Flash(ish) Finding: Drug Spending vs. Net Drug Spending vs. PMPM in Medicaid 2020
Re-Examining money from sick people
One of our founding principals at 46brooklyn was, and still is, to liberate the myriad drug pricing data sources scattered across the web and bring them together in meaningful ways as a means of shortening the divide between the public and drug pricing truths.
As we perform a self-reflection of our work on this endeavor to date, we must acknowledge that we still have a long way to go; with many data sources yet-to-be liberated. For example, while 46brooklyn has numerous dashboards that track a drug’s national average drug acquisition cost (NADAC) and other drug reference pricing benchmark’s movements over time, these measurements are often insufficient to measure a patient’s expected cost for a drug (as our inbox often tells us), nor are they necessarily a good measure of a plan sponsor’s likely drug expenditure on any given drug due to many hidden, retrospective price concessions. Said differently, we are regularly running into issues where simple measurements of these pricing benchmarks fail to tell us the more complete “story” as it relates to drug prices in this country. Which is of course frustrating given the numerous stories of drug unaffordability, particularly on products like insulin as this mother’s story show us:
One of the key challenges we face in determining “actual drug costs” are prescription drug rebates, or as we like to refer to them as “money from sick people”. Money from sick people muddies the waters in many conversations we’d like to have on drug prices – meaning we largely cannot have the conversation without caveating it in some way, shape, or form about how we address and account for the ~$200 billion elephant in the room. We explored this not too long ago with insulin, as did the U.S. Senate Finance Committee in 2021, which both show us how significant insulin rebates can be. But those reports are related to just one drug, and we wanted to re-visit the topic with a focus on the entirety of drug spend.
So that got us wondering if there was a way to liberate existing public data to explore the holistic impact of rebates on drug spending, and in doing so, help contextualize what their “value” to the system actually is (and by extension, what might a world with no drug rebates might look like?).
So if all this sounds like something you’re interested in, read on.
Getting our ducks in a row
As our history demonstrates, we have a particular fondness for logical extremes, as they can help us contextualize an argument in ways we might not otherwise be able to (or would require too much nuance to get out – and we already struggle with brevity). To broadly study the impacts of rebates, we were inevitably drawn to further study the public data in Medicaid. We say inevitable because of two facts.
Medicaid statutorily secures itself the best possible drugmaker rebates in the U.S. within the confines of the Medicaid Drug Rebate Program (MDRP). Provisions of MDRP state that if anyone in the broader U.S. drug market negotiates a better rebate for a given drug, then Medicaid is automatically entitled to the same price concession. What is good for the goose is good for the gander, so to speak.
Additionally, federal regulations limit the amount of money for prescription drugs that can come from patients in Medicaid. Medicaid is a program largely for the poorer individuals in our country, so it only makes sense that there would be fixed copayment maximums for many recipients of Medicaid. However, this fact also means that the Medicaid patient is largely agnostic to how the drug pricing games are played behind the scenes. If drug manufacturers want to set a high price to offer a bigger rebate because they believe that is how they’ll secure the most customers (i.e., health plan drug coverage), the Medicaid patient isn’t as likely to be harmed by the higher list price like say the Medicare or commercially insured individual may be. Said differently, we know from insulin that while rebates might be big, the size of those rebates may not be helping patients actually afford their insulin based upon the number of individuals rationing it (at least 25% in one study). By focusing on Medicaid, we can avoid some of the confounding impacts of rebates because of the required patient cost-share structure in the program.
Knowing this, we set out to find data on Medicaid drug expenditures relative to Medicaid drug rebates. If we could find such data (we did), then we would be able to perform an assessment of the Medicaid drug program, and the prices it pays, in a more complete way than we perhaps have to date.
For our analysis, we’re focusing on an oft-used measure of health expenditures per member per month (PMPM) spending. While we haven’t really used PMPM before in our analyses at 46brooklyn, PMPM spending is a useful measure for several reasons in today’s analysis.
First among these is the fact that the majority of lives in Medicaid are placed within Managed Care Organizations (MCOs) or what some may refer to as health insurance companies. MCO services are paid for via capitated payments; payments that are pre-arranged to deliver services on a PMPM basis. This means that an assessment of Medicaid drug spending on a PMPM-basis is most closely aligned with how the majority of Medicaid drug spending is actually experienced by the states paying the bill. More so than just a simple payment philosophy alignment; PMPM is a useful measure of both utilization and cost. This means it gives us a sense of overall program cost performance more than we would derive with just an average prescription cost or average number of prescriptions used per patient. While those measures can be useful, PMPM combines them in one value that helps us understand what it cost to provide a healthcare benefit, such as prescription drug coverage, to an individual.
