Copyright 1999 Federal News Service, Inc.
Federal News Service
FEBRUARY 25, 1999, THURSDAY
SECTION: IN THE NEWS
LENGTH:
5047 words
HEADLINE: PREPARED STATEMENT OF
HEIDI
MARGULIS
VICE PRESIDENT, GOV. AFFAIRS
HUMANA INC.
BEFORE THE
HOUSE COMMERCE COMMITTEE
HEALTH AND THE ENVIRONMENT
SUBCOMMITTEE
BODY:
Introduction
Good morning,
Mr. Chairman. I am Heidi Margulis, Vice President of Government Affairs at
Humana, Inc. I am pleased to be here this morning to talk with you about the
effects of risk adjustment on Medicare+Choice organizations.
Humana has been an active participant in the Medicare program since the mid
1980's - we currently provide coverage for over 500,000 Medicare+Choice
enrollees and are committed to continued participation in this program. We have
also been active in the discussions that have occurred between managed care
plans and HCFA on the topic of risk adjustment.
One of
HCFA's goals in implementing a risk adjustment system Iezzoni,
LI, et al. "Paying more fairly for capitated care," New England Journal of
Medicine, 339(26), December 24, 1998, pp. 1933-38. , Medicare+Choice Rates - 45
Day Notice, http://www.hcfa.gov/stats/hmorates/45d1999/45d.htm. for
Medicare+Choice is to ensure that Medicare payments to health plans are accurate
and that they reflect the health care needs of enrolled members. We believe that
this is a laudable goal and are committed to working with HCFA and other
interested parties in this endeavor. Payments that are risk adjusted based on
health care diagnostic data appear to be, on the surface, an improvement over
the current methodology, but only if designed fairly and implemented correctly.
My testimony addresses the issues that Humana has identified and the potential
effects of risk adjustment on Humana's 500,000 enrollees, and
the Medicare+Choice program.
Risk Assessment and Risk
Adjustment It may be helpful to first describe the difference between
risk assessment and risk adjustment. Risk assessment is a means
of determining objectively how much an individual or a subgroup differs in cost
from the average of the entire group. Individuals who are projected to incur
more costs for medical services are considered relatively high risks (and, thus,
have higher risk scores) than those who are expected to incur lower costs. See
"Health Risk Assessment and Health Risk Adjustment, Crucial
Elements in Effective Health Care Reform," Monograph Number One, American
Academy of Actuaries, May 1993 for a more detailed explanation. Risk assessment
can be accomplished using only demographic data, with diagnostic information, or
through use of health status surveys.
Risk adjustment may
be called "health-based payment." It is a process that can be used to determine
the amount of funds that should be allocated to account for the differences in
risk characteristics. While all covered individuals should be allocated a "base"
or minimum payment, only for those enrollees with high risk characteristics
should a health plan receive additional risk adjustment
transfers.
Brief Actuarial History of Risk Adjustment
Health plan actuaries have been using various forms of risk
adjustment for years for pricing premiums for health insurance
coverage. Age/sex rating, experience rating, and tier rating have been
components of the methods used to determine premiums to be charged for a
specific category of individuals. The insurance industry's practice of health
underwriting has been based on the ability to appropriately project next year's
costs based on current claims experience (for large employer groups), or on past
medical conditions along with age and sex (for individuals), or on a combination
(for small employer groups). This type of cost and illness information has been
used in a way that is generally similar to how the new health risk
adjustment methods operate.
Humana has had actual experience with
some of the early adopters of risk adjustment methods. One of
the best-designed early "natural experiments" was the Health Insurance Plan for
California (the "HIPC"), a small group purchasing pool. Bertko, J. and Hunt, S.
"Case Study: The Health Insurance Plan of California," Inquiry, 1998,
35:148-153. The HIPC implemented an inpatient-data risk adjuster for the 1996/97
contract year, after two years of development and simulation. Humana's small
employer division, Employers Health Insurance, participated in the HIPC as one
of two original PPOs and was actively involved in the design and implementation
of the HIPC's risk adjustment method. Humana and the HIPC
learned several lessons as we progressed from the "good idea" stage to full
implementation, including: -- A full and open process between vendors (health
plans) and the payment agency (the HIPC's parent, California's Managed Risk
Medical Insurance Board) was very helpful in designing a practical method; --
Any new data collection process will have flaws which only "trial and error" can
uncover and which can then be corrected; and -- A simulation period for a brand
new payment method is invaluable for learning the details of the approach and
for evaluating the "real world" impact (on premiums and behavior).
