Forecasting Medicaid Expenditures for Antipsychotic Medications

Spending reductions resulting from an ongoing transition from use of branded to generic second-generation antipsychotics may have critical implications for Medicaid policy. In October 2008, risperidone became the first second-generation antipsychotic with high sales volume to become available generically (1). Over the next four years, several additional second-generation antipsychotics came off patent, including olanzapine, quetiapine, ziprasidone, paliperidone, and aripiprazole (1). Patent protections for most other current second-generation antipsychotics, including asenapine, iloperidone, and lurasidone, are expected to expire by 2016 (1). The resulting reduction in Medicaid expenditures for antipsychotic medications could substantially affect Medicaid medication budgets and the need for formulary restrictions on access to second-generation antipsychotic medications.

The market entry of second-generation antipsychotics beginning in the 1990s resulted in sharply higher Medicaid expenditures for antipsychotic medications (2), as these branded medications largely replaced generic first-generation antipsychotics in schizophrenia treatment. Between 1999 and 2005, annual antipsychotic expenditures per Medicaid beneficiary more than doubled, while antipsychotic prescriptions per beneficiary increased 30% (2). By 2009, antipsychotics accounted for nearly 15% of all Medicaid expenditures for medications (3). Policy makers’ concerns about increased spending on second-generation antipsychotics led many states to impose formulary restrictions on second-generation antipsychotics (2,4–7). However, this rationale for formulary restrictions for antipsychotics may be less relevant in an era when the availability of generic versions of most second-generation antipsychotic medications with high sales volume promises to substantially reduce Medicaid expenditures on second-generation antipsychotics.

Although the pharmaceutical industry and government agencies have published estimates of savings from the entry of generic versions of other types of medications (8–11), published estimates of the implications of generic second-generation antipsychotics on expenditures are not available to date. Using aggregate data on antipsychotic prescriptions and spending in Medicaid, this study developed estimates of the savings to the Medicaid program from use of generic risperidone during the years 2008 to 2011 and used these estimates to derive forecasts (projections) of Medicaid expenditures for all antipsychotic medications (12).

Methods

Data

Information on number of and expenditures for Medicaid outpatient prescriptions for antipsychotic medications was drawn from Medicaid state drug utilization data for 2008 to 2011 (12). The database contains quarterly totals (N=816) for all 50 states and the District of Columbia by medication. Records are listed by National Drug Code (NDC) and include the product name. Generic medications were differentiated from branded medications by using the product name. Risperidone records included risperidone in tablet and oral solution form as well as a long-acting injectable (Risperdal Consta) and an extended-release wafer (Risperdal M-Tab). These different forms were identified by using their corresponding NDC. All expenditure amounts were converted to 2011 constant dollars by using the Consumer Price Index for All Urban Consumers (13).

To account for manufacturer rebates to Medicaid, expenditure totals were reduced by 28% for branded medications and 24% for generics (14). No prescriptions data were available for 21 (2.6%) of all 816 possible combinations of state and quarter. Values for 13 of these 21 missing records were imputed by using the previous quarter’s observation carried forward. Values for the remaining eight missing records, all of which pertained to one state that had no information recorded for 2008 or 2009, were imputed by using values for that state from 2010, which were then adjusted for time trends in aggregate Medicaid antipsychotic use between 2008 and 2010.

Projected Savings From Use of Generic Risperidone

Projected expenditures for risperidone had patent protection continued were calculated first by using a constant price and then by using a variable price. In the constant-price model, it was assumed that the price per unit of branded risperidone stayed constant from the second quarter of 2008, when risperidone’s patent protection ended, until the end of 2011. This price was multiplied by the actual number of prescriptions for generic risperidone during this period, and the product was added to expenditures for prescriptions of branded risperidone to obtain total projected expenditures for risperidone under continued patent protection.

However, the constant-price assumption is contrary to evidence that the prices of branded medications tend to increase over time at a faster rate compared with the prices of generics (15,16). Consequently, in the variable-price model, it was assumed that the price per unit of branded risperidone would have changed at the same rate annually compared with the weighted average price of prescriptions for branded antipsychotic medications. Branded antipsychotic medications included aripiprazole, asenapine, clozapine, iloperidone, lurasidone, olanzapine, paliperidone, quetiapine, risperidone, and ziprasidone. The average price of a prescription for each antipsychotic was obtained by dividing Medicaid expenditures by the number of prescriptions for each medication.

