Transcript OACT`s Excess Cost Growth Method

Projecting Long-Range Aggregate
Health Expenditures:
OACT, CBO, and SOA Approaches
The Middle Atlantic Actuarial Club
Presentation by:
Todd Caldis, Ph.D., J.D.
Senior Economist
National Health Statistics Group, CMS\OACT
September 18, 2009
Introduction
• The long-range projection challenge
• Examine and compare the three principal
approaches in use for projecting longrange health expenditures
• General observations (brief)
• Questions and discussion
The Projection Task
• CMS, Office of the Actuary (OACT)
• Congressional Budget Office (CBO)
• Society of Actuaries (SOA)
The Projection Problem
• Almost uninterrupted trend of health sector
growth faster than the growth of our
overall economy
• Growth can’t go on forever---an economy
devoted only to health care can’t exist
Health Sector Growth
Chart 1 - National Health Expenditures (NHE) as a Percentage
of Gross Domestic Product (GDP)
1960-2007
18%
16%
14%
NHE
Share
of
GDP
(%)
12%
10%
8%
6%
4%
2%
0%
1960
1964
1968
1972
1976
1980
1984
1988
Year
Source: Centers for Medicare & Medicaid Services, Office of the Actuary
1992
1996
2000
2004
International Comparisons
Chart 2 - Total Expenditures on Health as a Percentage Share of GDP,
OECD Countries, 2006
16.0
United States
14.0
12.0
Belgium
10.0
Share
8.0
of
GDP (%)
6.0
Austria
Australia
France
Germany
Canada
Denmark
Switzerland
Netherlands
Portugal
Sweden
GreeceIcelandItaly
New Zealand
Norway
Spain
United Kingdom
Hungary
Finland
Japan
Ireland
Luxembourg
Slovak Republic
Czech Republic
Korea Mexico
Poland
Turkey
4.0
2.0
0.0
Source: OECD Health Data 2008.
Note: For the United States the 2006 data reported here do not match the 2006 data point for the United States in
Chart 1 since the OECD uses a slightly different definition of "total expenditures on health" than that used in the
National Health Expenditure Accounts.
Excess Cost Growth Trends
Table 1 - Average excess cost growth rates, selected time periods 1975-2007
Time period
Average constant-dollar, per capita
growth
NHE (rounded)
GDP (rounded)
Average Excess
Cost Growth
Rate
Periods beginning with 1975:
through 1980 (5 years)
4.5%
2.7%
1.8%
through 1985 (10 years)
4.6%
2.5%
2.2%
through 1990 (15 years)
4.9%
2.4%
2.5%
through 1995 (20 years)
4.5%
2.1%
2.4%
through 2000 (25 years)
4.2%
2.3%
1.9%
through 2007 (32 years)
4.0%
2.1%
1.9%
Periods beginning with 1980:
through 1985 (5 years)
4.8%
2.3%
2.5%
through 1990 (10 years)
5.1%
2.3%
2.9%
through 1995 (15 years)
4.5%
1.9%
2.6%
through 2000 (20 years)
4.1%
2.2%
1.9%
through 2007 (27 years)
3.9%
2.0%
1.9%
Periods beginning with 1985:
through 1990 (5 years)
5.5%
2.3%
3.3%
through 1995 (10 years)
4.4%
1.8%
2.6%
through 2000 (15 years)
3.9%
2.2%
1.7%
through 2007 (22 years)
3.7%
1.9%
1.8%
Periods beginning with 1990:
through 1995 (5 years)
3.3%
1.3%
2.0%
through 2000 (10 years)
3.1%
2.2%
0.9%
through 2007 (17 years)
3.2%
1.8%
1.4%
through 2000 (5 years)
2.9%
3.1%
-0.2%
through 2007 (12 years)
3.1%
2.1%
1.1%
1.3%
2.0%
Periods beginning with 1995:
Periods beginning with 2000:
through 2007 (7 years)
3.3%
Note: NHE rates are adjusted for age-gender effects.
Source: Centers for Medicare and Medicaid Services, Office of the Actuary.
The Projection Challenge
• Health sector can’t grow faster than the
economy forever---so it won’t.
– Cannot project off of historic growth trends
alone in the long-run
• But when, how, and at what rate will
excess cost growth slow down?
– What evidence to rely on
Method #1: OACT Excess Cost
Growth Model
• The first excess cost growth projection
model
• Result of an evolution that started in the
late 1990s
• Uses a core assumption about the
average rate of excess cost growth for the
last 51 years of the 75-year horizon
• More complicated modeling apparatus
brought in to refine the assumption into its
finished form
Projection from a Current Law
Perspective
• Idea of a policy-neutral baseline, the state of the
world if specified existing policies remained
unchanged.
