Estimates of Medicaid and Non-Medicaid Net Prices of Top-Selling Brand-name Drugs Incorporating Best Price Rebates, 2015 to 2019

Key Points Question Can a new method for estimating Medicaid discounts clarify trends in net Medicaid and non-Medicaid spending and prices for the top-selling brand-name drugs? Findings This cross-sectional study of 18 top-selling brand-name drugs demonstrated a new method to account for the Medicaid best price discount, one component of the total Medicaid discount that is sometimes overlooked by researchers. Including the best price discount reduced the estimated Medicaid drug prices in the sample by 3% in 2019, and up to 54% for individual drugs. Meaning The findings of this cross-sectional study suggest that rigorous and transparent methods to estimate Medicaid discounts can help reveal patterns in prices and spending, which can be helpful in developing strategies for better aligning drug prices with clinical benefits.

The FSS Max Cap caps the FCP in the second and subsequent years of a contract.
We do not estimate this measure; it is used conceptually.

Sample of Selected Top Selling Brand-name Drugs
The selected sample drugs were associated with 68 NDC-9 codes, combined across package sizes, in the 2015-2019 Medicaid utilization data.

Estimating Medicaid Manufacturer Drug Rebates
Estimating Medicaid's Best Price Provision As part of this process, we compared the FSS price to the FCP and Big Four Price to determine when the FSS price was a good estimate of best price for a large sample of 156 products (including the 18 drugs analyzed in the study). Separate FSS and Big Four prices were listed for the majority (58%, weighted by Medicaid spending) of NDC-quarters, implying that the FSS price was a good estimate of best price because it was not capped by the FCP as is the Big Four price. For a share (19%) of NDC-quarters where the published FSS price was equal to the Big Four price, three-quarters of the records listed an FSS price that was the same as or higher than the FSS price from a previous period before the two prices were equal. This situation implies that the Big Four price increased (instead of the FSS price decreasing) to the FCP; for these cases we also assume that FSS prices were a good estimate of best price. For the remainder (23%) of NDC-quarters, there is no published Big Four price, so we used the FSS price to estimate best price even though the FSS price may be based on the FCP.
For the 18 products in this study, we found that single pricing was uncommon, and have thus assumed that the published FSS price reflects the most favored commercial customer discount rather than the FCP. However, single pricing was more common in the broader sample of drugs. If single pricing is more typical, this methodology may underestimate the best price and therefore overestimating the best price rebate.

Computing the Medicaid Basic Rebate including the Best Price Provision and the Inflation-linked Rebate
To estimate Medicaid's basic rebate provisions, we computed the Medicaid rebate at the NDC-9 level, combining across package sizes. We estimated AMP as the average AMP within the NDC-9 group, weighting by gross Medicaid sales. To estimate best price, we first smooth the FSS prices at the NDC-11 level to account for discontinuities and price jumps resulting from five-year contract periods. We then aggregate across package sizes to the NDC-9 level, using the lowest smoothed FSS price among package sizes for the NDC-9 group. We then choose the larger of the published FSS price or the smoothed price per NDC-9. We are assuming an increase in the FSS price from one contract to another reflects a real price increase in the price offered to the most favored customer.
We used the earliest launch quarter in the NDC-9 group (across package sizes), identifying launch quarter by NDC-11 from the Medicaid Drug Rebate Program data. We estimated baseline AMP as 91% of the WAC Basic rebate = larger of: (1) 23.1% AMP or (2) AMP -Best Price Inflation rebate = costs in excess of baseline AMP from launch date inflated by the CPI Total rebate = basic rebate + inflation rebate, capped at 100% AMP from the earliest launch quarter within the NDC-9, using WAC prices listed in SSR Health or Medi-Span data to fill data gaps, which accounted for about half of NDCs. Though SSR Health compute the Medicaid rebate at the product-strength level, we compute it at the NDC-9 level, consistent with CMS and CBO.
To estimate total Medicaid rebate amounts for each drug, we computed the unit rebate from both the basic rebate and inflation-linked rebate as a percent of AMP at the NDC level and applied that to total Medicaid spending. To estimate the total Medicaid rebate amount, we applied the unit rebate amount as a percent of AMP at the NDC-9 level to total Medicaid spending rather than multiplying the unit rebate amount by the number of units reimbursed by Medicaid, as published in SDUD. We do this due to data quality concerns in Medicaid SDUD unit reporting.