What Hcfa
Did - Right and Wrong
I will turn now to a few specific comments about the
proposed risk adjustment system. In its efforts to implement a
risk adjustment system, HCFA has considered and responded to
several critical issues. First, HCFA realized that given a January 1, 2000
implementation date, the use of inpatient data (while imperfect) is the only
practical option. Second, HCFA has adopted a prospective payment method, using a
6-month data lag. This means that payments to a health plan will be made based
on information that at most is between 6 and 18 months old. Third, HCFA plans to
implement the system using a "back-loaded" transition approach, somewhat
limiting the degree to which health plan payments are affected in the early
years of the transition period.
The system that HCFA plans to implement uses
the Principal Inpatient Diagnostic Cost Group ("PIP-DCG") model, which groups
diagnostic data according to expected cost Ellis, RP, Pope, GC, Iezzoni, LI, et
al. "Diagnosis-based risk adjustment for Medicare capitation
payments," Health Care Financing Review," 17(3), 1996, pp. 101-28.. This model
has been extensively tested on Medicare fee-for-service data alone. However, it
relies exclusively on inpatient hospital data; no data on services provided
outside the hospital are used. There are several shortcomings to a system that
uses only inpatient data, including a payment bias against Medicare+Choice
plans.
Many managed care organizations have implemented programs to treat
patients on an outpatient basis when appropriate. For example, Humana has
developed several disease management programs for our enrollees - ranging from
asthma to diabetes to complex chronic conditions to congestive heart failure. As
part of these programs, our health plan enrollees collaborate with their
caregivers to manage their care, often eliminating or shortening inpatient stays
and improving health status. High levels of patient satisfaction are associated
with these programs as well as reduced costs. When health plans implement
programs that manage care and keep enrollees out of the hospital, they bear the
full cost of those programs. Without such programs, enrollees would be more
likely to be hospitalized, an outcome that is costly and unnecessary as the
hospital may no longer be the most effective setting for such care. The PIP-DCG
risk adjustment method penalizes plans that have such disease
management systems because such plans will have fewer inpatient admissions.
The proposed risk adjustment system also excludes
"short" hospital stays, those that are shorter than two days. In so doing, HCFA
again penalizes those health plans that are able to provide treatment during a
short inpatient stay. As an example, an individual with a particular diagnosis
who is enrolled in Medicare FFS must be hospitalized for three days prior to
discharge to a sub-acute care facility. An individual with the same diagnosis
enrolled in a Medicare+Choice plan may be hospitalized for only one day, then
moved to a sub-acute facility (which is not part of the inpatient hospital). The
Medicare+Choice plan would not receive any additional payment for the treatment
of this individual, since the patient did not have a qualifying inpatient
admission.
There is a similar problem for conditions that can be treated
equally well on an inpatient or outpatient basis - so called "discretionary
diagnoses." In these cases, health plans are only paid if the condition is
treated on an inpatient basis. While the PIP-DCG system does make some effort to
exclude such cases Iezzoni, LI, et al. "Paying more fairly for capitated care.",
some discretionary diagnoses are still included on the final list of diagnostic
groups that lead to additional payment above the base payment amount such as
many types of congestive heart failure. It is unlikely that there will be a
large scale effort on the part of health plans to move care back into the
hospital to increase payment. However, a very real potential effect is that
health plans will be less likely to be innovative - either to invest in new
disease management programs or in new technologies that would allow patients to
be treated on an outpatient basis.
There are also technical shortcomings to
the proposed risk adjustment system. First, there is a
difference in the time period used to calibrate the PIP-DCG model and what will
actually be used to pay health plans. The current model was developed using data
from one calendar year to predict expenses for the immediate next calendar year
(i.e., calendar 1995 data were used to predict calendar year 1996 expenses). In
HCFA's 45-day notice Medicare+Choice Rates - 45 Day Notice,
http://www.hcfa.gov/stats/hmorates/45d1999/45d.htm., a 6-month time lag for the
actual implementation of the PIP-DCG model is described - this model will use
data from a 12-month period (July 1- June 30) to predict expenses for the year
beginning six months later (i.e., data from July 1, 1998-June 30, 1999 will be
used to predict year 2000 expenses) using the original, "no lag" risk weights. A
more appropriate technical solution would be for a different set of risk weights
to be used; these weights would be calibrated to incorporate the 6-month time
lag.