• A perspective that characterizes much of
OACT’s work
• For long-term projections the state of the world if
benefits now promised are maintained
– Assumes that revenue to pay benefits would
and could be found
Basic Formulas: Definition of
Excess Cost Growth
• Difference of growth factors(algebraically
equivalent to a difference of rates):
• Health expenditures and Age-Gender are per
enrollee; GDP is per capita.
Basic Formulas: The Excess Cost
Ratio
• The XRatio can be thought of an excess cost
growth factor.
A Projection Formula
• Can implement with aggregate expenditures in
the formula if also include a population factor in
the equation:
The Core Excess Growth
Assumption: GDP+1
• Assumption that average per enrollee,
age-gender adjusted rate of health
expenditure growth will exceed per capita
rate of GDP growth by 1% for the years 25
through 75 of the projection horizon.
– If GDP growth is 5% and age-gender adjusted
health expenditure growth will be 6%, a
percentage point more
• Can be implemented with either nominal
or real GDP
Implementing the Method for a 75
Year Horizon
•
•
•
Years 1 -10 determined by OACT short-run
projections.
Years 11 – 25 based on straight-line transition
from excess cost ratios for year 10 to the
consolidated excess cost ratios for year 25.
Years 25 – 75 projections based upon excess
cost ratios generated from the CMS
Computable General Equilibrium (CGE) Model.
Generating Long-Range Excess
Cost Ratios (Years 25 – 75)
• OACT CGE Model
– A Ramsey-type model optimizing for a
representative agent
– Incorporates assumptions about technological
change and costs effects for the health sector
• CGE solved for a time series of excess
cost ratios whose resulting Part A actuarial
balance is financially equivalent to the
cumulative 75 year balance under a linear
GDP + 1 excess cost growth rate
History of the GDP + 1 Excess
Cost Growth Assumption
• Until 2001 projections in Trustees Reports implicitly
based upon a “GDP + 0” assumption of no excess
cost growth for projection year 26 forward.
• Pure “GDP + 1” implemented in 2001 based upon
recommendations of the 2000 Technical Review
Panel for projection year 26 forward.
• “GDP + 1” affirmed as “within the range of
reasonable assumptions” by 2004 Technical Review
Panel.
• Refined GDP +1 with OACT CGE “smoothing”
implemented starting with 2006 Medicare Trustees
Report.
Thinking About “GDP + 1”
• Based upon expert consensus.
• Reflects expectation of a substantial slowdown
in historic rates of excess cost growth.
• Reflects belief that technological change will
remain an important driver of excess cost growth
for most of the 75 year projection horizon.
• An understandable and discussable core
assumption
The Idea of “Natural Brakes”
• Rationale that reconciles idea of a
spending slowdown with current law
– Cost sharing provisions in current law
Medicare
– Diffusion of cost-saving practice patterns
(Spillover)
Method #2: CBO’s Excess Cost
Growth Model
• First long-term projection using this
method appeared in November, 2007 in
“The Long-Term Outlook for Health Care
Spending”
• The same framework was employed with
modified assumptions for CBO’s most
recent long-term health expenditure
projections released in June, 2009
• Like OACT, CBO projects from a current
law perspective
CBO’s Disaggregated View of
Excess Cost Growth
20
Medicaid
15
10
Total
Medicare
Other
5
0
-5
- 10
1976
1980
1984
1988
1992
1996
2000
2004
Source: Congressional Budget Office, The Long-Term Outlook for Health Care Spending,
November 2007.