Forecast of Antipsychotic Expenditures

Forecasts of Medicaid expenditures for antipsychotic medication in calendar years 2016 and 2019 were developed by using data on trends in risperidone expenditures and information on Medicaid expenditures for other second-generation antipsychotics and first-generation antipsychotics. Calendar year 2016 is the expected year of patent expiration for lurasidone, the most recently approved of the currently used antipsychotic medications. Projections were also made for 2019 because an initial examination of the study data suggested that once the patent for a branded medication expires, it takes approximately three years for the generic form to account for 99% of all Medicaid prescriptions for that medication.

To develop forecasts, three modeling assumptions were necessary. First, it was assumed that a generic version of each second-generation antipsychotic would gain approval one year after patent expiration. This is consistent with recent prior experience for several second-generation antipsychotics (1). Second, it was assumed that Medicaid’s average costs for generic second-generation antipsychotics would decrease at the same rate as Medicaid’s average costs for generic risperidone. Finally, it was assumed that prescriptions for generic versions of second-generation antipsychotics would displace branded prescriptions at the same rate as generic risperidone replaced the branded version.

Projected Medicaid Enrollment Under the ACA

Although forecasts of antipsychotic expenditures were generated by holding Medicaid enrollment constant, Medicaid enrollment is expected to increase substantially as a result of the 2010 Affordable Care Act (ACA) and other demographic trends. As a sensitivity analysis, projected Medicaid enrollment increases (17) were incorporated into an alternate forecast of antipsychotic spending. To do this, an additional assumption was made regarding the proportion of additional Medicaid enrollees who will be prescribed an antipsychotic medication. The ACA may have affected this proportion by greatly expanding categorical eligibility for Medicaid. Although antipsychotic use in Medicaid is highly concentrated among the 17% of Medicaid enrollees in disability enrollment categories (3), the Medicaid enrollment expansions authorized by the ACT pertain predominantly to persons who are not eligible for Medicaid under a disability category. Consequently, antipsychotic prevalence is likely to be lower in the expansion population than in the traditional Medicaid population.

The sensitivity analysis was based on federal actuarial projections that total Medicaid enrollment will increase by 22 million individuals (38%) between 2011 and 2019, while disabled group enrollment will increase by only .6 million individuals (6.3%) over the same period (17). It also was assumed, on the basis of prior analysis of Medicaid claims data (18), that 24.6% of disabled enrollees and 1.9% of nondisabled enrollees in Medicaid were prescribed an antipsychotic.

Results

Changes in Risperidone Use and Cost

Between 2008 and 2011, use of generic risperidone increased by 2.9 million prescriptions (386%), from .7 million prescriptions in 2008 to 3.6 million prescriptions in 2011 (Table 1). Meanwhile, use of branded risperidone (not including Risperdal Consta or M-Tab) decreased by 2.07 million prescriptions (99%), from 2.1 million prescriptions in 2008 to .03 million prescriptions in 2011. Overall, as a percentage of all second-generation antipsychotic prescriptions, use of risperidone decreased slightly, from 28.4% in 2008 to 27.8% in 2011.

During the same time frame, expenditures per prescription for generic risperidone decreased by 80%, from $166 in 2008 to $33 in 2011. This rapid decrease in generic prices helps explain apparent anomalies between prescription volume trends and expenditure trends during this period. For example, even though the volume of prescriptions for generic second-generation antipsychotics increased by 36% (from 2.8 to 3.8 million) between 2009 and 2010, expenditures for generic second-generation antipsychotics decreased by 13% (from $208 to $180 million) during the same time frame because of a sharp decrease in average prices for a prescription. Meanwhile, the cost of a prescription for branded risperidone increased 10%, from $204 in 2008 to $224 in 2011, and the average cost of a prescription for a branded second-generation antipsychotic increased 34%, from $265 in 2008 to $356 in 2011. Total expenditures for risperidone (the sum of expenditures for generic and branded risperidone) decreased 77% over the study period, from $546 million in 2008 to $124 million in 2011.

