Analyzing HHS’ 100 Days of Action report

On August 20, 2018, the U.S. Department of Health and Human Services (HHS) released an analysis of the accomplishments that the Trump administration has achieved in the first 100 days since the release of the President’s American Patients First Blueprint. Authored by former CVS Health and Pfizer executive Dan Best, the Report on 100 Days of Action on the American Patient First Blueprint provides some interesting insight into the Trump administration’s progress on delivering on their promise of lower drug prices.

While there is a lot that needs deciphering, HHS made two very significant claims in this report. They assert that over the first 100 days since the Blueprint's release there have been:

  1. 60% fewer brand-drug price increases compared to the same period in 2017, and

  2. 54% more generic and brand-drug price decreases compared to the same period in 2017.

On its surface, this would seem to be substantial change, but what does this mean for the people burdened with high drug costs? This is like your grocer telling you that they raised the MSRP on 60% fewer SKUs this year compared to last, when all you want to know is if you’ll save money on a gallon of milk.

As Donna Young at S&P Global Market Intelligence pointed out, there are all sorts of issues with this analysis. HHS’ methodology report, released on August 21, 2018, acknowledged these issues and fittingly ended with the following statement: “we are unable to conclude much about the magnitude of the price changes themselves… or their effect on overall prescription drug spending.”    

It’s hard for us to accept this statement and move on. It’s like getting sucked into a riveting book, only to realize your dog ripped out and ate the last several chapters. 

So we started digging, in hopes that we could provide a bit more context to leave this topic a bit less unfinished. Our goal was to use only publicly-available, free data (in line with both our budget and our philosophical views on data transparency) to try to at least write some semblance of an ending to this story: What actually happened to drug prices in the 100 days since the release of the Trump administration blueprint to lower drug prices?

60% fewer brand drug price increases in 2018 vs 2017? Confirmed!

“When we released the blueprint with [President Donald Trump], few believed that, within months, drug manufacturers would begin to change their annual ritual of significant price hikes. Yet that is exactly what happened in the months following.”
- U.S. HHS Secretary Alex Azar (August 20, 2018) -

We started off trying to replicate the first claim, that there were fewer brand-drug price increases compared to the same period in 2017. To be clear, we do not have access to the manufacturer price data that HHS used, nor do we know for sure if they used Average Wholesale Price or Wholesale Acquisition Cost (although based on the source, we suspect it was WAC). But the good news is that for brand-name drugs, it doesn’t really matter. CMS has published a handy “NADAC Equivalency Metrics” table that shows that freely-available National Average Drug Acquisition Cost (NADAC) is a very consistent offset to either AWP or WAC.  So we felt comfortable using NADAC pricing to conduct this analysis. 

Once we knew NADAC would work, we simply pulled the weekly NADAC prices for each brand-name drug NDC Description (e.g. ABILIFY 30 MG TABLET) over the time period specified by HHS. Between May 11, 2018 and August 15, 2018 (only 96 days 🤔) there were 14 sets of weekly NADAC prices – one published each Wednesday.  Between May 11, 2017 and August 15, 2017, there were actually only 13 Wednesdays, and therefore, 13 sets of NADAC prices, so to ensure we had comparable datasets in both years, we moved the starting point of the 2017 date range back to May 10, 2017 to capture the additional 14th week of data.

With comparable data sets, we then moved on to the task of counting the number of price increases and decreases to see how our numbers compared to what HHS published. First, we defined a “price change” to be any week-to-week price movement with an absolute value of 0.5% or greater to remove minor week-to-week movements in NADAC pricing. Next, we adjusted our 2018 numbers for all NDC Descriptions that experienced intra-period manufacturer pricing reversals (i.e. a price increase that was almost immediately reversed by the manufacturer). These pricing reversals appeared to be inflating both the price increase and decrease count in the NADAC pricing data, and distorting the comparison with HHS’ data.

As a quick aside, 82 of the 83 NDC Description-level pricing reversals we found came from Pfizer-owned labelers. As you may recall from July, Pfizer rolled back their summer price increases after discussions with President Trump and administration officials. Figure 2 shows this reversal for one of Pfizer’s higher-volume brand drugs, Lyrica.  You can access a full list of these pricing reversals in our 100-days Analysis Support file (sheet #1), and visualize the rest of them using 46brooklyn's Drug Pricing (NADAC) Dashboard.

Figure 2
Source: CMS NADAC Database; FDA product/package tables; 46brooklyn Research

After making the aforementioned adjustments, we came up with a 64% reduction in brand-drug price increases, compared with HHS’ 60% reduction, as shown in Figure 3. In the context of the claims in the HHS report, it appears the administration was pretty much on the ball. But there’s more to the story than just tracking overall movement of list prices.

