CMS Says AMP is Coming; Acquisition Cost Data Reduce Generic Costs 37% in ‘Bama; Why Ortho Demand May Slow in the Mid to Longer-Term

Print Friendly, PDF & Email

Richard Evans / Scott Hinds

203.901.1631 /.1632

richard@ /

April 6, 2011

CMS Says AMP is Coming; Acquisition Cost Data Reduce Generic Costs 37% in ‘Bama; Why Ortho Demand May Slow in the Mid to Longer-Term

  • CMS confirmed to us their intention to publish average manufacturer price (AMP) data to the general public; this is in addition to Secretary Sebelius’ February 2011 commitment to provide average acquisition cost (AAC) survey data to the states
  • We analyze Alabama’s publicly available AAC dataset, and show that AAC values are on average 87% below AWP, and 74% below Federal upper limits (FUL)
  • By switching to AAC, we show that Alabama saved an average of 37% per generic prescription, even after nearly doubling pharmacists’ dispensing fees. Potential savings are comparable in other states, all of whom are likely to move away from AWP
  • AWP’s relevance is fading; no state is likely to use the benchmark for much longer, and we expect the two major providers of pricing data to stop publishing AWP – First Databank in September of this year, and Medi-Span as soon as either AAC or AMP is published by CMS
  • The fading of AWP may force the re-benchmarking of commercial (i.e. PBM) contracts, practically all of which currently are benchmarked to AWP. We’re increasingly convinced that PBM contracts will shift to acquisition cost benchmarks, and that generic dispensing margins compress as a result
  • We analyze hospital discharge data over a 7 year period, and show that much of ortho demand growth – especially for hips – came from growing numbers of elective implants in ever younger patients. We also show that like spinal surgery, where payors are tightening eligibility standards, that knee and hip implant rates vary significantly across regions. At a minimum patients cannot keep getting younger indefinitely, and there is a non-trivial risk of payor tightening. Cardio demand faces neither of these pressures; nevertheless cardio and ortho expectations and valuations are nearly identical
  • Correction: At the end of our March 2, 2011 call “Post-2014 Reform-Related Volume Gains” we misidentified our “median case” as an uninsured rate of 18 percent. In fact, our median case should have been 15.7 percent uninsurance. Correcting this error shifts the modeled median case gross profit by less than 0.5 percent (from 5.6 percent to 6.0 percent) and does not change any of our conclusions

Alabama moves to AAC reimbursement benchmark; CMS confirms it plans to publish both AMP and AAC

In September 2010, Alabama became the first state to move its Medicaid drug reimbursement benchmark to a measure other than AWP (by far the most common) or wholesale acquisition cost (WAC) when CMS approved the state’s Average Acquisition Cost (AAC) survey methodology. AAC is based on the results of a mandatory, semi-annual survey of pharmacies to determine – and subsequently tie Medicaid reimbursement to – the actual cost of acquiring and dispensing a prescription.[1] The AAC data are updated weekly between the twice yearly surveys, and mandatory participation by pharmacies is done on a rotating basis, such that each Medicaid-participating pharmacy is required to submit its actual invoiced acquisition costs per NDC number approximately once every two years

Before September 2010, a pharmacist filling a Medicaid prescription in Alabama was paid (the lesser of the FUL, WAC+9.2% or) AWP-10% for the cost of the ingredient, plus a $5.40 dispensing fee. This discount to AWP was slightly lower than the national median discount (-14%) for a multi-source drug; the dispensing fee was slightly higher than the national median ($4.34). Ingredient cost reimbursement now is set to 1.0 x AAC (of course, capped at the FUL), with a significantly higher dispensing fee of $10.63


To understand the practical implications of this reimbursement change, we looked at the most-prescribed generic drugs in Alabama’s Medicaid program during the first half of 2009 (the most recent data available), and estimated a weighted average AWP for each.[2] Across our sample, the weighted average AWP – $46.59 – was almost 7 times larger than the weighted average AAC (as calculated by the Alabama survey) – $5.89 – across the same sample. Because reimbursement is capped, the more meaningful comparison is to the weighted average maximum price per Rx of the sample (FUL, with 0.9 x AWP replacing FUL for generics with no federal upper limit) – $22.02 – which is nearly 4 times actual acquisition costs (Exhibit 1)

When we apply Alabama’s Medicaid reimbursement formulas – both before and after the switch to AAC pricing – note that on average, the total dispensed cost per prescription, including the dispensing fee, declines by 37 percent (Exhibit 2). Without Alabama’s decision to nearly double the dispensing fee, the drop in reimbursement per prescription would have been a full 72 percent

