Copyright 1999 Federal News Service, Inc.
Federal News Service
FEBRUARY 25, 1999, THURSDAY
SECTION: IN THE NEWS
LENGTH:
1677 words
HEADLINE: PREPARED TESTIMONY BY
JOHN
BERTKO FSA, MAAA
PRINCIPAL
REDEN & ANDERS, LTD
BEFORE THE
HOUSE COMMERCE COMMITTEE
SUBCOMMITTEE ON HEALTH AND
ENVIRONMENT,
SUBJECT - RISK ADJUSTMENT'S IMPACT ON HEALTH
PLANS
BODY:
SUMMARY:
- Reden & Anders,
Ltd., an actuarial consulting and data analysis firm, has extensive experience
in risk adjustment and works with 15 large Medicare+Choice
contractors, investigating the effects of HCFA's proposed PIP-DCG risk adjuster
on payments to health plans. Based on that experience, I would like to make the
following observations:
- Health-based risk adjustment,
such as the PIP-DCG method, represents an improvement over the current age, sex
and status-based AAPCC payment method, but should only be implemented after all
outstanding issues are resolved.
= Although practical considerations require
initial use of an inpatient-data method, HCFA should re-examine several issues
that lead to payment bias against health plans.
- There is a need for HCFA
to provide full disclosure of all of the formulas of the method, as well as
access to beneficiary data on status and health conditions so that health plans
can replicate HCFA's results.
- Outstanding data collection and analysis
issues must be resolved before the risk adjustment payment
process begins; if these problems are not addressed, health plans participating
in the Medicare+Choice program would be underpaid.
TESTIMONY:
Good
Morning, Mr. Chairman and Members of the Committee. I appreciate the opportunity
to provide you with information about my experience with risk
adjustment for health plans using the Principal Inpatient Diagnostic
Cost Group (or PIP-DCG) method. I am a consulting actuary and Principal with
Reden & Anders, Ltd., an actuarial consulting and data analysis firm, which
is part of Ingenix, the information division of United Health Group. Reden &
Anders' risk adjustment clients include 15 large
Medicare+Choice contractors with operations in nearly 40 diverse markets. Over
the past year, we have been analyzing data from our clients, in an effort to
help them better understand how their payments for Medicare enrollees will
change as a result of the implementation of a risk adjustment
system. I will be drawing from this experience and providing an actuary's
perspective on the effects of and issues associated with the proposed PIP-DCG
risk adjuster.
Risk Adjusted Payments Are an Improvement over the Current
Method
In my opinion, risk adjusted payments using diagnostic data represent
a step forward in making appropriate payments to Medicare+Choice contractors.
Technically, a payment system using health risk adjusters is somewhat more
accurate than the current payment system. This is especially true for groups of
individuals with more health problems - inmost cases, health plans will spend
more to treat these individuals and, as a result, deserve higher payments.
Similarly, for groups of healthy individuals, risk-adjusted payments to health
plans will be appropriately reduced. Use of risk adjustment
also helps to achieve the policy goal of better matching payment to the needs of
a covered population.
In any diagnosis-based risk
adjustment method, each individual is assigned a" relative risk score"
based on his or her past illness history. In comparison with the average risk
score of 1.00 for an entire population (both sick and healthy individuals),
someone with a heart condition may have a relative risk score of 3.0, which
means we expect the person to have expenditures that are three times the average
next year. By contrast, a healthy 70 year-old male may have a risk score of .70,
meaning that we would expect his expenditures for the next year to be only about
70% of the average of the group. Under the PIP-DCG method, most enrollees (i.e.,
those who do not have a qualifying inpatient stay) will be assigned to the
"healthy" (base) category. This means that health plans will receive a base
payment amount (approximately $5,000 for these individuals). The base payment
varies by age and gender, as well as by disability, welfare, and working aged
status. HCFA will analyze each enrollee's medical encounters over the previous
year to determine whether the enrollee had an admission that falls into one of
the 15 PIP-DCG categories with increased payments. An inpatient admission
creates a PIP-DCG score that is worth from $2,000 to $26,000 per year in
additional payments.
Because the PIP-DCG method relies only on inpatient
data, it should be considered a first step, or a "Work in Progress," but an
important improvement over the current method which uses only age, gender, and
status (disability, institutionalization, welfare or Working Aged status) for
payments. Other methods that use data from both the inpatient and outpatient
settings ("full" data methods) are under development and are being used in some
settings (e.g., by the Buyers Health Care Action Group in Minneapolis). HCFA has
indicated that it plans to use both inpatient and outpatient data for
risk adjustment beginning in 2004.
