Medicare Part B Therapy: Issues and Beneficiary Analyses

1

Examining Long-Term Care Episodes and Care
History for Medicare Beneficiaries A
Longitudinal Analysis of Elderly Individuals with
Congestive Heart Failure Stephanie Maxwell
Timothy Waidmann APHA Annual Meeting Boston,
MA November 6, 2006
2

• Background
• Congestive heart failure (CHF) is the leading
  medical condition among the elderly.   
• Significant policy concern regarding CHF
  hospitalization rates
• CHF is a common target of disease management
  programs. 
• Few large-scale studies have explored the CHF
  patients use of long-term care (LTC) services
  and Medicare services combined

3

• Overview of Study Design
• Longitudinal (36-month) analyses
• National cohort of elderly who were first
  hospitalized for CHF in 1999. 
• Identify patterns over three years of Medicare
  service use and spending, enrollment in Medicaid,
  and nursing home entry. 
• Estimate hazard models of risks of
  re-hospitalization, nursing home admission and
  death, controlling for health status. 

4
Data Sources (mainly 1999-2002 files)

• 100 Medicare claims files (all service
  types)
• 100 Medicare enrollment files
•  
• 100 MDS patient assessment records
• Area Resource File and Interstudy HMO Files
  
•  
•  

5

• Defining the Cohort in the Claims Data
• Final cohort 296,462 elderly
• Cohort consists of elderly hospitalized, in 1999,
  for their first hospitalization for CHF.
• The principal diagnosis field of acute hospital
  records was searched for a set of diagnosis codes
  indicating CHF as the primary reason for
  hospitalization.
• Scanned a 5-year look-back period of hospital
  claims (1994-1998 claims) to screen out
  individuals whose first CHF admission occurred
  before 1999.
• To assure a comparable look-back period, we
  included only those age-eligible for Medicare in
  January 1994 in the cohort.

6
Statistical Methods

• Bivariate analyses -- of outcomes
  stratified by patient and area characteristics
• Survival models -- to estimate the effects
  of covariates on the instantaneous risk of an
  outcome, through measuring the elapsed time
  before an outcome is observed.
• Two-part use and spending models estimated
  models for the first six months following CHF
  hospitalization and also for the three years
  following CHF hospitalization.
•  

7
Outcomes Measures of Hazard Models

• Survival
• Subsequent CHF hospitalization
• Subsequent non-CHF hospitalization
• Medicaid enrollment
• Nursing home entry
•  

8
Outcomes Measures of Two-Part Use and Spending
Models

• CHF hospitalizations
• Other hospitalizations
• SNF stays
• Home health use
• Hospital outpatient use
• Physician services use
•  

9
Person-Level Independent Variables

• Demographics (age group, race, sex)
• Charlson comorbidity score
• Length of stay of the index CHF hospitalization
  
• Nursing home use prior to index CHF
  hospitalization
• Utilization and spending variables between the
  index CHF hospitalization and outcome
• Quarterly physician spending
• Quarterly hospital outpatient spending
• Quarterly acute hospital spending (except in
  models of death and non-CHF hospitalizations)
• CHF hospitalizations (except when used as an
  outcome)
• Oher hospitalizations (except when use as an
  outcome)
• SNF stays
• Medicare home health use
• Nursing home use (except when used as an
  outcome)
•  

10
County-Level Independent Variables

• Urban influence
• HMO penetration
• Median county income
• Supply rates per 1000 elderly
• All physicians
• Cardiologists
• Short-term hospital beds
• Long-term hospital beds
• SNF beds
• Nursing home beds

• Presence of a facility in the county
• Short-term hospital
• Nursing home
• Rural health clinic
• Federally qualified health clinic
• Population mortality rates for 10 selected
  medical conditions

