Wednesday, October 11, 2000
Thursday, October 12, 2000
Members of the Technical Panel:
Dale Yamamoto, Chair of the Panel
Center for Health Systems Research and Analysis
University of Wisconsin-Madison
Department of Health Management and Policy
University of Michigan
Ann Arbor, MI
Professor of Economics
Department of Economics
Principal Research Associate
Health Policy Center
The Urban Institute
Consultant to the Panel:
Director and Consulting Actuary
Price Waterhouse Coopers
Staff of the Health Care Financing Administration, Office of the Actuary:
Richard Foster, Chief Actuary
Sol Mussey, Director of the Medicare and Medicaid Cost Estimates Group
John Shatto, Actuary
John Wandishin, Actuary
W. Kent Clemens, Actuary
Mark Freeland, Deputy Director, National Health Statistics Group
Stephen Heffler, Deputy Director, National Health Statistics Group
Jackie Carroll, Executive Officer
Donna Holt, Secretary
Staff of the Office of the Assistant Secretary for Planning and Evaluation:
Christy Schmidt, Deputy to the Deputy Assistant Secretary for Health Policy,
Ariel Winter, Executive Director of the Panel, ASPE, DHHS
Gene Moyer, Consultant to ASPE and to the Panel, DHHS
Henry Krakauer, ASPE, DHHS
George Greenberg, ASPE, DHHS
Social Security Administration
Phil Ellis, Economist,
General Accounting Office
Congressional Budget Office
University of Maryland
Proceedings of the Meeting:
The meeting was called to order by Mr. Winter, who welcomed the group and thanked them for all their hard work. He pointed out that Mr. Yamamoto would be asking them to approve the minutes from the July 26-27 meeting and that once they were approved, the minutes would be available for members of the public and staff.
He asked that the members sign the five letters he had drafted thanking individuals who made presentations to the Panel at the last meeting in September.
He recognized Mr. Foster for some remarks. Mr. Foster also expressed his appreciation for the panel's hard work.
Mr. Winter then turned the meeting over to Mr. Yamamoto. Mr. Yamamoto asked for and received from Mr. Robinson a motion to approve the minutes from the July meeting. Mr. Chernew seconded the motion. The motion was unanimously approved.
Mr. Yamamoto then turned to Mr. Foster for a discussion on the Office of the Actuary's projections of managed care.
In response to some questions by Mr. Robinson and Mr. Cutler, Mr. Foster said that the Office of the Actuary was doing a simulation in which the managed care population was assumed to be stable to see how much difference that would make in the HI projections.
There was a discussion of the adjustment for biased selection in managed care, which is done for HI, but not for SMI. As healthier persons select into managed care, the remaining sicker persons should cost SMI more. Mr. Foster suggested some papers about the topic from various sources and felt that this was an area deserving further investigation.
Mr. Gutterman and Mr. Chernew also brought up the geographic differences in enrollment in managed care and asked whether there needed to be a geographic adjustment or not.
Mr. Foster suggested that the House Ways and Means Committee's pending legislation to boost managed care plan payment rates may provide a natural experiment to allow a test of whether a geographic adjustment is needed.
Mr. Robinson mentioned that the current managed care enrollment projections by age and sex grow at the same rate across cells, which causes varying managed care penetration rates by age/sex group. Mr. Foster said that they could probably investigate the changes in the last few years to see how different the managed care growth rates by age/sex group have been.
Mr. Wandishin said that the little evidence they have suggested that the older age groups in managed care were growing faster than the younger age groups. Therefore, leaving more older beneficiaries in fee for service Medicare makes the projections conservative (i.e., overestimates costs) because fee for service beneficiaries cost more per capita.
Mr. Robinson volunteered to lay out a methodology and to suggest a recommendation on the subject.
Mr. Yamamoto then turned the meeting over to Mr. Cutler to discuss the subgroup 1 draft recommendations.
Subgroup 1 had five recommendations. These were
1. The committee recommends that forecasting continue to be done in three intervals: the first ten years; years 10-25; and years 25-75, with the middle 15 years being a transition between the first 10 years and the latter 50 years. [The committee recommends that after 25 years, Medicare cost growth be considered as a whole, rather than separate forecasts for parts A and B.]