CBO’s Basic Projection Formula
Source: The Long-Term Outlook for Health Care Spending (CBO: November,
2007)
CBO’s Excess Cost Growth:
The Definition of xt
Source: The Long-Term Outlook for Health Care Spending (CBO: November, 2007)
Relation between CBO and
OACT Excess Cost Growth
• A conceptual connection useful for comparing CBO
and OACT approaches
• In practice OACT has a single series of Xratios
applicable to all parts of the US health sector; CBO
has three series related to each other, but distinct,
each series being applicable to a different part of the
US health sector
Other OACT/CBO Similarity
18
Medicare Share of GDP
OACT Projection Model Results with CBO Excess Cost Rates vs. Published CBO Model
Results
2018-2082
16
14
12
Share 10
of
GDP
8
%)
OACT with CBO Excess Rates
CBO Model
6
4
2
0
2018
2028
2038
2048
2058
Year
OACT with CBO Excess Rates
Source: Centers for Medicare and Medicaid, Office of the Actuary
CBO Model
2068
2078
CBO’s Disaggregated Excess
Cost Growth Projection Method
• Projects excess cost growth for the US
health sector partitioned into Medicare,
Medicaid, and Non-Medicare, NonMedicaid
• For each part of the health sector
implements different initial rate of excess
cost growth-measure of historic average
• Beginning in the 12th year of the 75-year
projection period each part’s rate of
excess cost growth begins decelerating
“Identification” of CBO’s
Projection
• Assumption about interconnection across the three parts
of the health sector of the rate of excess cost growth
deceleration
– In 2007 Medicare and Medicaid assumed at 25% and
75% of the rate of deceleration in Non-Medicare,
Non-Medicaid
– In 2009 Medicare assumed to decelerate at 1/3 the
rate of NMNM; Medicaid at the same rate as NMNM
• Choose an NMNM deceleration rate such that real per
capita Non-Health personal consumption expenditure is
always non-negative (No recomputation in 2009)
CBO’s Deceleration Factors
• For 2007 projection, the excess cost growth rate
decelerated at 4.6% per year for NMNM, 3.45% per year
for Medicaid, and 1.15%
• 2009 projection assumed that NMNM would continue to
decelerate at approximately the same rate as in the 2007
projection, 4.5% per year and changed the assumed
connection with deceleration elsewhere
– Medicaid at same deceleration rate as NMNM, 4.5%
– Medicare decelerating at one-third the rate of NMNM,
1.5% per year.
• Remember---deceleration refers to deceleration of the
rate of excess cost growth unique to each of the three
parts of the health sector
Other Aspects of CBO Long-Range
Health Expenditure Projection
• For the first 10 years uses short-range projections where available
(e.g. Medicare); for other parts of health sector otherwise assumes
that historic rate of ECG prevails
• For year 11 of the projection period each part of the health sector
assumed to be at its historic average rate of ECG
– 2007: 2.4% (Care), 2.2% (Caid), 2.0% (NMNM)
– 2009: 2.3% (Care), 1.9% (Caid), 1.8% (NMNM)
• Deceleration of ECG in each part of health sector starting in year 12.
• NMNM nominally includes Medicare-related and Medicaid-related
items
– E.g. State Medicaid share and Medicare OOP
– CBO makes assumptions to pull such items out of NMNM,
compute NMNM growth without them, grow the removed items at
the same rates as Medicare or Medicaid, and then recombines
the removed items with the rest of NMNM
CBO’s Theory of ECG
Slowdown
• Assumes no change in federal policy, i.e.,
current law
• Private sector flexibility to respond to cost
pressures
• State laws may change
– Medicaid program therefore more able to
respond to cost pressure than Medicare but
less able than private sector
• Assumes spillover effects in practice norms and
pricing
Comparison of OACT and CBO
Methods
• Results are driven by substantively
different assumptions about rates of
excess cost growth
• Reflect different views of how current law
would play out into the future
• Mechanical implementations are different
and raise issues of method unique to each
model
Comparison of ECG Assumptions
Excess Cost Ratios
OACT Assumptions vs. CBO Assumptions
2019-2083
1.04
1.035
1.03
OACT Part D
1.025
Excess
1.02
Ratio
OACT Part B
CBO - Medicare
1.015
1.01
OACT Part A
OACT Medicare (All Parts)
1.005
1
2019
2029
2039
2049
2059
2069
Year
OACT Part A
OACT Part B
OACT Part D
OACT Medicare (All Parts)
Source: Centers for Medicare and Medicaid Services, Office of the Actuary
CBO - Medicare
2079
Effect of Excess Rate Assumptions
on Projections
Gross Medicare Share of GDP
OACT vs. CBO Projections
2019-2083
18
16
14
CBO Model
12
Share
10
of
GDP
8
(%)
OACT Model
6
4
2
0
2019
2029
2039
2049
2059
Year
OACT Model
CBO Model
Source: Centers for Medicare and Medicaid Services, Office of the Actuary;
2069
2079
Method #3: Getzen-Society of
Actuaries (SOA)
• Released in 2007 under auspices of the Society of
Actuaries
• Constructed by Professor Thomas Getzen under
contract with SOA and in close consultation with a
working group of actuaries
• Intended as a tool for evaluating retirement health
benefit plans; base scenario with allowance user
discretion in choice of inputs
• Allows fast excess cost growth to continue until NHE
attains a prescribed share of GDP or a specified limit
year is reached
Required Inputs
• Per capita rate of real income growth (Income%)
• Income elasticity, rate at which real income growth
converts into increased demand for health care (IncElas:
A constant)
• Per capita rate at which technology and other factors
contribute to increased demand for health care (Tech%:
A percentage independent of income growth)
• Baseline Health Sector Share of GDP in 2011(ShareBase
• Average CPI (appears to affect nominal but not real
action in the model)
Significance of Inputs:
Fixed Annual GDP Share Increment
• Annual increment is fixed annual
increment by which health sector share
grows if no limits are encountered
• If ShareBase = 17.5% would reach 42.1% in
2083 and 100% of GDP in @2250.
