Extent Of Coverage And Circularity Of Estimates

4. EXTENT OF COVERAGE AND CIRCULARITY OF ESTIMATES PRI/PwC: PRI/PwC estimates “that the Medicare population’s share of total drug spending covered by insurance will increase to 70 percent, compared to 58 percent in the absence of the law.” (p. 4). The implication is that 42 percent of drug spending would be out-of-pocket in 2006 absent the law, and this would fall to 30 percent under the new law. REPLY: The PRI/PwC critique appears to rest heavily on a circular argument. And unsubstantiated, highly questionable estimates are used for the insured and out-ofpocket shares of drug costs. a. Circularity of volume estimates By way of documentation, the authors offer a remarkably uninformative footnote that says they calculated what the text says they calculated. No methods, calculations, or evidence regarding supporting assumptions are provided. The 58 percent/70 percent calculation appears to rest in part on the assumption of the minus 0.3 price-elasticity of demand. The PRI/PwC’s assumed price-elasticity of demand and its 58 percent/70 percent estimate are its reasons for saying that the overwhelming bulk of the new $400 billion Medicare prescription drug program would go to replacing existing drug spending by patients or by their insurers, consequently doing very little to buy new medications for patients who today need them but can’t afford them. While this would seem to be a clear conclusion of the PRI/PwC critique, that critique does not ever estimate a split in spending between new prescriptions and replacement prescriptions. The PRI/PwC critique (p. 6) goes on to use its own 58 percent/70 percent estimate to complain that our calculations rest on a very high price-elasticity of demand of 1.3. They try to back us into that corner through their own unrealistic assumptions about priceelasticity of demand and their own entirely undocumented 58 percent/70 percent estimates of the shares of medications covered by insurance before and after implementation of the new Medicare drug benefit. In this way, the PRI/PwC critique erects its own world of interlocking or circular assumptions, with the appearance of empirical support, and uses them to say we are wrong. They imagine an unrealistic world and say we are wrong because we refuse to inhabit it with them. They could have said, “we think they are wrong.” That would have been fair. But when they claim that they are relying on science to argue we are wrong, they badly overreach. b. Unfounded estimates of covered and out-of-pocket shares Moreover, the PRI/PwC’s assumption that 58 percent of Medicare beneficiaries’ drug spending will be covered by insurance before the new Medicare prescription drug benefit kicks in, in 2006, simply does not square with reality. (PRI/PwC assert, in note ix, that this was “calculated …by first estimating the average amount of drug spending covered by different types of insurance…,” but no underlying data whatsoever are shown.) CBO had similarly estimated that only 40 percent of drug spending for Medicare beneficiaries was paid out-of-pocket in 1999.33 This might seem consistent with—and may even have rested in part on—CBO’s apparent occasional use of an inaccurate estimate that only 25 percent of Medicare beneficiaries had no drug coverage in 1999. But as discussed elsewhere in this response, that figure reflects an estimate of the share of beneficiaries lacking insurance for drugs for an entire year, not the average share of beneficiaries without insurance at a given time. The latter is the better measure of insurance coverage and the lack of it, we believe. And even if only 40 percent of drug spending for Medicare beneficiaries was out-ofpocket in 1999, that is certainly not true today, and it predictably will be less true in 2006. One reason is the steady erosion of retiree health insurance coverage. In 1999, CBO estimated that employer-sponsored coverage paid 26 percent of Medicare beneficiaries’ drug costs. Another reason is the steady rise in beneficiaries with no drug coverage, which we have conservatively estimated at 40 percent, following Laschober and colleagues, as discussed elsewhere in this response. With 40 percent of people on Medicare lacking insurance for drugs entirely, with the continuing erosion of retiree coverage, with many others lacking adequate drug benefits, and with rising dollar co-payments, the gaps in coverage are great. It is simply not credible, especially absent supporting evidence, that 58 percent of Medicare beneficiaries’ drug costs were covered by insurance in 2003, leaving only 42 percent to be paid by patients themselves. By this measure, as with the number of people covered, PRI/PwC (and the similar CBO figures) appear to over-estimate Medicare