Another technical issue relates to the criteria that were used to
determine whether a particular diagnosis would be included in the group of
diagnoses that lead to increased payments for health plans. To be included, at
least 1,000 beneficiaries in the original sample had to have the diagnosis. Such
a decision rule helps to stabilize payments in the model; however, by setting a
minimum threshold, admissions with very high costs may be excluded and plans
will not receive any additional payment for these very high cost cases.
Risk Adjustment Implementation Issues
There are several
important issues related to the implementation of the proposed risk
adjustment system. One of the reasons we are reluctant to have risk
adjusted payments implemented this year is that we have received insufficient
information from HCFA on the details of the risk adjustment
process. For plans to have confidence in the risk adjustment
system that HCFA implements, we must be able to understand the system and be
able to replicate HCFA's results. To this end, we believe that HCFA must
disclose all of the formulas used in the risk adjustment
process - we cannot replicate results given with the information we have
received thus far.
To date, HCFA has not disclosed all of the formulas used
for the various components of the risk adjustment process even
though plans have asked for this information for several months. As one example,
a re-scaling factor is used to transform the current AAPCC county ratebook into
the new risk county ratebook that forms the basis for calculating an
individual's risk-adjusted payment. Thus far, HCFA has only provided a brief
description of this formula - not all the components of the formula. Months ago,
the American Association of Health Plans (AAHP), the industry's trade
association, and others submitted to HCFA a list of desired information that
would allow plans to make the same kinds of calculations that HCFA is making. A
summary of the types of information needed is included in one of the attachments
to this testimony.
HCFA has indicated that when it does release data to the
health plans on March 1st, it will do so on a summary basis. Again, this will
not allow plans to compare their own results with those of HCFA - the plans need
individual data to determine whether they are using the same data and whether
they are applying the risk adjustment technology appropriately.
We understand that HCFA has faced a daunting time schedule in attempting to
implement a risk adjustment system for January 1, 2000 and
believe that they could do more to disclose relevant information to health plans
if they had more time.
Many of the key implementation issues relate to the
gathering, transmission, and analysis of data. Each health plan submits its data
to a fiscal intermediary, which in turn submits the data to HCFA. To date, plans
have not been able to confirm that the data submitted to HCFA are being
transmitted, received and used correctly or whether there are other systems'
issues HCFA has identified. If there are HCFA or fiscal intermediary systems
problems that need to be fixed, plans are concerned those fixes may be delayed
due to Year 2000 compliance issues. There may be as yet undetected problems in
the data transfer process, potentially leading to incorrect payments to plans.
While Humana is generally pleased with the performance of its Fiscal
Intermediary, Palmetto Government Benefit Administrators, we have had to work
out several time-consuming processes to understand the nature of the Medicare
FFS edit "error messages" that became part of the process. This happens, we
believe, because HCFA is forcing the "square peg" of managed care data into the
"round hole" of a Fiscal Intermediary's FFS information system. Here are just a
few of our issues: -- Inability to obtain a relevant list of error codes. If we
had obtained a list with the coding logic that creates an error message, we
would correct the problems at the source of the error and avoid further
submissions with these so-called errors. -- The Fiscal Intermediary provides
errors grouped in an almost useless format - by provider. We need to have a more
"user-friendly" or managed care- relevant error report - such as returning our
own list with the error reason annotated in the same format. -- There are a lot
of claims - Beginning March 1, we will be submitting encounter data in batches
of 11,000 every two weeks - because of capacity constraints at the Fiscal
Intermediary. Humana is being forced to build a special program just to organize
the error list electronically to allow reconciliation and correction. The HCFA
contractor that combines all of the Fiscal Intermediaries' claims has its own
turnaround and through-put issues. As reported to me by our staff, error edits
can take between one day and three weeks for each batch. The process of
informing us of errors is incredibly inefficient, as demonstrated by the
following: -- The HCFA contractor "kicks out" one error at a time on a claim and
then returns it. When corrected for that specific error, the contractor may then
find another error on the same claim and return it again. This process is
repeated until all codes on a claim have been accepted. -- Each separate error
requires a new claim line, which then clogs up our claim system with unnecessary
claims history. -- We cannot obtain the logic behind the edits to identify and
fix the source of the errors and are forced to continue this awkward,
time-consuming, and costly process. This is just one example of where resources
are used for unnecessary administrative costs rather than for patient care.