Estimating 30-Day Supply Prices
As described in the main methods section, prices were standardized to a 30-day supply used the defined a typical daily dose per product, published by the World Health Organization, 6

Estimating Net Medicaid and Non-Medicaid Spending and Prices for Selected Top Selling Brand-name Drugs
We used net spending and rebate data from SSR Health to estimate spending at WAC prices. As a proxy for Medicaid spending at WAC as a share of total spending at WAC, we estimated a similar ratio, Medicaid's share of gross spending, by dividing Medicaid gross spending from CMS by total gross sales from IQVIA, which is 12.7%, weighted by Medicaid gross spending at the NDC-11 level. We applied this ratio to total spending at WAC to approximate Medicaid and non-Medicaid spending at WAC. As CBO estimates that sales at AMP are equal to 98% of sales at Medicaid retail prices, dividing CMS' Medicaid sales by IQVIA sales is likely to be a good estimate of Medicaid market share.
We then computed Medicaid total net spending (Medicaid gross spending from CMS minus estimated Medicaid rebates) and non-Medicaid total net spending (non-Medicaid total gross sales minus non-Medicaid rebates). Non-Medicaid gross sales was computed as IQVIA total gross sales minus Medicaid total sales from CMS and the non-Medicaid rebate was computed as IQVIA gross sales minus SSR net sales minus total estimated Medicaid rebates. We used these to compute Medicaid and non-Medicaid net spending to total WAC spending ratios. Finally, to estimate Medicaid and non-Medicaid net 30-day price, we applied these ratios to the 30-day WAC price from Medi-Span, using the average 30-day WAC price by product, weighting by gross Medicaid sales.

Additional Methods on the Components of Medicaid Rebate-Related Discounts
The best price provision was substantial for certain drugs. For example, for apixaban (with gross spending of $366.9 million in 2019), the Medicaid savings due to the best price provision was $246 per 30-day supply in 2019, equal to 54% of the gross price for a 30-day supply on top of the basic rebate of 23%. For ustekinumab (with gross spending of $474.9 million in 2019), the Medicaid savings due to the best price provision was $2,687 per 30-day supply in 2019, or 16% of the gross price for a 30-day supply on top of the basic rebate of 23%. Of the 18 brand-name drugs assessed, 6 to 7 were reduced by the best price provision in each year and between 11% and 22% of gross Medicaid spending was associated with drugs that triggered the best price provision. Medicaid wholesale acquisition cost (WAC), 30-day supply, $(25th, 50th, 75th percentile) 3  Notes: Wholesale Acquisition Cost (WAC); Average Wholesale Price (AWP); National Drug Code (NDC). All spending, prices and rebates are weighted by average quarterly gross Medicaid spending at the NDC11 level in the 5-year period to compute annual estimates by drug. Drugs are aggregated to an annual index, weighting by average gross Medicaid spending in the 5-year period. Prices were standardized to a 30-day supply by product in cases in which the prescription exceeded 30 days. See text for more details on methods. 3 We extracted WAC at the NDC-9 level from SSR Health and Medi-Span; author's analysis of overlap data showed WAC prices in SSR Heath and Medi-span are highly consistent. 4 Medicaid gross price is estimated AMP. See text for more details on methods. 5 The best price provision was triggered by Eliquis, Enbrel, Januvia, Jardiance, Remicade, Stelara, Symbicort, and Vyvanse over the study period. Counts of the number of 18 top selling brand-name drugs estimated to trigger the best price provisions and the inflation cap counts the number with at least one NDC that triggers the provision.

eLimitations
The estimated non-Medicaid market price does not necessarily reflect exact Medicare Part D and commercial prices because the SSR health data include the effect of discount cards, coupons, the 340B program, VA and Department of Defense purchasers with access to statutory discounts. These limitations may explain the two cases in which we estimated that the non-Medicaid net price was below the Medicaid net price (empagliflozin and budesonide/formoterol in some years). An additional data limitation that should be explored in future research is that SSR-based estimates of non-Medicaid net prices that do not account for Medicaid rebates as we have in this study are generally far below our estimated best prices. Particularly in the later years, these non-Medicaid net prices may not always reflect pricing dynamics in the private market by Part D and commercial plans for the study drugs. Related to this point, we find that from 2017 to 2019 Medicaid net prices increase more quickly than non-Medicaid net prices (after dropping Humira) even though most of the non-Medicaid market does not enjoy Medicaid's inflation rebate protections. SSR Health estimates prices using the assumption that AMP is 3% lower than WAC, rather than 9%, which is used in this analysis (and CBO research 8 ), thus estimates were difficult to compare. In addition, there may be variation in the relationship between AMP and WAC that may affect analysis of price trends over time as drugs differ in composition over years, although CBO finds that the relationship is fairly consistent across subgroups (such as specialty, nonspecialty, and high-priced drugs). For the 18 products in this study, we found that single pricing appears to be relatively uncommon, and have thus assumed that the published FSS price reflects the most favored commercial customer discount rather than the FCP. However, if the FSS actually reflects FCP, we may be estimating too low of a best price and therefore overestimating the best price rebate relative to AMP. We may also be underestimating the best price rebate if some rebates are not reflected in the FSS price. In addition, spending on physician administered drugs may not be captured entirely in these data to the extent that they are paid through medical claims rather than pharmacy claims. Physician administered drugs may also be subject to a different AMP calculation, which can include discounts, rebates, and payments in addition to sales. For such drugs, our AMP estimate based on WAC price is likely to be an overestimate, which may result in an overestimate of the rebate. Finally, the sample of 18 top-selling drugs studied here accounted for only 8% of Medicaid spending, potentially limiting the generalizability of results.