Copyright 1999 Federal News Service, Inc.
Federal News Service
MARCH 18, 1999, THURSDAY
SECTION: IN THE NEWS
LENGTH:
3344 words
HEADLINE: PREPARED TESTIMONY BY
WILLIAM
F. BLUHM, MAAA, FSA, FCA
CHAIR, RISK ADJUSTOR WORK GROUP
AMERICAN
ACADEMY OF ACTUARIES
BEFORE THE HOUSE WAYS AND MEANS
COMMITTEE
HEALTH SUBCOMMITTEE
SUBJECT - "ACTUARIAL REVIEW OF HCFA'S
HEALTH STATUS
RISK ADJUSTMENT METHODOLOGY"
BODY:
Good morning Chairman Thomas and members
of the committee. My name is Bill Bluhm and I am a principal with the actuarial
consulting firm of Milliman & Robertson in Minneapolis. I am appearing today
in my capacity as the Chair of the Risk Adjustors Work Group of the American
Academy of Actuaries. Our work group was formed at the request of the Health
Care Financing Administration (HCFA) to complete an actuarial review of the
health status risk adjustment methodology the agency will use
starting on January 1, 2000 to pay Medicare+Choice health plans.
As you are
aware, the use of a health status risk adjustment formula is
required by the Balanced Budget Act of 1997 (BBA). That law directed HCFA to
report to Congress on the proposed risk adjustment method and,
further, provides for, "an evaluation of such method by an outside, independent
actuary of the actuarial soundness of the proposal. "(BBA, Section 1853). Last
fall, the Health Care Financing Administration asked the American Academy of
Actuaries to perform this evaluation. As the Academy's Vice President of Health
at that time, I appointed a volunteer work group consisting of health actuaries
who are either consultants to or staff members with health plans and health
insurers to review HCFA's proposal. A list of the members of the work group is
attached to my testimony. Our analysis was included as part of the agency's
report to Congress which was issued earlier this month. The Academy's work was
provided pro bono, although HCFA did reimburse the members for travel expenses
associated with the meetings of the work group. HCFA's Proposal
Currently,
HCFA's payment rates for Medicare+Choice plans are adjusted to reflect the risk
characteristics of the plans' participants as defined by the demographic factors
of age, gender and the beneficiary's status (institutionalized or
non-institutionalized; Medicaid recipient or non- Medicaid; employed or not;
disabled or not). Beginning in the year 2000, HCFA is required by the BBA to
supplement these demographic adjustments with a health status risk adjustor.
HCFA plans to assign a risk score to each Medicare beneficiary based on
diagnosis information for that individual, taken from previous hospital
inpatient stays. The risk scores were developed using a list of"principal
inpatient diagnostic cost groups" (PIP-DCGs), which were developed for this
purpose. The previous medical costs for inpatient hospital stays incurred by the
individual are used to determine their expected future medical risk and,
therefore, how much the Medicare+Choice health plan in which they are enrolled
should be paid. New enrollees in Medicare will be assigned an estimated risk
score based on HCFA's analysis of existing Medicare fee-for-service (FFS) data.
Conclusions
The new risk adjustment system represents a
significant change for health plans, contracting providers, and health plan
members. While the Academy work group believes the conceptual basis of the
risk adjustment method proposed by HCFA is "actuarially sound,"
as we have defined it for this purpose, we have serious concerns about the
method's implementation, operation, and impact. These issues include:
-
Exclusions of certain risk categories from the risk adjustment
methodology, such as oneday hospital stays, which may penalize health plans that
effectively manage the delivery of health care.
- Lack of adequate testing
of the potential impact of the new methodology on health plans and
Medicare+Choice beneficiaries, although the phase-in will significantly soften
the impact of changes in reimbursement levels from what it might otherwise be. -
Administrative feasibility of the implementation of the new system because of
timing and data collection issues.
- The processing of extraordinary amounts
of newly collected data and completing a series of complex calculations
introduces an element of uncertainty that cannot be anticipated until health
plans and HCFA have full opportunity to understand the implications.
- Use
of only fee-for-service data as the basis for the development of risk
adjustment weights.