There are related problems with HCFA's own system (Common Working File). We
received a 22,000-page report to be reconciled. Although most of these claims
were accepted, we understand that some of the remainder may have "disappeared."
These are just a few examples of our frustrations and concerns about the
start-up phase of this new data collection process. While HCFA has made a
valiant attempt to prepare for start-up, there are still too many unresolved
issues. Among them, our Chief Financial Officer must "attest" to the accuracy of
our data submissions. He takes this responsibility seriously and is greatly
concerned about the remaining problems. Second, any loss of data unfairly
penalizes health plans since most hospital admissions create additional payments
for sick members. Missing data means reduced payments in 2000. Each qualifying
missed claim represents approximately $1,900 to $26,500.
Phase-In of
Risk Adjustment
HCFA included a phase-in schedule for the
risk adjustment system. A transition approach has a long
history in the Medicare program - such rules were used for the implementation of
Medicare's Prospective Payment System (PPS) and Resource-Based Relative Value
System (RBRVS).
During the transition for PPS, for example, hospitals
received a blend of a hospital-specific payment rate and a Federal payment rate.
In the first year of the transition, hospital payments were more heavily
weighted toward the hospital's own costs, while toward the end of the
transition, hospital payments were more heavily weighted toward the Federal
payment rate. 1986 Annual Report to Congress: Impact of the Medicare Hospital
Prospective Payment System, Health Care Financing Administration, May 1989.
Effects of Risk Adjustment As many Subcommittee Members and
staff may know from a HCFA briefing on January 14, 1999, preliminary estimates
by HCFA analysts indicate that HCFA's fully phased-in PIP-DCG risk adjuster
would reduce payments to health plans for the 195 plans that were measured.
Presentation to Congressional staff by the Health Care Financing Administration
on January 14, 1999. It must be noted that a risk adjustment
method can be designed to be budget-neutral; HCFA, however, has released a
method that apparently is intended to reduce payments to health plans even
further than intended by Congress.
There are two main issues related to the
impact on health plans: (1) Should the new risk adjuster be budget neutral? and
(2) Are the results of PIP-DCG risk adjustment method biased
because of the reliance on only inpatient data?
Although HCFA analysts and
other researchers Brown, R.S., Bergeron, J.W., Clement, D.G., Hill, J.W., and
Retchin, S.M., "Does Managed Care Work for Medicare? An Evaluation of the
Medicare Risk Program for HMOs," Princeton, NJ: Mathematica Policy Research,
Inc., December 1993. have previously submitted studies using Medicare Fee For
Service data indicating concerns about overpayment of health plans in excess of
10%, HCFA's own impact assessment using actual preliminary health plan data
showed a payment reduction of 7.6% for a typical month. Because this analysis
used the initial submission of somewhat incomplete admission data, the actual
payment reduction impact is likely to be less than 7% with better data. Reducing
the PIP-DCG method's biases would likely eliminate more of the preliminary
estimate of overpayment.
Some or all of this overpayment issue has already
been addressed. Implementation of a risk adjuster that is not budget-neutral
would be the sixth reduction in payment to health plans relative to FFS
Medicare. The first reduction is the long-established 5% reduction in the
payment to health plans relative to the average FFS payment - a reduction that
is continued through use of the 1997 county ratebooks under the BBA. This
reduction was originally made to assure savings in the Medicare Risk program.