Armed with a dataset that replicated HHS’ results, we then set out to apply some sort of a “drug mix” to size these price changes. Here’s why overlaying a drug mix is important. Think of it this way: If your grocer lowered the prices on 25% of its goods, you may initially assume that you would be getting discounts on a quarter of your purchases. But what if those cuts occured on niche, underused items like canned liverwurst and powdered goat’s milk? Those price cuts may never impact your wallet. Meanwhile if the grocer increased prices on 10% of its more common goods like bread and booze, you may need to add a good mixer to your grocery list, because you just experienced some unfavorable mix shift (which will explain your higher grocery bill)!

Prescription drugs are the same. While every patient’s individual drug mix is crucial to consider, when discussing the overall impact of price increases and decreases, it’s important to apply a typical drug mix to the data. This allows us to truly estimate whether the price changes will have a meaningful impact on the payer.

The most accessible (and very large) source of "drug mix" data is of course the Medicaid State Utilization database. We dug into this database and chose the top five states that appeared to have reported complete data to CMS in Q1 2018 – New York, Texas, Florida, North Carolina, and Massachusetts (as an aside, we would have preferred to use all states in this analysis, but it’s very clear from the aggregate reported Medicaid spending that Q1 2018 data was only partially reported by many states in July. So we opted to only choose a handful of larger states that appeared to have complete Q1 2018 data). We pulled both Q1 2017 and Q1 2018 spending by brand-name NDC Description for these five states to use as our base “mix” for each year.

Conveniently, both Q1 2017 and Q1 2018 totalled to $2.6 billion in brand-name drug spending for the five states combined. We then applied the May 11, 2017 through August 15, 2017 price changes to Q1 2017 base spending and the May 11, 2018 through August 15, 2018 price changes to Q1 2018 base spending. Figure 4 shows the results of this analysis.

Figure 4
Source: 46brooklyn's 100-days Analysis Support file

Using these five states, we calculate that the 2017 price increases resulted in $26.0 million of inflation (1.0%) off a Q1 2017 base, whereas the 2018 price increases resulted in only $8.5 million of inflation (0.3%) off a Q1 2018 base.  You can access all of our detailed work in 46brooklyn's 100-days Analysis Support file (flip to sheets # 5 and 6).

One of the more interesting findings came from comparing the top 100 drugs by spending in both years. In 2017, 19 of the top 100 drugs showed increases of 3% or more. In 2018, only 6 of the top 100 showed increases of 3% or more.  So based on this five-state sample, it does not appear that manufacturers were concentrating their pricing freezes on the drugs with lower utilization. This is good news.

So overall, we found a 0.7 point reduction in aggregate brand-name drug price inflation across these five states.  This may not sound like a lot, but overall brand-drug spending in Medicaid was around $40 billion in 2017. While it’s completely unfair to extrapolate these five states to all of Medicaid, we'll do it anyway just to get our bearings.  If we apply even half of the 0.7 point inflation reduction across all of Medicaid, it would generate gross savings approaching $150 million per year, which is significant.

Or is it? We must remember that the brand drug prices we are working with are all gross, pre-rebate price. With rebate dollars being hidden from the public, there is no way for us to know if the true net cost to the system is coming down. So while these gross pricing trends may show good progress, it’s important to stress that these are surface-level achievements that may not necessarily translate into lower drug spending. This is less of a critique of the administration than it is of the drug supply chain that makes its living off of rebates (and then blaming each other for the problem).

54% more overall price decreases in 2018 vs 2017? Not so fast!

“Our research revealed that for a fair assessment of true drug costs in relative value studies, researchers should not look to WAC or AWP, but to publicly available average acquisition cost benchmarks that are well-designed and statistically sound.”
- Julie Suko, director of editorial services First Databank (March 31, 2014) -

We then moved on to HHS’ second claim, that there were 54% more price decreases among brand and generic drugs combined.  We immediately knew there was going to be a problem replicating this analysis using public NADAC data when we realized that HHS was almost certainly using manufacturer-reported prices to assess generic price changes.

Simply put, the “MSRP-like” prices that manufacturers set (and forget?) on generic drugs don’t have much of anything to do with the invoice price. This view is captured perfectly in the Julie Suko quote at the top of this section. There is some fascinating irony in that the source of this quote was from First Databank, because it turns out that the manufacturer pricing that HHS used to measure relative generic pricing changes came from First Databank (Figure 5).

Another helpful view of the disconnect between AWP or WAC and invoice cost again comes from Myers and Stauffer’s NADAC Equivalency Metrics document. The pricing data pros at Myers and Stauffer looked at the gap between NADAC and both AWP and WAC by “number of labelers” (Figure 6). If both AWP and WAC were accurate and representative pricing benchmarks, there really should be no noticeable relationship between the number of labelers and the pricing gap/offset between NADAC (i.e. pharmacy invoice price) and AWP/WAC (i.e. manufacturer price). But that is not what Myers and Stauffer found. They found substantial escalation in the gap as the generic drug matures and more labelers enter the market.