2. The committee believes that the forecast of market basket changes is reasonable.
3. The committee suggests viewing short term [and long-term] forecasts in a two part framework stressing growth rates and levels separately. The growth rate component is due to technological change, which the committee recommends setting at 1 percent annually above the rate of growth of real per capita GDP. Level effects include payment and other reforms that do not have impacts on the long-term growth of costs.
3a) The technology component need not be neutral across service categories -eg, the difference between minimally invasive bypass surgery (inpatient) and modern cataract lens extraction (outpatient). [The committee recommends some trend in the next decade towards less inpatient care and more outpatient care. The appropriate magnitude of such effects could be estimated through studies of several diseases to determine how technological change in treatment has affeted utilization of different service categories.]
3b) The growth rate should apply after age and sex adjustment.
3c) Level effects result from several factors: (i) gaming associated with payment system changes, as with the new prospective payment systems for home health, SNF, and outpatient hospital; (ii) fraud and abuse control activities; (iii) particularly large changes in payments for a service.
3d) The committee believes that transitory utilization changes are partially corrected over time. Some of this is already forecast. For example, lower home health visits in 1998 and 1999 are projected to be offset by rapid growth in the early part of this decade. In other cases, this is not forecast. For example, below average case mix changes for inpatient hospitals in 1998 and 1999 is not matched by forecasts of above long-run growth in case mix in the early part of this decade.
3e) The committee recommends monitoring carefully up coding in response to new prospective payment systems for home health, skilled nursing facilities, and outpatient hospitals.
4. The committee recommends that several additional studies by undertaken along the lines of the RAND study of up coding, and that data systems be set up to monitor quarterly trends in key variables. The goals of the study and data gathering would be to decompose the factors leading to changes in these variable. In particular, the committee recommends consideration of the following areas:
4a) Utilization and case mix changes in outpatient hospital, home health, and skilled nursing facilities as a resuly of new prospective payment systems and cuts under BBA97
4b) Inpatient hospital case mix changes
Each of the recommendations and subrecommendations was discussed in turn, but no decisions were made at this meeting. There were some suggestions that the recommendations include more specifics. Mr. Nichols asked if the Panel would vote on the recommendations as they were discussed, but the Panel decided to look at all the recommendations from all groups, meld those which seemed similar, and then to vote on the entire list.
At this point, the panel took a five minute break.
After the break, Mr. Foster described some simulations done by the University of Maryland's LIFT model. They allowed productivity to vary in order to allow the other assumptions to be the same as those of the Trustees. In order for the alternative II assumptions to be met and for growth to be equal to that assumed, productivity had to "skyrocket." Health care eventually reached about 31% of GDP.
Mr. Chernew suggested that this discussion be held over until the subgroup 3 discussion right after lunch. He felt it was closely related to the subgroup 3 recommendations.
Mr. Yamamoto then turned the meeting over to Mr. Robinson and Mr. Chernew for a discussion of the subgroup 2 recommendations.
This subgroup had the following recommendations:
1. Assumptions regarding shifts in site of care should be explicit and consistent across models. This is especially important for shifts from inpatient (HI) to outpatient care (SMI).
2. For the short run forecast, the quantity and price components of spending for each service category should be modeled explicitly, with total spending by category being computed as the product of a price component and the quantity component.
3. Short run models of utilization should be based on age and gender utilization patterns for all service categories.
4. Within each age/gender cell, distinctions among spending for decedents and survivors should be made.
5. Managed care enrollment, and its impact on FFS utilization and case-mix, should be treated consistently across the models.
6. The Office of the Actuary should coordinate revisions to the projection models with current and future research projects. Where possible, models such as those being developed by RAND should be considered for incorporation into the projection methodology. In other cases, further investment in the construction of these models might be warranted.
7. The model documentation should be made more explicit regarding key underlying assumptions and the reasons for various adjustments (e.g. legislation) to historical trends.
These recommendations again were discussed in turn.. Mr. Chernew said that the subgroup and the entire Panel had not given much thought to the question of how much better the numbers would be if the recommendations were followed.