Critically – and not surprisingly – we found that the newer the generic, the bigger the savings from the switch to AAC (Exhibit 3). When generics first enter the market, pharmacies’ acquisition costs fall faster than either AWP or FUL, so dispensing margins on newer generics tend to be larger. With AAC-based reimbursement the benchmark is the pharmacies’ acquisition cost, so outsized dispensing margins on newer drugs either do not develop, or if they do develop, quickly fade. I.e. not only does AAC reduce dispensing margins during a lull in major generic approvals (characteristic of the timeframe for our dataset); AAC also should reduce or even eliminate new-generic windfall dispensing profits

All together, our sample of generic drugs represented about 20 percent of Alabama’s total 1H09 Medicaid prescription drug spend, and we estimate that AAC reimbursement could reduce total spend on generics alone by 37 percent; which implies total savings of about 7 percent of total prescription (brand and generic) spend, all else equal. This is very much in line with the estimates published by HHS, which suggested $30mm or 6 percent savings for Alabama in year one of the program. But note that Alabama, though the first to move to a reimbursement formula centered around the actual cost of acquisition, is hardly the only payor who could reap substantial savings by adjusting their benchmark

We performed an illustrative analysis on all other states with generic reimbursement tied in some form to AWP (every state except Massachusetts and Rhode Island, which use formulas strictly based on WAC). We assumed that ingredient cost moved to simply the lesser of FUL and (Alabama’s) AAC; while dispensing fees nearly doubled (i.e. increased by the same factor as Alabama’s, or 10.64/5.40 ≈ 1.97). We kept aggregate prescription drug spending and relative generic volumes constant, in order to simplify and focus just on the difference between the current reimbursement benchmark and an AAC model. State-by-state reimbursement rates and analysis results are in an Appendix. We found that on average, states could save approximately 7 percent of their Medicaid prescription spend (with a range of 3 to 10 percent) with an AAC benchmark. Given the state budget crises and near-certain cuts to Medicaid about which we’ve previously written[3], these savings appear too large to ignore

For commercial payors, FTC estimated that PBMs reimburse retail pharmacies at rates generally similar to most state Medicaid agencies (lesser of AWP – 15% / MAC list; plus a $1.75-$2.50 dispensing fee). To model a typical PBM, we used a 15% discount to AWP, the FUL as an approximation of a MAC list price, and the high end of the dispensing fee range. In this context, moving to AAC implies drug cost savings potential of 52 percent on retail-dispensed generic prescriptions alone. However, we also know that on mail-order prescriptions, PBMs typically do not apply a MAC list (though they do take a steeper discount to AWP – usually around 50 percent), which implies that mail-order generic savings should be even higher than retail. All in, this suggests a conservative estimate is at least 9 percent total savings across all drug spending

Of course, all of these savings are predicated on the availability of accurate, timely, and accepted AAC survey data. Reporting under Alabama’s survey is mandatory, and the results suggest they have reached the critical mass necessary to fully wean themselves from AWP (the number of NDCs that we sampled for which we were unable to find an AAC was well under 1 percent). However, the inefficiencies and administrative overlaps inherent in running 50 separate acquisition cost surveys almost certainly means that an enforceable national survey emerges. HHS Secretary Kathleen Sebelius – in her February 3 letter to state governors – first acknowledged that HHS was undertaking a national actual acquisition cost survey, results of which would be available to states as a reimbursement benchmark, in 2011

Our contact at CMS confirmed the survey, though acknowledged that some specifics, including regarding its dissemination, were still under consideration. However, we were able to confirm that in addition to its survey of generic acquisition costs, that CMS “will provide the weighted average of the most recently reported monthly average manufacturer prices (AMP) and the average retail survey price on [its] website.” This in itself is a significant development as it is, to our knowledge, the first time that CMS has confirmed it will follow through on the Affordable Care Act provisions calling for publication of AMP, and that the data will be publicly available, rather than available only to state Medicaid programs

Results of the AAC switch in Alabama, and activity at CMS, provide further support to our belief that the AWP benchmark ultimately will disappear, forcing commercial drug benefit contracts to be re-benchmarked to either AAC or AMP. Public availability of national AMP, and likely AAC, data obviates the need for AWP. First DataBank announced plans to discontinue AWP in September of this year when it carried out the re-setting of AWP markups as part of its price fixing settlement in 2009. Medi-Span, in the wake of the same settlement, announced that it will continue publishing AWP until “relevant industry or governmental organizations develop a viable, generally accepted alternative price benchmark to replace AWP”. Clearly both AAC and AMP are viable alternatives, thus we’re likely to see the two largest publishers of drug price data cease publication of AWP in the near to mid-term, which to our minds catalyzes the migration of commercial drug contracts to either AAC or AMP, with a commensurate reduction in pharmacy dispensing margins

Demand trends for medical devices – can ortho implant recipients get any younger?