11
Summary of Findings

• Over 3 years following index hospitalization for
  CHF
• 36 had additional CHF hospitalizations
• 68 had hospitalizations for other conditions
• 42 had SNF stays
• 15 entered a nursing home (non-Medicare)
• 7 enrolled in Medicaid
• 56 died
• 11 had NH use prior to their index CHF
  hospitalization
• Average 3-year spending 35,000
• Non-CHF hospitalizations was largest source of
  spending

12
Findings Death

• SNF use is the dominant risk
• Age -- 5 additional years ? 13 to 30 higher
  risk
• Charlson -- additional comorbidity ? 10 higher
  risk
• Index LOS -- additional day ? 2 higher risk
• SNF use ? 200 higher risk
• Physician spending per quarter (thousands) ? 15
  to 40 higher risk
• NF use ? 15 to 47 higher risk

13
(No Transcript)
14
Findings CHF Hospitalizations

• Approximately 15 increased risk associated with
  
• 5-year age increase
• Additional comorbidity
• Race Black
• Physician spending per quarter (thousands)
• Home health use
• Whites have higher death risks and blacks have
  higher rehospitalization risks. This is
  consistent with each other in suggesting that
  whites are more severely ill once hospitalized.

15
(No Transcript)
16
Findings Other Hospitalizations

• Compared to CHF hospitalization, key
  differences
• are regarding race and home health use
• Blacks ? 10 to 20 higher risk for CHF
  hospitalizations
• But blacks ? 5 to 10 lower risk for other
  hospitalizations
• Home health use ? 15 higher risk for CHF
  hospitalizations
• But home health use ? 20 lower risk for other
  hospitalizations

17
(No Transcript)
18
Findings Nursing Home Entry

• SNF use and prior NH use are dominant risks
• SNF use ? several hundred percent higher risk
• Prior NH use ? 100 higher risk
• Additional CHF hospitalizations ? 20 higher risk
• Other hospitalizations ? 5 to 20 higher risk

19
(No Transcript)
20
Findings Medicaid Enrollment

• SNF use and NH use are the dominant risks
• (200 to 300 higher risk)
• Three factors each increasing risk by 6 to
  24
• Prior NH use
• Hospitalizations
• Home health use
• Race black ? 40 to 100 higher risk

21
(No Transcript)
22
Methodological Contributions to the CHF
Literature

• Large-scale, national study of CHF population
  with a long follow-up (36 months).
• Survival analysis jointly accounts for
  utilization and mortality risk. This is
  important when studying elderly or
  high-mortality conditions. Logistic regression
  may give misleading impressions.
• Controlled for health status using comorbidity
  index and prior nursing home use.
• Controlled for area variation using state and
  6-level urban influence variable. In terms of
  urban influence, risks hinged on large metro
  county residence. An urban/rural flag would
  incorrectly attribute practice patterns typical
  in large center cities to the surrounding metro
  areas and to smaller cities.

23
Conclusions

• Higher CHF rehospitalization among African
  Americans. Target for disease management
  programs?
• Bivariate findings suggest decreasing intensity
  of care with age. Multivariate models do not.
• Importance of more than CHF hospitalization in
  cohort.
• Geographic variation in utilization and health.

24
Main Study Limitation Missing Data on Social
Support, Income, Functional Status

• This study had mixed findings regarding the
  effect (sign) of home health use on outcomes.
  Our findings on home health use in relation to
  SNF use may point to influential characteristics
  not available in our data social support,
  individual income, and ADL information on
  community residents.
• The importance of these factors in understanding
  LTC use is well-established in the literature.
• This studys findings suggest that these factors
  may be important in understanding medical use as
  well, when examining a chronic and ultimately
  debilitating disease like CHF.

25

• Principal Investigators
• Stephanie Maxwell, PhD and Timothy Waidmann, PhD
• smaxwell_at_ui.urban.org twaidman_at_ui.urban.org
• 202-261-5825 202-261-5718
• Health Policy Center
• The Urban Institute
• 2100 M Street, NW
• Washington, DC 20037
• fax [phone]
• Funder
• Centers for Medicare and Medicaid Services