Mr. Foster thought this was important. If there was an error which a recommendation sought to correct, that recommendation should be flagged. If the recommendation basically agreed with what the Actuaries are doing, it would also be nice to have it on the record. Some of the recommendations, however, may be difficult to implement in the near future.
There were suggestions about combining some of the recommendations and considerable discussion of providing priority measures in the list so that the actuaries would have a guide about which should be implemented first.
At 1:00, Mr. Yamamoto adjourned the Panel for lunch.
At 2:00, Mr. Yamamoto called the Panel to order.
Mr. Robinson said that in the earlier discussion the Panel had agreed not to put into the recommendation for changes to the model any kind of "escape clause" for testing whether or not the model works adequately or not.
Mr. Robinson and Mr. Chernew then discussed the remaining two recommendations. The first of these had to do again with the projection of managed care and the selection issue. Mr. Chernew and Mr. Robinson agreed to recast this recommendation taking into account some of the suggestions made by the Panel.
Finally, everyone seemed to agree with the recommendation for additional documentation of the model.
Then Mr. Yamamoto moved on to the Long-range Growth Assumptions subgroup: Messrs. Gutterman, Cutler, and Chernew.
This subgroup had only one recommendation: the real age-adjusted per beneficiary expenditure for both HI and SMI should be assumed to grow at 1 percentage point above the real per capita GDP growth.
Mr. Gutterman asked whether the base rate should be real GDP or real wages. Mr. Chernew said that the subgroup would consider the question.
Mr. Chernew then went through the analyses which led to the recommendation.
Finally there was a kind of vote of the Panel and the audience, which, with small reservations, unanimously supported the recommendation.
There was then a discussion of procedure for the rest of the day and the next day. Then the group took a short break.
Upon reconvening, Mr. Yamamoto turned the meeting over to Messrs. Gutterman, Robinson, and Nichols for the Uncertainty/Stochastic Forecasting group's recommendations.
Rather than recommendations, this group had a set of benefits and concerns about stochastic modeling. Several of these have yet to be written up, so Mr. Robinson went over what they had done.
Mr. Cutler asked if recommendations could come out of the list. Mr. Gutterman and Mr. Nichols responded that they had not been able to reach consensus on recommendations. Largely, it seems, they could not reach consensus because they did not feel that a stochastic model could replicate alternative II and they could not quite agree about what alternative II represented in the stochastic model.
They passed out the following six draft recommendations, but said that they had not agreed on them:
1. The development of the stochastic model for SMI included in Appendix C of the 2000 TR was on the whole successful. The Panel recommends that
a) This model should be further developed and enhanced, and
b) Other stochastic models should be developed; both HI and SMI.
The Panel notes that at the current time the results of such models be supplementary rather than eliminate other means of simulating uncertainty of the programs.
2. On a regular basis, identify sources and causes of changes in historical health care costs.
3. On a regular basis, evaluate and report on the accuracy of past HI and SMI projections, over a 1, 6, and 10-year timeframe, including a decomposition of the sources of deviation of experience from their projections.
4. A new set of indicators should be developed, focusing on sources of uncertainty and sensitivity of health care cost projections under alternative scenarios and the degree of affordability of costs of SMI, including:
a) Alternative managed care experience,
b) Over a 75 year period, comparisons of projected SMI premiums to representative or, average OASI benefits,
c) SMI general revenue financing to [not specified]
d) [not specified]
5. To avoid confusion in the HI long range test of close actuarial balance, either eliminate or relate the actuarial balance under the modified average-cost method now shown in appendix A of the TR.
6. Regarding alternatives I and III, [not specified]
While they agreed that a stochastic model was beneficial, they could not decide how it should be utilized in the Trustees Report.
Mr. Foster asked if this meant that they considered the work in the SMI report was "barking up the wrong tree." They assured him that they did not, but they felt that one major fault of the model the Chief Actuary was using was that it did not contain a feedback loop.
Mr. Gutterman then said he felt that the group had one finding and one recommendation. The finding was that the work so far is commendable in terms of its application. The recommendation is that the work should be enhanced to take into account some of the caveats the subgroup had mentioned. It should also be expanded into the HI model, but not necessarily to replace anything in either model. There seemed to be general agreement with this.