We’ve taken a step further to analyze the unit demand for medical devices, adding a third theme of analysis to the two we’ve published previously

Our two prior analyses argue that: 1) in the near term, reported unit growth for (particularly discretionary) products and services will improve, because of easing comps[4]; and 2) expectations for reform-related demand growth in 2014 probably will not be met[5]. On net, this argues that ortho (relatively more discretionary) unit demand is likely to surprise on the upside in 2011, but that longer-term unit growth expectations (across healthcare generally, but especially for more elastic demand such as ortho) may not be met. In contrast, 2011 demand growth for cardio is unlikely to surprise; cardio unit demand is largely non-discretionary, so there was little or no pull forward of demand in ‘08/’09 that would have suppressed reported unit growth in 2010. And longer-term, while cardio unit demand also may not benefit from the levels of reform-related volume growth that consensus assumes, the (relatively non-discretionary / inelastic) cardio demand forecast is less reactive than ortho to reform-related insurance trends, so 2014 cardio estimates presumably have less downside. Nevertheless, despite these sharp differences, ortho and cardio growth estimates and valuations are remarkably similar

The preceding analyses both assume that payor policy is constant, i.e. that insurers do not become either more or less restrictive with respect to eligibility criteria for implants. This is overly simplistic, and inconsistent with the risks of changing eligibility standards as highlighted by recently increased restrictions for spinal implants. Accordingly, our third theme of medical device demand analysis is to examine demand growth for implants over the last several (7 to 8) years, focusing particularly on growth according to payor (Medicare v. commercial) and indication (elective v. non-elective), and also on changes to the average age of implant recipients

The typical ortho implant recipient has changed fairly dramatically (more commercial, more discretionary, and younger) across this period, where the typical cardio implant recipient (with the exception of ICDs[6]) has remained fairly constant. Exhibits 4 thru 10 show how the typical implant recipient has changed between 2001 and 2008 with respect to whether the payor was Medicare or commercial, and whether the procedure was elective or non-elective.

With the exception of ICDs (Exhibit 9), ortho demand has grown much faster than cardio demand, elective procedures have added much more to growth than non-elective procedures, and commercial payors have contributed more to growth than Medicare. These trends are further summarized in Exhibits 11 and 12. Exhibit 11 shows the change in share of total procedures by payor and indication, and again shows that change in demand for ortho is characterized by an increase in commercial elective procedures, and that this trend is far more pronounced than in cardio. Exhibit 12 shows growth in discharges between ’01 and ’08 by implant type, and shows the share of growth attributable to either Medicare/elective or commercial/elective recipients. Generally speaking ortho discharges are growing more rapidly (excepting ICDs), are more likely to be elective, and commercial payors play are more substantial role in demand growth


Across all payors the average age of ortho implant recipients (except for spinal fusion) is falling (Exhibit 13), where the average age of implant recipients for cardio generally is stable. In part the decline in average age across all payors is attributable to an increased proportion of commercial patients relative to Medicare patients; we also analyze the changes in average age of commercial-paid patients (Exhibit 14), and see that the average age of commercial-paid knee implants has fallen slightly, where the average age of commercial-paid hip implant recipients has fallen significantly. Also, we find evidence that the rate of growth in hip and knee implants appears to have outstripped the rate of growth in patient conditions / complaints that presumably correlate with implants (Exhibit 15)

This leads to the question of whether recent demand growth in orthopedic implants is sustainable, in either of two senses: first, at a minimum, there must come a point at which (especially hip) patients cannot keep getting younger and presumably healthier; and to the extent this trend has driven growth in ortho implants, the trend eventually must stall irrespective of changes in payor eligibility standards. Second, to the extent physicians’ / institutions’ discretion in selecting implant recipients has led to un-necessary (or at least less convincingly justified) implantation (e.g. spine), then the ortho demand trend may be susceptible to a tightening of payor eligibility standards

In 2007 the Dartmouth Center for Evaluative Clinical Sciences published an analysis of procedure rates across 306 hospital referral regions. Specifically, the Center compared the percent of Medicare enrollees in each region that had a hip fracture, knee replacement, hip replacement, or back surgery. Hip fracture rates are effectively a benchmark – whether a patient has a fracture is independent of any subjective tendency on the part of the physician or institution to recommend care, so the variance across regions in the rate of hip fracture among Medicare enrollees predominantly reflects natural variance in the rate of the event, and little if any variance in practice standards. The Center then compared rates of knee replacement, hip replacement, and back surgery across the regions to rates of hip fracture across the regions. In contrast to hip fracture rates, knee hip and back procedure rates include not only the natural rate of occurrence of an underlying diagnosis, but also the variance in physicians’ and institutions’ practice patterns. The Center found that rates of knee replacement were nearly 4 times more variable than hip fracture, rates of hip replacement nearly 5 times more variable, and rates of back surgery more than 7 times as variable (Exhibit 16). These results argue that physicians and institutions have, and are exercising, enormous discretion over whether patients receive knee, hip, or back procedures; and, that the relationship between patient status and likelihood of surgery for knees and hips looks more like back surgery (loose) than hip fracture (tight). In turn, this suggests that knee and hip eligibility criteria are sufficiently loose to have opened the door to over-implantation – as has clearly happened with spine – and that knee and particularly hip demand trends face some non-trivial risk of tightening eligibility standards[7]