There is a substantial risk for the Medicare
system if the risk adjustment methodology does not work as
intended. The negative consequences could include withdrawal of Medicare+Choice
health plans from the market, financial problems or insolvency for health plans
and the potential for a reduction in benefits provided to beneficiaries. Because
of these concerns, the work group believes HCFA's decision to implement the new
methodology under a phased-in approach is a sound one and will limit changes
from the current payment system while HCFA and the health plans assess the
impact of the new methodology.
While HCFA has done much work in a short time
period to develop the new methodology and design implementation strategies,
additional work remains to fully define HCFA's risk adjustment
method and test application of the method to make sure it achieves the intended
results. The work group recommends that HCFA further modify the risk
adjustment model with the knowledge gained during the first year of
operation. Definition of Actuarial Soundness
The Academy was asked by HCFA
to evaluate the actuarial soundness of its proposal. For this purpose, there is
no widely recognized definition of "actuarial soundness." The work group
therefore analyzed HCFA's proposal in terms of: (1) established actuarial
criteria for risk adjustment, (2) Actuarial Standards of
Practice, and (3) the general principles and practices of actuarial science.
Actuarial Standards of Practice are guidelines developed by the Actuarial
Standards Board to help actuaries in their work. Specific actuarial goals and
criteria for risk adjustment are described in the Academy's May
1993 monograph titled, "Health Risk Assessment and Health Risk
Adjustment: Crucial Elements in Effective Health Care Reform." The
criteria used to evaluate risk adjustment systems are:
Accuracy: Because payments to health plans will be determined based on the
risk adjustment mechanism, accuracy and avoidance of
statistical bias is critical.
Practicality and Reasonable Cost: The
risk adjustment mechanism should not be so complex that
implementation is extremely cumbersome, thereby adding significant cost to the
system.
Timeliness and Predictability: Carriers setting premium rates should
be able to predict the impact of risk adjustment on their
premiums with a fair degree of accuracy and in a timely manner, in order to
avoid solvency concerns and disruption to members.
Resistance to
Manipulation: The risk adjustment mechanism should aim to make
it impossible for specific carriers to benefit financially by "gaming" the
mechanism.
The Academy's review took into account all aspects of the
proposed methodologies that impact on its "actuarial soundness," including, but
not limited to the proposed formulas, the availability, quality, and relevance
of the data required, and the ability to be implemented as intended.
In
addition, the Academy has evaluated the appropriateness of the proposed methods
in relation to available alternatives (including non- administrative data models
such as surveys, enhanced age/gender/status, and the status quo) and in light of
the modifications being made to the underlying base rates by county over the
same time period. Limitations of the Work Group's Analysis
It is important
to note that the work group's analysis and conclusions relied on the information
supplied by HCFA. During the review process, HCFA provided the work group with
preliminary results of the potential payment impact of the risk
adjustment methodology on Medicare+Choice plans. However, the work
group was not able to verify the accuracy of the data collected by HCFA or the
calculations used by HCFA to determine the impact on health plans.
In
addition, HCFA did not provide the work group with an assessment of the impact
of the risk adjustment methodology on beneficiaries, and the
scope of our opinion is similarly limited.
HCFA's risk adjusted payment
system is still a "work in progress", and it should be understood that our
opinion on the actuarial soundness of HCFA's proposals are based on the system
as they were described to us at the time we performed our review.
The work
group was not able to undertake a detailed analysis of the mathematical formulas
used to develop the risk adjustment methodology, but rather
focused its review on the conceptual and theoretical basis of the system.
Because HCFA is still working on the proposed methodology and there are a number
of unresolved implementation issues, our report is a qualified review of the
actuarial soundness of the proposal. Analysis and Recommendations
The new
methodology for making health status risk adjustments to
Medicare payments appears to meet the requirements of the Balanced Budget Act of
1997, provided the system is implemented carefully. On balance, and with a
phase-in, the proposed risk adjustment method appears to be a
reasonable first step in what should be a long-term evolutionary process. HCFA
is to be commended for the progress to date and for recognizing the limitations
of the proposal arising from the available data, timing requirements and areas
for future improvements.