Placing ourselves in the position of a plan sponsor, such as an employer, PMPM costs help us understand what it takes to provide benefits as our business grows and we hire more people. Ultimately, PMPM is precisely the type of measure we want in assessing the value of rebates to the Medicaid prescription drug benefit.
Assumptions
As with all of our work, we need to start with some data, and some assumptions. As we alluded to in the intro, and fortunately for us, all of our data for this flash finding is publicly sourced, so feel free to check our work, and thank those who make drug pricing data more available and accessible to the public.
For this analysis all the data we need can be found in one spot – MACStats. If you’re not familiar with the Medicaid and CHIP Payment and Access Commission (MACPAC), we would encourage you to visit their website and explore their reports, as they are the equivalent of what MEDPAC is to Medicare (which if you’ve followed our work, we’ve mentioned MEDPAC before when discussing Part D re-design). We’re only going to rely upon two of MACPAC’s numerous data sources for this analysis:
Exhibit 28: Medicaid Gross Spending and Rebates for Drugs by Delivery System
This exhibit will provide us with data on how much Medicaid spent on drugs (both gross and in the net after rebates), by state, for prescriptions within direct state-run programs (i.e., fee-for-service) as well as prescriptions under managed care organizations (i.e., MCOs); and
Exhibit 29: Percentage of Medicaid Enrollees in Managed Care by State
This exhibits will provide us with data on how many people got Medicaid benefits in the state and how many of those people’s benefits were managed by MCOs.
Combining these two data sets gives us valuable insight into what we want to study today. By joining drug expenditures to the number of covered patient lives, we can perform one of the most basic measurements in the insurance industry, which is simply how much it cost PMPM to deliver prescription benefits in Medicaid.
Because of the way the data is formatted, we can explore 2020 prescription drug payments both ‘at the pharmacy counter’ – assessing PMPM with gross pharmacy expenditures; and ‘in the net’ – assessing PMPM after accounting for all of Medicaid’s substantial discounts. A comparison of the two - gross and net - will give us an appreciate of the value of rebates, which is the goal of our analysis.
As with all analysis, our assessments in this regard will be limited because of our data sources. One of the key limitations we should note at the start is in regards to Medicaid’s reported rebates from drugmakers. Rebates are not collected until months after the claim was dispensed. This means that there is a temporal difference between the claim cost assessments we’re making and the rebate collections we’re relying upon. A claim whose expense occurred at the end of 2020 likely does not have its corresponding rebate collected within the rebate column of Exhibit 28. Similarly, some of the rebates collected in 2020 were for claims whose expenses were from the prior year(s) [there isn’t really a rebate collection ‘statute of limitations’, so rebates could be changing on claims going all the way back to 1990]. However, we feel this limitation is appropriately controlled for the purposes of this analysis because regardless of how, when, or why the rebate was collected, the Medicaid program still needed to be funded in 2020. Meaning that this temporal difference in rebate collection should be largely irrelevant to the aggregate measure of program performance we are undertaking today. As a result, the realities of the program are better reflected if we acknowledge that the system doesn’t perfectly recognize a rebate immediately upon dispensing the claim.
Another limitation worth mentioning up front is that drug mix differs between Medicaid MCO and FFS claims in ways that may be significant. This occurs due to intentional design, where each program (FFS vs. MCO) is trying to secure what it perceives to be the lowest net cost for drugs, but also due to eligibility differences. Different types of people take different types of drugs. However, due to the aggregate nature of this analysis, and given that some programs are nearly exclusively all MCO or all FFS, this limitation should be sufficiently controlled within the aggregate nature of the analysis performed today. It should also be noted that some states might be considered hybrid MCO/FFS models. For example, the Michigan Medicaid program is technically considered a managed care model; however, some medicines used for HIV, hepatitis C, cystic fibrosis, seizure disorders, and behavioral health are “carved out” from managed care and thus treated as FFS. Again, all of these differences may be significant in why our observations are what they are, but hopefully not as impactful to a directional analysis of which way things seem to favor.