beneficiaries’ current level of insurance coverage for prescription drugs. And the share covered by insurance can be expected to fall further between 2003 and 2006 as retiree coverage erosion continues to drop and as co-payments continue to rise. We doubt that improvements in HMOs/PPOs’ drug benefits for Medicare+Choice enrollees will be durable or adequate enough to offset more than a small portion of this coverage erosion and co-payment rise. The Kaiser Family Foundation (KFF) recently published projections that show a slightly higher, 46 percent, out-of-pocket share for 2006 (up from 43 percent for 2003).34 Notably, however, the out-of-pocket shares were described as being "consistent with” CMS projections for non-institutionalized Medicare patients.35 Yet one might expect the out-of-pocket share for the community-dwelling, non-institutionalized people on Medicare to substantially exceed the share for Medicare beneficiaries overall, because the latter includes many nursing home patients, the great majority of whom have Medicaid drug coverage.36 Other evidence casts the CBO and PRI/PwC out-of-pocket estimates much further into question. It is well recognized that Medicare patients pay out-of-pocket for a far larger share of their prescription drug costs than non-Medicare patients do. Yet evidence from the Medical Expenditure Panel Survey (MEPS) for 1999 showed that the out-of-pocket share of drug costs was 39 percent for non-Medicare patients—comparable to the figure cited by CBO and PRI/PwC for Medicare patients. Further, the MEPS survey found that in 1999 Medicare patients paid fully 57 percent of their drug costs out of pocket. Public and private coverage paid just 43 percent of Medicare beneficiaries' drug costs.37 These figures are virtually the inverse of those estimated by PRI/PwC and CBO, and of the similar KFF estimates. (The MEPS data appear to exclude drug costs for nursing home patients, as the KFF figures may. But it seems unlikely that prescription drug use by Medicare patients in nursing homes alone could raise the covered share for Medicare patients overall from 43 percent to the 58 percent estimated by PRI/PwC). The KFF data and CBO make use not only of MEPS data but also the Medicare Current Beneficiary Survey (MCBS). Each has advantages—and substantial limitations. Even with the most careful analyses, these surveys simply must be recognized as providing estimates of current drug spending that are unreliable bases on which to rest projections. Consider just one issue on which Poisal reported recently, for example–the complexity of adjusting such surveys to account for patients’ under-reporting of their medication use and spending. For the MCBS, he notes, “a difference of just 1 percent in the underreporting estimate can change total projected annual outlays” for prescription drugs among non-institutionalized Medicare patients by over $500 billion.38 c. Baseline spending estimate disparities Some of the foregoing difficulty with the PRI/PwC calculations may stem from the apparently very substantial disagreement between the CBO estimates of outpatient prescription drug spending on Medicare beneficiaries and the estimates for spending on Medicare beneficiaries derived from retail drug spending data compiled by the CMS Office of the Actuary. One possible element in the disagreement: CBO’s outpatient prescription drug spending may be measuring something different and greater than CMS’s retail prescription drug spending. For example, prescription drugs for patients in nursing homes ought to be included in CBO’s outpatient estimates, but in CMS data, some of them are tallied as nursing home costs rather than with retail prescription drugs. Consider the latest data from CMS, for 2002. The CMS Office of the Actuary reported that total U.S. retail prescription drug spending in 2002 was $162.4 billion.39 Medicare beneficiaries are reported to use some 35-40 percent of that medicine. CBO reports the 40 percent share.40 Now, 40 percent of CMS’s $162.4 billion is $65.0 billion. But this is visibly different from CBO’s own $87.0 billion estimate of outpatient prescription drug spending by Medicare patients in 2002. CBO’s reported figure is 33.8 percent above that derived by applying CBO’s own Medicare share of outpatient prescription drug spending to CMS’s total retail spending estimate. Worse, CBO’s estimates of spending on drugs for Medicare beneficiaries for 2002 rose by $6 billion from the estimates made in 2001 to the estimates made in 2002, meaning that CBO’s number has been drifting farther above CMS’s number. The two figures should be very close together if they are gauging the same reality. They appear to differ somewhat in definition, so even baseline spending estimates may be muddled.