The second reduction is the five year phase-in of a "growth reduction" of 2.8%
under BBA, which is an arbitrary payment reduction. The third reduction, while
not intended as a reduction, is related to decreasing the geographic payment
disparity between high and low cost counties which affects counties where the
majority of beneficiaries reside. The fourth reduction is reduced payments to
FFS providers - an indirect reduction -- and the fifth reduction, also not
intended to be a reduction, is the unfair imposition of a "user fee" to cover
the cost of beneficiary education materials for all beneficiaries - not just
those in managed care. Finally, HCFA has proposed a risk
adjustment implementation that further reduces payments to
Medicare+Choice contractors, rather than using risk adjustment
to allocate proper funding to health plans that enroll sicker members. We
strongly recommend that risk adjustment be implemented on a
budget-neutral basis to avoid "double jeopardy" of multiple payment reductions.
The other major issue is the bias against managed care health plans through
use of HCFA's version of the PIP-DCG method. There are several areas where this
bias will have an effect on payments. The following are examples: -- HCFA's
elimination of "short stay" admissions from the PIP-DCG model payments. Humana,
like most Medicare+Choice contractors has successfully reduced the length of
stay at acute facilities. The elimination from payment "scoring" of hospital
stays of less than two days penalizes health plans that have reduced hospital
costs. For Humana's senior and disabled members, 22% of hospital visits were in
the "short stay" category, based on a recent study. While many of these stays
will be for less serious conditions, the effect of the elimination was
approximately a 1.5% reduction in payment. In contrast, there is much less
incentive in Medicare FFS to achieve significant reductions in length of stay
given Medicare's requirement of a 3-day stay prior to discharge to a sub-acute
care facility. -- HCFA's inclusion in PIP-DCG payments of certain conditions
that are more commonly treated in FFS medicine by an inpatient admission. There
is a wide variation in treatment practice across the U.S. Wennberg, J.E.,
Freeman, J.L., Shelton, R.M., and Bubolz, T.A., "Hospital Use and Mortality
among Medicare Beneficiaries in Boston and New Haven," New England Journal of
Medicine, 321(17):1168-73, October 26, 1989. and great efforts by health plans
to appropriately treat members in the lowest cost setting. Since this setting is
more frequently an outpatient clinic or physician office, conditions that are
"site-of-treatment discretionary" should be moved to the "base" payment
category, so health plans are not penalized (by failing to trigger additional
payments associated with a PIP-DCG group) through shifting patients to these
less expensive ambulatory sites. Using definitions of discretionary conditions
from an older study by members of the DCG research team, Ellis, R., and Ash, A.
"Refining the Diagnostic Cost Group Model: A Proposed Modification to the AAPCC
for HMO Reimbursement," February 1988, report prepared for the Health Care
Financing Administration. our consultants found that keeping these discretionary
conditions in the PIP-DCG model could reduce payments to health plans by 1% to
3%. -- A perverse incentive created by use of only inpatient admissions rather
than more complete diagnostic data. Humana recognizes the practical need to
begin risk adjustment with only inpatient data. While pragmatic
issues may mandate the use of an inpatient data method at the start, health
plans should not be penalized at every decision point. If some of the biases can
be corrected, then health plans will be paid more appropriately for providing
care in a cost effect manner.
We would point out that this practical
approach penalizes the "good deeds" that health plans accomplish, such as
preventing heart conditions. For example, Humana has over 3,900 members in a
disease management program to prevent or reduce Congestive Heart Failure (CHF).
Our specialists estimate that 60% of admissions linked to CHF can be eliminated
- therefore, we hope to eliminate all admissions next year for 2,300 of the
4,000 seniors in the program. However, if all these members are Medicare+Choice
members, we will then lose about $12,000 per admission by not triggering the
additional payment for PIP-DCG 16 - for a total of about $28 million. On a
per-person basis in the region affected, reimbursement would be reduced by about
$100 per member per year. We may need to reduce members' prescription drug
benefits by nearly 33% to offset this revenue loss.
Possible effects of
inappropriate implementation of risk adjustment on our
Medicare+Choice enrollees could be significant. While the phase-in reduces
concerns in the first year, the amount of payment reduction that may be
incorrect in the second year is frequently greater than a health plan's entire
profit/surplus margin. One or more consequences will occur: health plans will
reduce supplemental benefits, premiums will be charged or increased, or health
plans will exit counties that are currently marginal or difficult markets. The
recent study in JAMA about "spillover" effects of Managed Care points out the
savings that would be eliminated from FFS Medicare without the beneficial
presence of health plans. Baker, L. "Association of Managed Care Market Share
and Health Expenditures for Fee-for- Service Medicare Patients," Journal of the
American Medical Association, 1999, 281: 432-437.
