Do Selection or Treatment Effects Explain Differences in Medicare EOL Costs Among Hospice and Normal...
-
Upload
maria-conley -
Category
Documents
-
view
212 -
download
0
Transcript of Do Selection or Treatment Effects Explain Differences in Medicare EOL Costs Among Hospice and Normal...
Do Selection or Treatment Effects Explain Differences in Medicare EOL Costs Among
Hospice and Normal Care Users?Donald H. Taylor, Jr.
Duke UniversityFunded by The Robert Wood Johnson
Foundation’s Changes in Health Care Financing and Organization (HCFO)
Initiative
Question
• Does hospice save Medicare money? – Are observed differences in cost due to
treatment effects (hospice) or selection effects?
Background/Context
• Hospice improves patient QOL
• Large cost to Medicare near end-of-life– Hospice/cost effect is mixed in past work– Medicare hospice use expanding
Data
• Screening sample for NLTCS
• 10,319 decedents (died 1993-2000)
• Medicare claims 2 yrs. prior to death
• Total Medicare-financed payments
Analytic Approach
• Match decedents on pred prob of hospice
• Compare cost--hospice users v. matches– Marginal & cumulative costs– Varied follow-up periods
• Predicted weekly costs in LYOL– Regressed cost on time, t2, t3, t4
Results
• 13% died in hospice (1,309/10,319)
• 143 began, but did not die in hospice
• Median LOS=15 days; mean=50 – cancer dx longer
Comparing Marginal Costs of Hospice Decedents to Matched
Controls
-2000 -1000 0 1000 2000
1
Fo
llo
w u
p P
erio
d
Cost Difference
365 days
180 days
150 days
120 days
90 days
60 days
28 days
21 days
14 days
7 days
Comparing Cumulative Costs for Hospice Decedents and Matched
Controls
-4000 -2000 0 2000 4000
1
Pe
rio
d P
rio
r to
De
ath
Cost Differences
365 days
180 days
150 days
120 days
90 days
60 days
28 days
21 days
14 days
7 days
Hospice Save Medicare Money?
• Marginal acctg—Yes. Hospice saves Medicare ~$1,800 (50%) in the last 2 wks
• Cumulative acctg--Yes. Hospice saves Medicare ~$1,600 (15%) during last 60 days
Predicted Weekly Costs for Persons Who Died in Hospice
0
500
1000
1500
2000
2500
Weeks Prior to Death
To
tal W
eekl
y C
ost
($)
0
500
1000
1500
2000
2500
Hospice users' hospice cost Hospice users' total cost Non-hospice users' total cost
Crossover Point
Predicted Cumulative Costs for Persons Who Died in Hospice
0
5000
10000
15000
20000
25000
30000
0246810121416182022242628303234363840424446485052
Weeks Prior to Death
To
tal C
ost
Un
itl D
eath
($)
0
5000
10000
15000
20000
25000
30000
Hospice users' hospice cost Hospice users' total cost Non-hospice users' total cost
Crossover Point
Policy/Comment/Questions
• Matching increased cost savings ~10%
• Strong effect—savings in last 2 wks show cumulative savings in last 2 months
• Answer is sensitive to follow up period
• Earlier use of hospice will improve QOL and potentially save money
02
46
810
Per
cent
0 .1 .2 .3 .4 .5
Probability of Death in Hospice
Distribution of Predicted Probabilities of Death in Hospice
Matching
• Logit of hospice use– Predictors of use: younger, white, later death,
higher Medicare costs in year prior, dx cancer, urban
• Pred prob ranged from 0.03-0.49
• Matches were close; largest difference between hospice user and match was 0.01
Limitations/Extensions
• Time to death is assumed exogenous to choice of hospice/no hospice
• Partial solution: predict time of hospice entry and costs prospectively from date of study entry– Predict daily prob of hospice entry
– Calculate costs between each day and time of death
– Calculate expected cost after hospice entry by integrating over all days (for hospice users and non-users)