Projected Savings From Risperidone

Table 2 shows projected Medicaid savings from generic risperidone at constant and variable prices. The variable prices were calculated by using the annual percentage change in the average cost per prescription of branded second-generation antipsychotics (Table 1). Risperidone’s projected price increased annually and, by 2011, exceeded the constant price by $68 (or 34%) (annual price changes are available from the authors). Under the assumption of constant prices for prescriptions for branded risperidone, access to generic risperidone resulted in an estimated savings of $25 million in 2008, $315 million in 2009, $524 million in 2010, and $599 million in 2011 (Table 2). Under the assumption of variable prices, estimated savings were $25 million, $384 million, $524 million, and $844 million, respectively.

Forecast of Antipsychotic Expenditures

Overall, in 2011 Medicaid spent $3,676 million on outpatient prescriptions for antipsychotic medications (Table 3), and prescriptions for second-generation antipsychotics accounted for nearly all (98.9%) of this total. Future Medicaid expenditures for second-generation antipsychotics were forecast by using Medicaid spending for outpatient prescriptions for antipsychotics in 2011 (Table 3) and information about risperidone prescriptions between 2008 and 2011 (Table 1). Medicaid’s average cost per risperidone prescription decreased by $167 (83.5%) between 2008 and 2011, from $200 per prescription for branded risperidone in the first quarter of 2008 to $33 for generic risperidone in 2011. To develop a forecast, it was assumed that Medicaid’s average costs for other generic second-generation antipsychotics would similarly decrease by 83.5% within three years of generic entry. It also was assumed that generic prescriptions would account for 99% of all prescriptions for a given medication within three years of generic entry, which, as noted earlier, occurred with generic and branded risperidone. Given these assumptions, Medicaid expenditures were forecast to decrease by an additional $1,794 million (48.8%) by the year 2016 and by $2,814 million (76.5%) by the year 2019. This would imply that Medicaid expenditures on outpatient antipsychotic prescriptions would total $1,882 million in 2016 and $862 million in 2019.

Sensitivity Analysis

Federally projected increases in Medicaid enrollment between 2011 and 2019 are likely to result in greater numbers of enrollees’ filling antipsychotic prescriptions, thereby offsetting the decrease forecast in Medicaid expenditures for antipsychotic medications (Figure 1). A sensitivity analysis that incorporated these projected enrollment increases suggested that the number of Medicaid enrollees prescribed an antipsychotic may increase by 16.3% by 2016 and by 20.5% by 2019. As a result, total Medicaid expenditures on outpatient antipsychotic prescriptions, adjusted for enrollment trends, may decrease to $2,189 million by 2016—a decrease of 40.4% from 2011—and to $1,038 million by 2019—a decrease of 71.7% from 2011.

Discussion

Within the next three or four years, the makeup of antipsychotic prescriptions in Medicaid will likely undergo a transition from predominantly branded medications to predominantly generic medications. In 2011, Medicaid spent more than $3.6 billion on second-generation antipsychotics. Five branded medications—aripiprazole, quetiapine, olanzapine, ziprasidone, and paliperidone—accounted for the vast majority ($3.3 billion [90%]) of this spending. According to this study’s forecasts, patent expirations for these medications could result in a further reduction in Medicaid expenditures for antipsychotic medications of $1.8 billion (49%) by the year 2016 and of $2.8 billion (77%) by the year 2019.

Anticipated reductions in spending on antipsychotics should be factored into reassessments of the ongoing need for Medicaid formulary restrictions on second-generation antipsychotics. Medicaid has historically accounted for 70%−80% of all antipsychotic prescriptions in the United States (19,20), and rapid increases in antipsychotic expenditures following the introduction of branded second-generation antipsychotic medications in the 1990s led some states to impose formulary restrictions on access to these medications in Medicaid (2,4–7). However, at best, these policies result in little or no net savings to Medicaid (21) and, at worst, adversely affect patients’ medication access (22–24) and outcomes (7,21). Consequently, if Medicaid is entering a period of rapidly decreasing antipsychotic expenditures, states’ original rationale for formulary restrictions may be significantly eroded.

Actual reductions in Medicaid spending on antipsychotic medications could be lower than forecast, as a result of increased antipsychotic prescription volume in states where Medicaid program eligibility was expanded because of the ACA. A sensitivity analysis that incorporated actuarial estimates of future Medicaid enrollment levels indicated that although enrollment increases will result in a moderately greater volume of antipsychotic prescriptions, the rapid decrease in cost per prescription will drive Medicaid expenditures lower over time. Antipsychotic prescription volume will likely increase only moderately as enrollment increases, because individuals who qualify for Medicaid under a disability enrollment category, among whom antipsychotic use is concentrated (3), will be underrepresented in the expansion population compared with existing Medicaid enrollees (17).