Growth Limiting Options
• Resistance Share: A user-specified health
sector share of GDP (25% in baseline
model)
– Beyond the resistance point the annual
increment is gradually reduced.
– Asymtotically health sector share can never
grow to more than twice the resistance share
• Limit Year: User-designated year beyond
which the annual increment is rapidly
phased to zero.
Other Options
• Change parameter values in 2020-2030
and 2030+
• Playing with per person baseline medical
costs
SOA Outputs
• Health sector share
• Per capita income
• Per capita medical costs
• Percentage increases
What’s distinctive about SOA
Model?
• Annual Increment forces user to think
about substantive sources of excess cost
growth
• Base model (used without resistance
share and limit year options) builds in
gradual deceleration of excess cost growth
• Resistance Share and Limit Year options
permit consideration of sudden-end
scenarios
Comparison of Projection Results
Com parison of Long Range Projections
of National Health Expenditures
2018-2082
60
50
CB O
40
OA CT
P er cent age
Sh a r e
SOA -Get zen
30
of GD P
20
10
0
2018
2028
2038
2048
2058
Y ear
OACT
Source: Centers for M edicare and M edicaid Services, Office of the Actuary
CBO
SOA-Getzen
2068
2078
Similarities Among All Methods
• Each based on assumptions about excess cost growth.
• Each envisions a slowdown in excess cost growth.
– OACT envisions natural brakes smoothly decelerating to zero ECG in
75 years
– CBO envisions different parts of the health sector decelerating at
different rates with a non-steady state at the end of 75 years
– SOA-Getzen in its unconstrained version envisions smooth deceleration
of health sector growth with no steady state reached at the end of 75
years.
• Each accepts the proposition that the health sector will continue to
grow despite a slowing of ECG rates
• In fact, each method projects a health sector share between 40 and
50 percent at the end of 75 years (SOA-Getzen in its unconstrained
version)
• Each of these models documents the implications of assumed trend
patterns better than they explain them
Research Issues Common to All
Projection Methods
• All need to work on strengthening the evidence
supporting key assumptions about slowing ECG
– OACT working with a simple insurance cost-sharing
model, also has a modest spillover project in
progress, and is in contact with an NIA-funded effort
to incorporate features of insurance cost-sharing
provisions into the Urban Institute’s microsim
– CBO may be working long-term to incorporate
insurance cost-sharing features into its CBOLT
microsim
– Not aware of research through SOA; maybe someone
here can comment
Constraints on Development of
These Methods
• OACT and CBO must produce current law
projections
– CL necessarily involves a scenario in which existing
program is sustained into the indefinite future
– In commentary both OACT and CBO acknowledge
that the actual sustainability of their long-term
scenarios is open to serious question
– In some circumstances current law scenarios are
declared unsustainable.
Other Development Constraints
• Professional standards
• Some stability in methods desirable;
erratic swings in LT projections due solely
to methodolgical brainstorms would send a
confusing message
• Tension between complexity and
transparency
Usefulness of These Projection
Methods
• Credible warnings about a long-term
systemic solvency crisis that our society
faces
• Would be a bonus if these methods
provided insight toward solving the crisis
– Bonus but not the fundamental mission
• Explanatory limits of these methods not an
excuse for complacency
References
• Medicare Trustees Reports
– http://www.cms.hhs.gov/ReportsTrustFunds/
– Has link to an official OACT memorandum about
projection methodology (by Caldis)
• Congressional Budget Office
– The Long-Term Outlook for Health Spending
(November, 2007)
– The Long-Term Budget Outlook (June, 2007)
• Society of Actuaries
– http://library.soa.org/research/health/research-hlthcare-trends.aspx
Contact Information
Todd Caldis
(410) 786-1217
[email protected]