Recent Trends in Medicaid Spending and Use of Drugs With US Food and Drug Administration Accelerated Approval

This cross-sectional study identifies the number and class of drugs approved through the US Food and Drug Administration’s accelerated approval pathway and analyzes state Medicaid programs’ use and spending on these drugs from 2015 through 2019.


I. eMethods
Details: Creation of study sample We constructed a data set at the product-indication level of all AA products approved by the FDA between December 1992-2020, using a biannual report from the FDA's CDER. We cross-checked each approval to CDER's compilation of all new molecular entities and biological approvals from 1985-2019 to identify any products that had a previously approved indication under the traditional approval pathway (i.e. "non-AA indication"). We elected not to include any AA products that were approved by the Center for Biologics Evaluation and Research (CBER), as we determined public access to these data to be unreliable.
We completed a crosswalk on brand names for the full AA list to match each drug to 11-digit NDC codes, using Redbook. When not found in Redbook, we used a publicly available search tool (e.g. ATLAS) to append additional NDC-11 codes. We matched the NDC-11 codes to our cleaned AA list and linked this file to spending and utilization data found at Medicaid.gov. For each year between 1992-2019, we downloaded the "Full Dataset (States + Totals)" and converted each file to a Stata .dta file. To perform our analyses more efficiently, we output a compressed data file that grouped "Total amount reimbursed," "Medicaid amount reimbursed," "non-Medicaid amount reimbursed," "number of units reimbursed," and "number of prescriptions reimbursed" by accelerated approval status, state, year, quarter, and utilization type. For AA drugs, we further grouped by prefix product name and NDC-11 code.

Details: Calculations for inflation-based rebates
To account for the Medicaid inflation penalty, we identified the first year each NDC-11 code was observed between 1992-2019 and estimated the median unit price at the NDC-11 code-level for each succeeding year through 2019. We used the median unit price in that year to represent the baseline average manufacturer price (AMP). For each subsequent year, we inflated the drug's baseline AMP by 2%, calculating a target price in each year from 2015 to 2019. The statutory inflation-based rebated for each NDC-11 code was then estimated as any unit price increase exceeding the 2% trend line from its baseline AMP multiplied by the NDC-11 code's total unit fills. This method enabled us to take into account both price and volume of sales in any rebate adjustment due to inflation. Each product's total annual inflation-based rebate was the sum of inflation-based rebates for each of its NDC-11 codes.
To finalize this method described above, we performed numerous variations to select the parameters that would best estimate each drug's AMP and our ability to account for year-to-year changes in price since market entry. Applying inflation-based rebates at the NDC-11 code-level required using a measure of central tendency to summarize unit prices across quarters in a given year and across utilization context. For this step, we tested using both the mean and the median. After closely assessing unit prices in the 25 th and 75 th percentiles among select AA drugs relative to their mean and median unit prices, we selected to estimate inflation-based rebates on the median unit price to safeguard against any arithmetic bias introduced by outliers. We also carefully considered the year that would represent a NDC-11 code's "baseline AMP," important for estimating inflation-based rebates that would accurately account for changes in price and volume of sales since a drug's market entry. It was critical to begin each NDC-11 code's inflation-adjusted trendline in the first year that is observed in the data, which often immediately follows FDA approval. Finally, since the State Medicaid Drug Utilization files are not indication specific, we used the first year each NDC-11 code was observed between 1992-2019 as the baseline year, even if the first year observed came before the AA indication approval date.

Details: External validation
To validate our findings, we compared our estimated gross and net Medicaid spending amounts to the available MACStats: Medicaid and CHIP Data Books, which include aggregated total gross