Turning to valuation, as disappointing demand trends began unfolding in late 2009 through 2010, sales and earnings expectations fell for many health care sub-sectors, including the ortho and cardio device companies. Ironically, expectations fell further / faster for cardio than ortho (Exhibits 17, 18); this despite the fact that slow reported 2010 demand appears to be a function of procedures that were pulled forward into 2008 / 2009 – which would have suppressed reported ortho results, but not cardio. Whereas sales and earnings estimates fell much faster for ortho and cardio sub-sectors than for health care as a whole, P/E’s for ortho and cardio generally tracked the broader health care trend (Exhibit 19), which suggests that cardio and ortho share prices moved on the revisions

Whatever the reason for greater downward revisions for cardio during late 2009 and throughout 2010, cardio and ortho valuations and forward expectations are now remarkably similar (Exhibit 20), despite the sharp differences in these subsectors’ exposure to the three key demand trends going forward. In the immediate term (as early as 1Q11 results) ortho is more likely to surprise on the upside, as results are no longer comparing against a period in which demand was inflated by job-loss related pull-forward. However in the middle- to longer-term, we believe the ortho demand trend will at least decelerate (especially for hips) as the trend to younger / healthier patients ultimately is unsustainable, and could flatten or even decline if payors tighten eligibility standards as has been done with spine, and as appears to be feasible in hips and knees. Finally, in 2014 we expect: 1) fewer net new insured (zero to 10 million v. CBO estimate of 30 million) from health reform; 2) falling average generosity of insurance as a result of reforms; and 3) as a result that reform-era expectations for (elastic / discretionary) ortho are more likely to be disappointing than for (less elastic / less discretionary) cardio

  1. Moving Medicaid pricing away from AWP was mandated by the Alabama state supreme court, which vacated almost $300mm in pricing fraud judgments against pharmaceutical manufacturers, in part because Alabama did nothing to “alter its course of business” (i.e. continued to tie reimbursement to AWP) despite evidence that AWP was an inflated measure.
  2. There can be dozens of NDCs—varying by labeler, dosage, formulation, combinations and package size—for a single generic drug, each with its own AWP. We sampled 1,040 of the most-utilized NDCs for 85 of the most-prescribed generic drugs to estimate a weighted average AWP per generic.
  3. See “Medicaid cost pressures intensify on states,” March 22, 2011
  4. Please see our December 1, 2010 note “Demand Trend Improves Starting 4Q, Ortho & Commodity Suppliers Benefit Most …” Quickly, the idea is that employees fear job (and health insurance) loss when the rate of job loss is high (‘08/’09), and so tend to pull forward discretionary demand (e.g. ortho). This inflates discretionary demand during (and for about 3 months after) the period of rapid job loss, leading to a difficult comp in the following period. As of 1Q11 results, we believe the comparable period will no longer be affected by high levels of job-loss related demand
  5. Please see our March 2, 2011 note “Post-2014 Reform-Related Volume Gains are Modest”. Briefly, the argument is that CBO’s estimate of 30 million newly insured contains 2 errors: first, we believe CBO over-estimates enrollment among persons who fall between Medicaid and employer-sponsored insurance; and, second, we believe CBO under-estimates the percent of persons who choose not to be insured after having been moved by their employer from employer-sponsored insurance (ESI) to the health insurance exchanges (HIEs). And, though CBO does not make an explicit assumption about the amount / generosity of insurance purchased by enrollees on HIEs, we expect these enrollees to opt for less generous plans (e.g. an average actuarial value (AV) of roughly 65 on the exchanges, versus a current average AV of roughly 82 under ESI)
  6. For obvious reasons unrelated to the demand trend for other implants, namely new clinical evidence that led to increased preventive / defensive use of ICDs in patients having certain risk factors for fatal arrhythmia
  7. We recognize that the Dartmouth study is in a Medicare population, and that because so much of ortho implant growth has been from commercial payors, that commercial payors are the ones whose policies we need to be most concerned with. Our sense is that commercial payors are far better positioned to tighten policy than Medicare, and that because commercial patients are younger and healthier, that the ‘looseness’ between patient status and odds of being given a knee or hip implant are greater in the commercial setting than in the Medicare setting
Print Friendly, PDF & Email