Analysts and actuaries
recognize the bias inherent in a payment method that uses only inpatient data,
since Managed Care plans are penalized for keeping members out of the hospital.
However, an inpatient data- based method is the only practical first step and
should be acceptable, if implemented with care. The inpatient method should also
be thought of as a natural transition to a method that uses both inpatient and
outpatient data. This is because the relative risk scores that are calculated
using only inpatient data are, generally, closer to the average for the group
(1.00) than risk scores calculated using "full" data methods, thus reducing the
effects on payment to health plans. On the other hand, health plans that
successfully treat high-cost conditions in an ambulatory setting will be
penalized, since no additional payment will be provided for high-cost
individuals who have not had an admission.
PIP-DCG Model Still Needs
Refinement
While I believe that the PIP-DCG model overall is well-designed
and has been tested extensively on the Medicare population, there are several
components of the model that should be re-examined. As noted above, any
inpatient-data risk adjuster will be biased against Managed Care plans because
of their ability to eliminate some inpatient admissions. Although this
circumstance must be accepted during a transition phase, other components of the
HCFA model add to this bias. As one example, HCFA has chosen to eliminate "short
stays" or admissions with a length of stay of less than two days from
contributing to an enrollee's risk score. Thus, if a person with a heart
condition is treated during one day and then sent to a Skilled Nursing Facility,
the health plan will receive only the base payment for that enrollee, since the
enrollee did not have a qualifying inpatient admission. In FFS Medicare, it is
much more likely that this person will be hospitalized two days or more, adding
to the payment bias. Similarly, there are other conditions that can be treated
either on an inpatient basis or an outpatient basis (in a physician office or
clinic).
Again, for any treatment that occurs in a non-hospital setting,
health plans will not receive any payment over the base amount. The result would
be that the relative risk scores for enrollees in Managed Care plans would be
lower than the relative risk scores of Medicare FFS enrollees. One way to
address this problem would be for HCFA to remove some of these conditions from
the PIP-DCG model.
HCFA Needs to Disclose Data and Methods
One of an
actuary's professional requirements is that he or she be able to replicate the
findings of another actuary's work. To date, we have analyzed data for our
clients, using our "best guess" regarding the various formulas and components of
the PIP-DCG Model. Therefore, at this point, I am unable to perform a thorough
replication. I strongly recommend that HCFA provide full disclosure and have
open discussions with health plans, consultants, academics, and other interested
parties. This involves disclosing all of the formulas used for every component
of the model, as well as providing access to data in HCFA's files about
beneficiary status and health conditions so that health plans can test and fully
understand the model's operation.
Implementation Issues
Implementation
of the PIP-DCG payment method has required the creation of a new data
transmittal and analysis process. As part of this process, encounter data must
be handed from hospitals to health plans to Fiscal Intermediaries to HCFA or its
contractor. If the "data baton" is dropped anywhere along the way, then health
plans are penalized automatically through lower payments. For example, sometimes
hospitals make mistakes regarding an inpatient admission and attribute it to the
wrong health plan. We have heard reports of hospitals delaying correction of
errors, with the result being that health plans cannot submit corrected
encounter records. Some plans may have difficulty gathering data from capitated
providers who pay their own claims. Many of the Fiscal Intermediaries have
awkward or slow processes in place to identify and report errors. As a result,
health plans can spend an inordinate amount of time trying to fix errors. If the
Fiscal Intermediaries were more forthcoming with what triggered errors, the
health plans could correct the errors before submitting data to the Fiscal
Intermediaries. HCFA has at least a few small problems in its systems for making
use of encounter data. With the lack of sufficient feedback on the data errors,
health plans are unable to confirm that data being used by HCFA matches their
internal records.
Summary
Risk-adjusted payments represent an
improvement over the current AAPCC payment method, but only if practical data
issues are first addressed and several components of HCFA's PIP-DCG method are
re-examined. HCFA staff are to be commended for their hard work in moving
towards implementation on such a demanding timetable. I suggest that
implementation not occur, however, until HCFA and health plans are certain that
all biases are removed from the model and important data process issues are
corrected.
END
LOAD-DATE: February 27, 1999