A final limitation worth noting is that this analysis is focused on drug cost, cause that is kind of our thing. But understand that it would likely be of greater benefit to also measure ‘total cost of care,’ which would include an assessment of medical costs, durable medical equipment costs, and all other manners of healthcare costs beyond drugs. However, we know that drugs, and people actually adhering to their therapies, are one of the best ways we can likely control the other aspects of healthcare costs. Nonetheless, it is important to acknowledge that this is a drug analysis and not a total cost analysis.
All things considered, if you were a state or federal legislator tasked with funding Medicaid in 2020, or even an interested party such as ourselves, the assessment we’re performing is arguably your report card of program cost performance regardless of how it was achieved and regardless of the limitations of the analysis. Said differently, this analysis reflects the realities of how the benefit was funded in 2020. However, as our assessment will show, it may be worth looking into the why a little deeper as some of the individual state program results show.
Crunching the numbers
In fiscal year (FY) 2020, according to Exhibit 28, the Medicaid program spent a combined gross $71.8 billion on prescription drugs – $25.2 billion in the state-run FFS program and $46.6 billion in managed care (Figure 1).
Figure 1
Source: MACPAC Exhibit 28, 46brooklyn Research
These costs were offset by MDRP drugmaker rebate collections in excess of $39 billion (Figure 2) – resulting in a net expenditure of $32.6 billion (or a discount of 55% off the gross price in Figure 1).
Figure 2
Source: MACPAC Exhibit 28, 46brooklyn Research
It’s not lost on us how significant this data is in the context of our national conversation around high prescription drug prices. Of all the funds spent on prescription drugs within state Medicaid programs, more than half of the gross expenditures boomerang right back. While Medicaid rebates are as large as you’re going to find thanks to MDRP, is it really a stretch to assume that the largest PBMs in the country, which control north of 75% of all pharmacy claims, are not positioned equal to Medicaid in regards to rebates? Said differently, who do we think is negotiating the ‘best price’ on behalf of Medicaid if not for these groups? At the minimum, it is worth at least looking into whether they’re getting 50% of all drug expenditures back via rebates, administrative fees, or other price concessions from the drug manufacturers and determining where the money is actually going. But we digress.
Digging deeper into Figure 2 you’ll note that one Medicaid program (Delaware) secured more rebates than program expenditures in 2020. This is not an error on our part in presenting the data and we already identified why this happens – rebates are collected retrospectively. As if to confirm our prior statements on the matter; the footnote of Exhibit 28 confirms exactly what we said can happen: “Delaware reported large prior period adjustments for FFS that ultimately result in a positive FFS rebate amount.” So while Delaware looks like it made out like a bandit this year, really it just means that there was some under-collection of rebates in prior years. In a broader context, this likely helped Delaware a great deal in financing the program this year which just goes to show you how valuable getting rebates right really can be.
In FY 2020, 77 million Americans received Medicaid benefits according to Exhibit 29, with the majority (70%) obtaining benefits within an MCO program at the state level (Figure 3).
Figure 3
Source: MACPAC Exhibit 29, 46brooklyn Research
The preference for managed care within Medicaid can be understood when assessing the per member per month (PMPM) spending of the aggregate Medicaid program. Taking the total gross expenditures in Exhibit 20, divided by the number of lives in each program, and dividing by 12 (for the 12 months in the year), we see that the PMPM experience is $77.32 for Medicaid as a whole, made of a composite $88.92 PMPM for FFS beneficiaries vs. $72.22 PMPM for MCO (Figure 4).
Figure 4
Source: MACPAC Exhibit 28 & Exhibit 29, 46brooklyn Research
Across the 77 million covered lives in Medicaid that PMPM difference ($88.92 - $72.22) equates to $1.3 billion. Contextualizing this further means that if you’re a state legislator looking to fund your program into the coming year, managed care would appear to be the obvious solution to save your state money.
However, before AHIP starts typing up their press release highlighting how awesome 46brooklyn’s research is, taking a step back, and performing the same assessment but relying upon net expenditures as opposed to gross, we see that the FFS PMPM is significantly less than managed care after accounting for rebates ($21.89 PMPM in FFS vs. $40.85 PMPM in MCO) [Figure 5].
Figure 5
Source: MACPAC Exhibit 28 & Exhibit 29, 46brooklyn Research
While on average, states would expend 20% more up front with FFS, Figure 5 suggests that expense would be worth it, as they’d save 50% more in the net on a PMPM-basis. Again, assessing the value difference across all 77 million lives suggests an FFS-savings of $1.5 billion in 2020 ($21.89 in FFS vs. $40.85 in MCO).
