Brand name drugs are typically more expensive than generic versions of the same drug, which in general have identical therapeutic effects. The Food and Drug Administration (FDA) evaluates and approves the bioequivalence of generic drugs. Controversy persists about the bioequivalence of a handful of medications, but nearly all other generic drugs provide identical therapeutic benefit.

State Medicaid programs provide health care coverage to those defined as either "categorically needy," whose coverage is federally mandated, or "medically needy," as determined by each state. Medicaid coverage includes prescription drugs in every state. Since 1987, the Health Care Financing Administration has set upper limits on Medicaid payment for certain drugs which are generically available from multiple sources and have been deemed therapeutically equivalent by the FDA (Nightingale 1998). These prices are referred to as Federal Upper Limit (FUL) prices, and are aimed at encouraging use of generic drugs in state Medicaid programs. States have some discretion in how they apply these price formulas; the various state payment plans are referred to as Maximum Allowable Cost (MAC) programs (National Pharmaceutical Council 1998). For drugs with these limits, Medicaid will only reimburse the pharmacy for the MAC price, regardless of whether generic or brand name product is dispensed. There may be exceptions to these limits if the prescribing physician indicates that the brand name drug is medically necessary. Federal law also requires drug manufacturers to negotiate rebate arrangements with state Medicaid programs; these rebates are based on the average manufacturer's price for a drug and are generally slightly higher for brand name than for generic drugs (National Pharmaceutical Council 1998).

Using aggregated state-level reimbursement data from the national Medicaid program, we sought to estimate potential savings from broader use of generic drugs and assess differences across states in spending on brand name and generic medications.

METHODS

Data Source

State-level data on prescriptions filled and amounts paid through most state Medicaid programs were obtained from data made available by the Health Care Financing Administration (HCFA) (2000). Data for 2000 were available for 48 states and the District of Columbia. No data were available for Arizona and Rhode Island. Some states provided data for only two of three quarters; in these cases we multiplied the available data as needed to yield an annualized projection. Data classifying drugs by generic entity, therapeutic class, manufacturer, formulation, and strength were obtained from the National Drug Data File (NDDF) (First Data Bank 2000).

Description of Variables

The HCFA data on state-level Medicaid drug spending were provided by state, year, and quarter. For each state Medicaid program, the data were categorized by the National Drug Code (NDC) and included the total number of prescriptions filled, units of medication dispensed, and amount paid by the state's Medicaid program for each distinct product. The NDC numbers were used to link the records to drug-specific information from the NDDF, such as the ingredient(s) in a given drug, formulation (e.g. standard tablet, timedrelease capsule, transdermal patch), strength, manufacturer, and the package size from which the medication was dispensed. An additional variable categorized whether a given NDC number represented a generic drug or a brand name drug.

Calculation of Generic Prices and Potential Savings

Calculation of generic prices and potential savings was limited to medications dispensed as pills, tablets, or capsules. For drugs available in both short-acting and long-acting forms, brand name drugs were included in the substitution calculation only if there were generic drugs available in the identical form. To ensure that substitutions were clinically reasonable, drugs were grouped by ingredient(s), formulation, strength, and package size. Package size was included because some generic medications are dispensed in very large packages of several thousand units while brand name medications are generally dispensed in much smaller packages. This difference in package size might magnify the difference in per-unit price calculated beyond the disparity between the costs of the medications themselves.

For brand name prescriptions with identical generic alternatives that met the above criteria, we calculated the reimbursement per unit (i.e., the price for a single tablet, pill, or capsule) for a comparable generic product, based on the lowest cost generic NDC code that was in common use in that state. Although we did not have direct access to each state's MAC prices, the lowest commonly used generic price per unit should provide a good approximation of the MAC price.

We then calculated the potential savings for each brand name NDC that met our substitution criteria. Multiplying the number of units dispensed in all the prescriptions for the brand name NDC by the price per unit of the generic alternative yielded the cost that would have applied if available generic drugs had been dispensed instead of brand name preparations of the same medication. We had in our data the amount reimbursed and the units dispensed for brand and generic NDCs. Using these variables, we calculated the potential savings from more consistent use of generic drugs for each individual medication in the following way:

(1) Price per [unit.sub.generic] = Amount [reimbursed.sub.generic] / Units [dispensed.sub.generic]

(2) Potential [savings.sub.brand] = Amount [reimbursed.sub.brand] -(Units [dispensed.sub.brand] X Price per [unit.sub.generic])

Some generic medications, such as enalapril and sotalol, first became available during the 2000 calendar year. For these medications we prorated the calculated potential savings, reducing the potential savings by the fraction of the year during which the generic version was not yet available. We did not incorporate manufacturer rebates into these calculations.

We performed an additional analysis to calculate the potential savings if the best MAC prices from a given state had been available nationwide. For each generic entity, we compared costs across all states to find the lowest per-unit price for all generic drugs in common use. We then repeated the calculation of potential savings as shown above, using the lowest per-unit MAC price across all states instead of the lowest generic per-unit price from within the state.

RESULTS

For 2000, the total amount reimbursed by Medicaid nationally in the 49 studied states was slightly over $20.9 billion. Of that total, about $4.3 billion (18.0 percent) was for medications that were available in both brand name and generic forms. In these 49 Medicaid programs, we identified potential savings of $229 million from use of generic drugs, representing 6.1 percent of expenditure on drugs available in both generic and branded forms and 1.1 percent of total drug spending. There was considerable variation among states, from a low potential savings of 3.3 percent of spending on such generically available drugs to a high of 10.3 percent of spending on generically available drugs. Table 1 lists the spending on generically available drugs and the calculated potential savings by state for 2000.

Our analysis of the potential savings using the best MAC price available across states is summarized in Table 2. The total potential savings almost double, to a total of $450 million dollars, representing 11.9 percent of spending on drugs available in both brand and generic forms and 2.1 percent of total drug expenditures. The amount by which the potential savings increase varied from state to state. For example, the potential savings in Ohio increased by 29.6 percent ($3.4 million in absolute terms) while the amount in California more than doubled (absolute increase of $31.7 million).

Several medications had particularly high amounts of potential savings from generic substitution. In 2000, use of generic clozapine would have yielded potential savings of $23.1 million (11.4 percent of reimbursement for that product); levothyroxine would have produced potential savings of $17.7 million (25.8 percent of reimbursement). Table 3 lists the ten medications with the highest amount of unrealized savings in 2000 in the first three columns. The combined potential savings for these 10 medications was over $109 million, or 47.8 percent of the total potential savings from all medications. The last two columns of Table 3 show the potential savings for these ten medications in the analysis using the best MAC price available. As with the state level results shown in Tables 1 and 2, the changes are not uniform across drugs, so that the potential savings for clozapine and enalapril increase more than 150 percent while the potential savings for alprazolam are almost unchanged.