In general, the work group believes the PIP-DCG
risk assessment methodology developed by HCFA meets the goals of risk assessment
I outlined earlier in my testimony. However, there are a number of concerns
about the health risk assessment formula that the work group raised in its
report:
Using Only Inpatient Data: A significant component of the PIP-DCG
model is the restriction of the risk adjustment method to
conditions identified by inpatient hospital claims. This feature has both
advantages and disadvantages. As one positive factor, this requirement matches
well with the information currently available to the Medicare program.
Currently, hospital claim information is more accessible and easier to audit
than ambulatory care data, and requires less additional work by health plans to
report to HCFA.
However, there are several drawbacks to a system that uses
only inpatient data. A major feature of managed care has been the measurable
shifting of inpatient care to outpatient sites and the substitution of less
invasive therapies to treat a given condition. When the risk assessment system
is restricted to inpatient claims, the members subject to effective managed care
can appear healthier than average, because of limits on what is measured.
If
outpatient (ambulatory) data is added to the inpatient claims information, a
better picture of the potential "risk" of each individual Medicare beneficiary
is obtained. We have therefore recommended that outpatient data be included in
HCFA's methodology as soon as it is feasible to do so. Exclusion of One-Day
Hospital Stays: The risk adjustment methodology does not "give
credit" for one-day hospitalizations, under the assumption that including them
may result in "gaming" of the system by health plans. If included, plans could
"game" the system by ordering unnecessary one-day stays for minor medical
conditions, in order to include beneficiaries in the health status risk
adjustment process, and thereby increase payments the next year.
The underlying concept of excluding one-day admissions does have merit. It
can reduce gaming of the system by requiring each hospitalization to be of a
certain severity (measured by a length of two days or more) and plans would not
have an incentive to hospitalize a patient overnight just to receive "credit."
However, the exclusion of one-day stays may unduly penalize plans which
efficiently manage the delivery of health care. This is because effective care
management tend to reduce stays to one day which might otherwise be two or more
day stays. Since those stays would then be excluded from the risk
adjustment process, this would penalizing plans for their efficiency.
According to the report from Health Economics Research (HER), which assisted
HCFA in designing the PIP-DCGs, excluding one-day stays reduces the predictive
power of the health status risk adjustment methodology. Also,
it might be noted that excluding one-day hospitalizations shifts the issue
of"gaming" from whether to hospitalize someone at all to a question of whether
to keep the patient for a second hospital day.
The work group suspects that
the disadvantages of excluding one-day hospitalizations may outweigh any
possible gain. It would be appropriate to analyze the risk
adjustment methodology based on whether it is easier to "game"
admissions or to "game" length of stay and any resulting adverse incentives for
health plans. HCFA may want to consider either using one-day stays as part of
the risk adjustment formula or giving a partial credit or other
adjustments for those hospitalizations in structuring payments to health plans.
Principal Diagnosis: The PIP-DCG model measures conditions by capturing the
principal diagnosis recorded on each inpatient claim. The use of the principal
diagnosis for the PIP-DCG model is based on existing coding practices for
inpatient claims used by hospitals. Since only the principal diagnosis is
generally used, it is possible that not all appropriate information is collected
or used. A qualifying condition could be listed as the secondary (or other)
diagnosis, which could be a contributing factor leading to the need for
hospitalization.
Alternately, there is a common belief that many secondary
conditions currently reported are not as reliable and should not be included in
the measurement system. Since the initial stages of the risk assessment system
will be using data that was recorded without the presence of direct coding
incentives, it may be reasonable to use only principal diagnosis information.
However, as the PIP-DCG system is implemented, the restriction to using only
principal diagnostic groups should be re-evaluated.
Number and Development
of the PIP-DCG Groups: Health Economics Research developed the diagnostic groups
using a HCFA survey of Medicare FFS data which sampled 5% of Medicare
beneficiaries. The claims information for this sample fell in the two-year
interval from January 1, 1995 through December 31, 1996. Beneficiaries who were
not alive and enrolled in Medicare for the entire time period were excluded, as
were individuals who would not have been eligible for the Medicare+Choice
program for various reasons. Because of these limits, the actual sample
represents roughly a 3.5% sample. We have included some technical
recommendations in our report, which can be included as HCFA revises the
methodology. Excluding Discretionary Conditions: The base cost group (those
individuals who are not assigned health status risk scores) also includes
Medicare beneficiaries with diagnoses that were determined by HER to be
discretionary, vague, or only occasionally resulted in inpatient admissions. The
exclusion of those "discretionary" conditions has the beneficial effect of
reducing potential bias in the formula against Medicare+Choice health plans with
well managed care delivery systems by not giving credit for discretionary
admissions and by removing the incentives to hospitalize a patient for minor
illness.