Note that even when we exclude the states within Exhibit 28, whose footnotes suggest there may be some funny business going on with the rebates, we still arrive at the same conclusion in the net. Specifically when we remove Arkansas, Delaware, New York, Tennessee, and Wisconsin from our net PMPM analysis, we still observe a net PMPM of $22.12 for FFS vs. $31.25 for MCO.
It’s worth noting that this sub-analysis covers 67 million of the 77 million total Medicaid lives as well as $56.5 billion of the $71.8 billion gross expenditures. This means that the impact of the FFS-to-MCO observable PMPM difference would equate to $612 million under this more limited analysis. This appears to support our earlier assumptions that the aggregation of the data within the Medicaid program is likely sufficient; particularly as we’re focused on the direction of our observations today, rather than the discrete value. To be clear, Medicaid programs are not bearing these reported drug expenditures directly because of the capitated model by which those benefits are funded; however, the benefit is funded on a PMPM basis that includes a likely profit margin for the MCO. Meaning that if you were an actuary calculating the PMPM for the pharmacy service, you’re likely baking in an amount greater than the calculated PMPM.
And although it is only anecdotal evidence at best, we can’t help draw a correlation between the weighted average PMPM expenditures of the 10 programs with the lowest MCO participation rates to the 10 programs with the highest (Figure 6). In reviewing the direction of Figure 6, while the PMPM gross amount looks relatively similar across states in either group, the gap between gross and net is significantly higher in the lower percentage of MCO lives group.
Figure 6
Source: MACPAC Exhibit 28 & Exhibit 29, 46brooklyn Research
Finally, we can see extreme differences relative to the mean when assessing the top 10 programs with the lowest PMPM vs. the highest PMPM (Figure 7). Again, in reviewing the direction of Figure 7, we can’t help but notice the more favorable group appears to be the lower on average MCO lives.
Figure 7
Source: MACPAC Exhibit 28 & Exhibit 29, 46brooklyn Research
For those interested in checking our math, the data set in support of this analysis can be downloaded here (though it is nothing more than a combination of Exhibit 28 & Exhibit 29).
Money from Sick People
The analysis presented here, while it does have its limitations, demonstrates that ‘money from sick people’ is a big deal. As shown, money generated off sick people can massively help finance a healthcare benefit. You would be hard pressed (i.e., it would be nigh impossible) to find another area that offers you an aggregate 55% discount on the product or service in the net. Whether it is a hospital bill, a doctor visit, or surgery, they are highly unlikely to offer you half of your money back a few months down the road via a “price concession.” Sure, they’ll put a high sticker price on their website and show you the “value” of your insurance in discounting that price down – just like we do with artificially inflated drug costs – but when the bill comes due, you must pay up for those services.
That is not the case with drugs.
At some point someone within the drug supply chain made a “rebate” off of whatever sickness required you to take the drug you were prescribed. And we have to put rebates in quote here because Medicaid isn’t the only program with statutory price concessions. The 340b program is entitled to the same rebate amounts as Medicaid, and given the growth of the 340b program, its important to note that it may not always be the health plan that is making a lot of money off of the drugs charged to patients.
However, the drug money will only come due if people actually can afford their medications, something we think the Medicaid program realizes. By insulating the patient from the games of drug pricing, Medicaid programs can go about their business of minimizing net program costs. Some programs do this exceptionally well, with very low net costs even relative to the averages seen in the aggregate Medicaid experience (which is already set-up for success with the MDRP). But we shouldn’t lose sight of the fact that even in the states with higher net PMPM figures, they’re still getting a ~50% discount off the cost of providing the healthcare benefit in the net. Which tells us that in a world devoid of drug rebates, the potential savings to the system could be a 50% reduction for all parties, so long as those parties are not currently capturing and with holding onto these rebate dollars. This analysis also further emphasizes that it is very difficult to contextualize what constitutes “the price of a drug” without considering the impact that discounts have on the list price itself – and the degree with which patients and plan sponsors are exposed to those list prices. To be clear, it can be a great thing that these programs have access to a way to make drugs 50% more affordable, to let them stretch scare resources farther, but only if patient’s are protected from the game. Otherwise, we have to question if it is all worth it in the end.
xxx