Administrative Cost
Concerns The cost of adapting to risk adjustment is just one of
many administrative costs resulting from the Balanced Budget Act (BBA). Although
Humana has begun to estimate these costs, we continue to collect and submit new
data, adapt our rating and budget processes, revise provider contracts,
products, enrollment systems and communication materials, collect data for newly
mandated clinical studies and train providers and staff. To provide some idea of
the magnitude of the administrative costs, we turn to the issue of provider
contracting. Humana has over 128,000 separate providers of all types -
physicians, hospitals, labs, DME vendors, etc. Because of BBA changes, we are in
the process of re-contracting with a significant number of those providers -
providers in whose offices we are concurrently auditing medical records for the
purposes of HEDIS and clinical studies and securing data to meet regulatory
requirements for physician incentive arrangements. The average cost of just re-
contracting runs about $34 per hour and 3 hours of effort, for an average
contract cost of roughly $100. Including the costs for drafting and regulatory
agency filing of contract forms, we estimate provider re-contracting costs will
exceed $3.2 million.
Beyond risk adjustment, the HCFA's BBA
regulations have imposed extensive new requirements for oversight, additional
clinical studies for quality measurement, and other compliance requirements. As
an example, the two required Medicare clinical studies for each Humana
Medicare+Choice plan requires an expenditure of $75,000 per plan for data
collection alone. Current accreditation costs for Humana requires an expense of
between $300,000 and $1.5 million.
Request for Deferral of Risk
Adjustment
We would ask that risk adjustment using
inpatient data for the Medicare+Choice program be deferred for at least one year
due to lack of disclosure of necessary methodological and formula-related
information to plans, data collection issues, known and unknown HCFA and Fiscal
Intermediary systems issues and the potential adverse effects this payment
method could have on beneficiaries and health plans. To date, health plans have
submitted detailed claims data to HCFA and HCFA is in the process of analyzing
those data. As you may know, due to provisions in the Balanced Budget Act of
1997, Medicare payments to health plans have increased in many counties by only
2% in the previous two years, a level below the cost increases that many health
plans have experienced. As a result, increased payment uncertainty due to a new
risk adjustment system will exacerbate what is already a
difficult situation for many health plans. Some plans have already decided to
discontinue participation in the M+C program in one or more counties. In spite
of the phase-in of risk adjusters, it is likely more plans will go this route in
the next two years if the PIP-DCG risk adjustment system is
implemented on the current schedule.
As an alternative to implementation of
the proposed system on January 1, 2000, we suggest that over the next year, HCFA
continue to analyze data submitted by the health plans and conduct a simulated
risk adjustment. This would allow hospitals, health plans, HCFA
and its contractors enough time to improve their data reporting systems so as to
ensure proper payment. A similar timeframe for implementation was used in
California for the Health Insurance Plan of California (HIPC), a small-group
purchasing cooperative. Bertko, J. and Hunt, S. "Case Study: The Health
Insurance Plan of California." We strongly believe that a simulation of
risk adjustment over the next year would provide invaluable
information to participating health plans and to HCFA.
Conclusion
Humana
supports the move towards a risk-adjusted payment system but only after several
key risk adjustment method issues are resolved and data
collection processes are improved significantly. We urge you and HCFA to defer
implementation of the new risk adjustment system for at least
one year to allow this improvement. We understand that HCFA faced a daunting
schedule and believe that the Agency could do more to disclose relevant
information to health plans if it had more time. As an alternative to
implementing the proposed system on January 1, 2000, we suggest that over the
next year, HCFA continue to gather and analyze data submitted by the health
plans, conduct a simulated risk adjustment, and work towards
improving the entire data process. This would allow all parties sufficient time
to improve their data reporting systems so as to ensure proper payment. We
strongly believe the additional time would allow health plans and HCFA to obtain
valuable experience and information, allowing all to have greater faith that the
new risk adjustment system is an improvement and is being
implemented correctly.
This concludes my prepared testimony. I would be
happy to answer any questions you may have.
END
LOAD-DATE: February 27, 1999