  Figure 6   Source:  NADAC Equivalency Metrics

Does this mean that manufacturers are simply “setting and forgetting” their generic prices? While we don’t have access to the all of the data we need to comprehensively answer this question (it’s not free and publicly available), the team over at Glass Box Analytics was kind enough to provide us with AWP “temporal charts” for four different generic drugs (Aripiprazole 5 MG, Esomeprazole 40 MG, Ezetimibe 10 MG, and Quetiapine ER 300 MG). We chose these four drugs because each experienced substantial invoice cost deflation over the past two years (90%, 81%, 97%, and 95%, respectively) as more competitors entered the generic market. 

Yet as shown in GlassBox Analytics’ temporal charts (Figures 7-10), AWP for these drugs hasn’t budged. Across dozens of NDCs that comprise these four drugs, only one NDC (Quetiapine ER 300 MG) was adjusted down to more closely align with true invoice price. No wonder why Myers and Stauffer found such a strong relationship in Figure 6 between the number of labelers and the NADAC-to-AWP pricing gap – AWP is not only stale, it’s "Ain’t What’s Paid."

Figure 7
Source: Glass Box Analytics

Figure 8
Source: Glass Box Analytics

Figure 9
Source: Glass Box Analytics

Figure 10
Source: Glass Box Analytics

So if the goal is to assess the true market price movement over this 100-day period, using manufacturer pricing for generic drugs is not going to be very helpful. As such, we attempted to replicate HHS’ analysis for generic drugs using NADAC pricing. The results are very different.

Using invoice costs instead of manufacturer costs, we found 3,002 brand and generic drug price decreases between May 11, 2018 and August 15, 2018, 1% lower than the same period in 2017. You can access all of our work in our 100-days Analysis Support file (sheets #3 and 4). This clearly tells a much less rosy story on drug price decreases than HHS’ 54% claim – one we believe may be more accurate.

We then used the same methodology used to size the brand price changes to size the generic price changes. Unfortunately, on this five-state mix, we found that deflation is subsiding. Figure 11 summarizes our findings, but it's worth flipping through sheets #7 and 8 in our 100-Days Analysis Support file for more drug-level detail. 

  Figure 11   Source: 46brooklyn's  100-days Analysis Support  file

Figure 11
Source: 46brooklyn's 100-days Analysis Support file

There appear to be two drivers of the reduced deflation. 

First, 2017 included several generic drugs that were at a steep point of their deflation curve (Aripiprazole, Quetiapine ER, Valganciclovir, Abacavir). In this analysis, Aripiprazole alone was responsible for nearly as much deflation in 2017 as all generic drugs combined in 2018. Fast forward one year and the 2017 high-deflation drugs are logically a smaller part of the 2018 mix, and there appear to be fewer high-volume/high-deflation generics replacing them to bring aggregate deflation back up. Classic negative mix effect – more complicated, but fundamentally no different than the grocery example provided earlier. (This topic deserves a separate research study to validate!) 

The second driver is that some of the more highly-utilized generic drug prices are quietly on the rise, at least moreso than was the case in 2017. 

So based on this five-state Medicaid sample, generic deflation has waned by 3 points.  If we (again, unfairly) apply even half of this to Medicaid's $13 billion or so in overall generic spending, that would result in an incremental loss of nearly $200 million in generic deflation, offetting the brand drug incremental savings we guesstimated in the previous section. 

There is admittedly so much assumption in this exercise that the absolute numbers are not all that meaningful. But going through this work has us convinced that we need to keep a closer eye on the generic market going forward. There is a lot of brute force effort being put into pushing down brand drug prices, but no guaranteeing that the savings will not just leak out somewhere else in the supply chain.

How do we lower drug prices when we don’t know what they are?

"All available evidence suggests that unwarranted healthcare price variation cannot be fixed by market forces and, instead, requires policymaker attention. Yet, treating prices as ‘trade secrets’ leaves government officials in the dark when it comes to understanding high healthcare prices and unwarranted variation in their states."
- Altarum Healthcare Value Hub in Revealing the Truth about Healthcare Price Transparency -

Ultimately, we don’t fault HHS for trying to somehow quantify the impact recent federal actions have had on drug pricing.  It’s human nature to want to know the impact of your actions. But as should be clear if you have made it this far in this report, there are no simple cause and effect answers in a system that is this complex. We can take action to reduce a price, only to realize that the price that we reduced was not really the net price (brand-drugs) or worse off, not a meaningful price at all (generic-drugs). Then there is the possibility that when one part of the supply chain decreases a price, another could gobble up the savings.

As Mylan CEO Heather Bresch explained back in 2016, price decreases are complicated because reductions “would not be guaranteed to flow through to the patient.” Meanwhile, due to the lack of easy-to-use, low-cost, drug mix data, we struggle mightily to assess whether or not we are even moving in the right direction.

Management guru Peter Drucker is often quoted to have said, “you can’t manage what you can’t measure.” If we have learned anything from this exercise, it is that it's immensely difficult for anyone not currently entrenched in the supply chain to measure drug prices. So we are left wondering, how we will ever manage them?

Ultimately, the proprietary data required to manage our way out of this problem is all out there under lock and key, with an army of supply chain members working to keep it that way. Hopefully this hidden data will soon see the light of the day to build the public’s trust in our nation's efforts to bring drug prices down.