However, we suggest that the diagnoses included in the base cost
group should be reviewed in the future as coding practices change under the
PIP-DCG system. If hospitals become more aggressive in their coding in the
future, the percentage of claims falling into a PIPDCG may change and weights
would need to be recalibrated, particularly if the PIP-DCG method is used beyond
the currently planned three-year period.
Chemotherapy: HCFA has
indicated that beneficiaries who are undergoing chemotherapy will be placed in a
diagnosis category based on the patient's secondary diagnosis (most likely
cancer). Since the medical conditions underlying the need for chemotherapy
represent high-cost, ongoing conditions that are predictive of future medical
expenses, it is appropriate that they be included in the risk assessment model.
The work group believes including chemotherapy as part of the diagnosis groups
will increase the ability of the methodology to predict future health care
costs.
Exclusion of Indirect Medical Education Costs: The model developed by
HER excludes indirect medical education (IME) costs from the Medicare FFS data
used to calculate the relative weights used in this system. The IME costs are
approximately two-thirds of the total graduate medical education costs currently
paid through Medicare (the FFS data does include direct medical education
expenses). While it is technically incorrect to include any graduate medical
education costs (since medical education costs will be paid outside of the
capitation rate in the future), any distortion is likely to be small. However,
it is possible there will be some internal inconsistencies in the model since
high-cost conditions captured in the PIP-DCGs may more likely be treated in a
tertiary care or teaching hospital.
Factors for Newly Enrolled Medicare
Members: HCFA decided to develop a special set of risk scores for those
individuals who are eligible for Medicare for the first time and do not have any
previous encounter data in the Medicare system. HCFA used FFS data to construct
average expenditures for categories of newly eligible members (beneficiaries who
become eligible for Medicare because of age or disability, or members who were
previously eligible for coverage but deferred entry into the Medicare system).
Newly eligible members will be assigned an estimated risk score based on HCFA's
estimate of their predicted medical expenditures. The validity of these risk
scores is unclear. The work group suggested that HCFA review its risk scores for
the newly eligible once current data is available.
Additional Testing:
Health Economics Research performed a number of tests on the PIPDCG risk
adjuster methodology to determine how accurately it predicts total expected
medical costs. The recommendations made by HER regarding several key components
of the model such as the use of inpatient data only, exclusion of one-day stays
and the number of PIP-DCG groups to be used, appear to be reasonable based on
the FFS data which was reviewed. While the HER report discusses potential bias
against managed care organizations that deliver care more efficiently than fee
for service providers, HER did not have managed care data to determine what, if
any, bias exists. HCFA has completed some preliminary testing of the potential
impact of the new risk adjustment methodology on
Medicare+Choice plans, including managed care organizations. In order to
understand the impact of the new system on the marketplace, the work group
suggests that HCFA update these tests as additional data is available, and as
health plans gain more experience with the operation of the risk
adjustment mechanism.
Cost-Benefit Analysis: The proposed system is
relatively new and it is likely that there will be difficulties in
implementation. It would be very helpful to establish more accurate estimates of
the cost of implementing the PIP-DCG methodology and any modifications (such as
using ambulatory data) and to determine the benefits to be derived from these
systems before final decisions as to implementation are made. We suggest that
consideration be given to producing a cost- benefit analysis of the PIP-DCG
methodology and any subsequent modifications. The analysis should specifically
include the costs incurred by health plans due to changes to the system.
Actuarial Oversight: HCFA apparently plans to conduct additional analysis of
the impact of the PIP-DCG methodology on managed care plans. It is unclear what
form that impact analysis will take. In addition, there is a need for continuing
monitoring and testing of the system and future modifications. The work group
suggests that additional actuarial review be included as the system and
subsequent changes are implemented.
END
LOAD-DATE: March 19, 1999