Mr. Gutterman said that there was a lot of work to do in writing this up. He also discussed three other recommendations concerning continuing research the subgroup felt the Office of the Actuary should do on the sources and causes of changes in health care costs, evaluating and reporting on the accuracy of past projections, sources of uncertainty and sensitivity of the projections under alternative assumptions, and the degree of affordability of SMI.
Mr. Gutterman also discussed the actuarial balance computation and brought up the question of why there are two such computations in the HI Report. Mr. Foster explained that the Trustees decided to use the discounted values in the report, but that the undiscounted values were put into an appendix to satisfy the HCFA(now known as CMS) actuaries.
Mr. Yamamoto asked the Panel if they thought the appendix reporting the alternative actuarial balance measure should be removed, but no one felt strongly that it should be and so they decided not to recommend that at this time.
Finally, the subgroup wanted to suggest that there be a similar actuarial balance computation in the SMI report, but could not decide what to recommend. Mr. Foster suggested that there was a danger that this might indicate that the program was out of actuarial balance. It is not by definition. After considerable discussion, the Panel decided to discuss it over dinner.
Mr. Yamamoto adjourned the meeting a 6:06 P. M until morning.
Mr. Yamamoto resumed the meeting a little after 9 and turned the meeting back to Mr. Gutterman for an extra 10 minutes.
The main remaining topic was a review of alternatives I and III.
The major question concerned consistency between HI and SMI.
Mr. Foster said that historically they have not been done consistently for reasons that are not exactly clear. While it would be nice to be able to do them consistently, it is not clear how to do so. While the actuaries may get there some time, they are not there now.
Two other consistency questions were brought up by Mr. Gutterman. One concerns the lengths of the short and long terms. The second concerns whether the 2 percentage point difference is more appropriate than the alternative assumption approach used for OASDI.
There was considerable discussion of the first issue, especially concerning the relative uncertainty of the short and long terms. There were questions about whether the interval between I and III represented in some senses a sensitivity analysis. Mr. Robinson suggested that stochastic modeling might be of help in characterizing the interval.
Mr. Gutterman wrote a proposed recommendation. While it was being typed up, the Panel moved on to Mr. Nichols and Report Presentation Issues/ Research and Data.
There were five recommendations. The first concerned allocating more resources to providing more current data. Mr. Gutterman suggested that it be merged with the stochastic subgroups recommendation on the same subject.
The second concerned adding additional actuaries to the office, providing a larger research budget to the Chief Actuary, and that there be more coordination of the research already being done by various Medicare analysts.
The third was a catch-all recommendation concerning consistency between the HI and SMI reports, a plea for additional clarity, and a suggestion that more emphasis be placed on cost per beneficiary in both reports.
The fourth recommendation concerned a long term comparison of past forecasts with actual experience for one year, five year, and ten year intervals. In the discussion there was concern that legislation in the intervening years would render the comparisons meaningless. Mr. Robinson said he felt this could be handled listing some factors representing the reasons the projections did not work out.
Mr. Gutterman brought up something that had been discussed previously, but which he felt needed to be in the presentation recommendations: More sensitivity tests need to be included in the report along with additional information to explain the tests.
At this point, Mr. Yamamoto called for a break.
After the break, the Panel listed the recommendations and findings with Mr. Winter acting as recorder.
Some were to be listed as findings rather than recommendations. The Panel had several votes about which items were to be included as it went through the list of subgroups and their recommendations.
The agreed-upon list of recommendations and findings follows, organized by chapter:
0-1: All assumptions but one are reasonable. Some that are reasonable could still be improved. The long-run expenditure growth assumption is unreasonable.
0-2: The overall projection methodology is reasonable, other than specific areas discussed in section II.
The Panel believes that the forecast of market basket changes is reasonable.
I-2: The Panel recommends setting the technological component of the growth rate at 1% above real per capita GDP.
I-3: The Panel recommends some trend in the next decade towards less inpatient care and more outpatient care.
