficiary age, gender, race, hospice use, E&M visit, as well as county

HCC, Medicare advantage and Medicaid rates; year fixed effects.

• ACO PBPY models adjust for ACO and year fixed effects and apply exchangeable correlation structure.

13

14

$- $20 $40 $60 $80 $100 $120 $140 $160 $180 Retrospective Prospective Retrospective Prospective Retrospective Prospective All Seriously Ill (N=1,600,629) Hospice Users (N=164,264) Decedents (N=255,852)

Medicare Expenditures, $1000 PBPY, truncated PBPY, actual

Impact of the Stop Loss Rule on PBPY Costs