The pace of the actual spending reduction will also depend on the extent of generic substitution in Medicaid and approval of new branded medications, among other factors. After generic risperidone entered the market, generic prescriptions rapidly achieved a nearly 100% share of all risperidone prescriptions for the same medication, but risperidone’s share of all prescriptions for second-generation antipsychotic remained virtually unchanged (Table 1). This pattern suggests that as generic forms of branded second-generation antipsychotics become available, they will make up an increasing fraction of prescriptions for all second-generation antipsychotics, leading to a sharp decrease in total spending for second-generation antipsychotics. However, this long-term decline in spending for second-generation antipsychotics could be slowed or even reversed by aggressive marketing by the pharmaceutical industry of reformulations of second-generation antipsychotic medications or by new market approval of one or more novel antipsychotic medications. However, a recent survey of the psychotropic medication development pipeline suggests that this is unlikely (25). No novel antipsychotic medications are expected to enter the market anytime soon, given that investments by major pharmaceutical manufacturers in the development of drugs for mental disorders has waned dramatically in recent years (25).

Evidence of savings from the use of generic risperidone indicates that expiration of patents for second-generation antipsychotics will be associated with a large financial windfall for Medicaid. Medicaid savings from the use of generic risperidone increased annually between 2008 and 2011, from $25 million in 2008 to at least $599 million in 2011. Savings increased over the study period as a result of an 84% decrease in Medicaid costs per prescription for generic risperidone, a nearly 100% conversion from prescriptions for branded forms to generic forms of risperidone, and an overall increase in use of risperidone and other second-generation antipsychotic medications.

This is the first study to use a variable-price model to derive estimates of projected savings from the use of generic medications. Savings from risperidone use in 2011 could have been as high as $844 million if variable pricing is used to calculate the costs of branded risperidone. Variable pricing assumes that the projected price of branded risperidone increased at a rate proportional to the increase in prices for branded second-generation antipsychotics. Previous reports by the pharmaceutical industry and federal agencies of savings from use of generic medications have generally used constant-price models (8–11). However, the variable price model is more consistent with empirical evidence (15,16). The prices of branded medications have tended to increase over the course of their patent protection period, perhaps as a response to decreasing price sensitivity among payers (15), whereas the prices of generic medications have tended to decrease over time (16). Nevertheless, both models are used to project counterfactual expenditure levels that are not directly observable, so whether a variable-price model results in a more accurate estimate of actual savings compared with a constant-price model cannot be verified.

This study had limitations that may affect the interpretation of the results. First, only outpatient prescriptions were recorded in the database used for this study. The omission of inpatient prescriptions resulted in underestimation of overall Medicaid expenditures for antipsychotic medications. Second, this study relied on aggregate data reported by state Medicaid agencies. The completeness of these data is unverified, and some states had missing information that was imputed. A sensitivity analysis conducted without these imputed values found slightly lower overall spending levels and savings compared with the main analysis, but both analyses found very similar expenditure trends. Third, the average manufacturer rebate rate used to adjust Medicaid expenditure amounts was drawn from a government report (14). Actual rebate amounts for specific medications and states are not publicly available and may have deviated substantially from this average. Finally, some of the results may be sensitive to the study’s time frame. The study time frame overlapped with an economic recession, during which the rate of new Medicaid enrollment increased temporarily (17), and preceded an expected surge in Medicaid enrollment (17). A sensitivity analysis suggested that these enrollment changes were unlikely to have more than a moderate impact on the study’s results, because the transition to use of much lower cost generic medications is the predominant influence on current spending trends.

Conclusions

Within the next five years, antipsychotic expenditures in Medicaid may decline to less than half of their current levels, as a result of the ongoing transition from use of branded to generic antipsychotic medications. Although many factors, such as increases in Medicaid enrollment pursuant to the ACA, increased marketing of branded medications by pharmaceutical manufacturers, and unanticipated changes in medication availability, could partially offset this spending decline, reduced average medication acquisition costs associated with the transition to use of generics will likely be the dominant factor affecting the overall spending trend. Such a reduction in spending warrants a reassessment of the continuing need for Medicaid formulary restrictions on second-generation antipsychotics.