I-4: The Panel recommends that, as a general principle, forecasting recognize that transitory utilization or case mix changes are partially offset over time.
II-1: Trustees should make consistent the function of projection periods between HI and SMI; intermediate term should be transition between short and long term with more aggregation than short term and more detail than long term projections.
II-2: The Panel recommends viewing short term [and long-term?] forecasts in a two part framework stressing growth effects and periodic shifts separately.
II-3: The Panel recommends that assumptions regarding shifts in site of care should be explicit and consistent across models. This is especially important for shifts between HI and SMI, such as movements from inpatient to outpatient care settings.
II-4: The Panel recommends that for the short run forecast, the quantity and price components of spending for each service category should be modeled explicitly, with total spending by category being computed as the product of a price component and the quantity component.
II-5: The Panel recommends that models of utilization should be based on age and gender utilization patterns for all service categories
II-6: The Panel recommends that within each age/gender cell, distinctions among spending for decedents and survivors should be considered.
II-7: The Panel recommends that explicit health status models should be considered for eventual incorporation into the projection methodology.
II-8: [placeholder for recommendations on managed care growth and managed care shift effect]
II-9: The Panel recommends that the model documentation should be made more explicit regarding key underlying assumptions and the reasons for various adjustments (e.g. legislation) to historical trends.
III-1: The Panel recommends that real, age adjusted, per beneficiary expenditures for both SMI and HI should be assumed to grow at a rate 1 percentage point above real, per capita GDP growth.
IV-1: The Panel recommends that the stochastic model used in the 2000 SMI Report should be further developed and enhanced and other stochastic models should be developed for SMI and HI.
IV-2: The Panel recommends that, on a regular basis and to the extent possible, the Trustees should evaluate and report on the accuracy of past HI and SMI projections over a 1, 5, and 10-year timeframe, including a decomposition of the sources of deviation of experience from their projections.
IV-3: The Panel recommends that a new set of indicators should be developed that focus on sources of uncertainty and sensitivity of health care projections to underlying factors, including managed care, under alternative scenarios.
IV-4: The Panel recommends that a set of indicators should be enhanced to assess the implications of estimated SMI cost growth.
IV-5: The Panel recommends that the range between alternatives I and III should be accordion-shaped.
V-1: The Panel recommends that more resources should be devoted to insuring that quarterly cost and use data are available with a shorter time lag for analysis by the Office of the Actuary (OACT).
V-2: The Panel recommends that more resources should be devoted to serving the analytic research priorities of the OACT and the Trustees. Without being prescriptive, this increased resource commitment should likely include both additional staff inside OACT as well as some targeted reallocation of research funds outside OACT that can address Trustee Report-related questions of paramount importance.
V-3: The Panel recommends that there should be greater coordination of relevant research projects and agendas across the many agencies of the Department (AHRQ, HRSA, CDC, ASPE, and HCFA(now known as CMS) itself) and between the Department, the congressional support agencies (MEDPAC, GAO, CRS, CBO), and private organizations.
V-4: The Panel recommends that the clarity of the Reports' presentations should be enhanced considerably with a number of specific changes and additions. They include:
A. Emphasize cost per beneficiary, or cost per worker, rather than cost per unit of service (which is harder to define and is often left obscure).
B. Both the HI and SMI reports should have a section with consolidated reporting of HI dollars per beneficiary, SMI dollars per beneficiary, and total dollars per beneficiary.
C. The discussion of some key tables -- II.F.1 in each report is a salient example -- should include an explanation of how the columns relate to (and may be derived from) each other.
Having written final draft recommendations, the Panel spent some time discussing next steps. They agreed that a November 15th meeting was necessary.
They further agreed that each subgroup head write their section of the final report by November 6 and send it around to the entire group for comment. Mr. Yamamoto would write an introduction/overview chapter.
By November 10, comments by other Panel members are to be sent to the entire Panel.
By November 13, final drafts of each chapter are to be sent to Mr. Winter, who will compile them into a draft report that he will distribute to the Panel.
At the November 15 meeting, the Panel will discuss recommendations related to managed care assumptions and finalize its report.
At 12:34 P. M., Mr. Yamamoto adjourned the meeting.