Trends in Use and Cost of Second-Generation Antipsychotics Among Children and Teens in Indiana Medicaid, 2004–2012

This was a retrospective, longitudinal survey of administrative claims data in the Indiana Medicaid claims database. The study population included all Indiana children ages 18 or younger who had Medicaid coverage for at least one month during any calendar year of the study period (2004–2012). In early analyses, we determined that there was minimal use of first-generation antipsychotics in our cohort. Hence, we focused our analyses on utilization and cost of second-generation antipsychotics (overall and of aripiprazole). Use was defined as at least one claim for a qualifying medication during the calendar year in question. Data from over 1.4 million unique children enrolled in Medicaid from January 2004 to June 2012 were extracted for analysis. The list of medications surveyed included aripiprazole, asenapine (sublingual), clozapine, iloperidone, lurasidone hydrochloride (HCL), olanzapine, olanzapine pamoate, paliperidone, paliperidone palmitate, quetiapine fumarate, risperidone, risperidone microspheres (intramuscular [IM]), ziprasidone HCL, and ziprasidone mesylate IM.

The variables of interest were the monthly cost of second-generation antipsychotics (average per member per month [PMPM] Medicaid reimbursement for all in-class medications, including aripiprazole, among users of a second-generation antipsychotic), aripiprazole (average PMPM Medicaid reimbursement for aripiprazole among users of aripiprazole), and second-generation medications excluding aripiprazole (average PMPM Medicaid reimbursement for in-class medications excluding aripiprazole among users of all in-class medications except users of aripiprazole); status as a ward of the state or foster child per calendar year (defined as having been a ward or foster child for at least six months per calendar year); antipsychotic safety benchmarks in the AHRQ-Rutgers study, including use of multiple antipsychotics, average count of unique antipsychotics used per member per calendar year (among those using at least one), counts of children ages five and under using an antipsychotic per calendar year, and counts of children using an antipsychotic with an ICD-9 diagnosis code for diabetes per calendar year; membership in a Medicaid subprogram, aid category (disabled, State Children's Health Insurance Program, Temporary Assistance for Needy Families [TANF], and other), and level of care for the plurality of months in each calendar year; and psychiatric diagnostic status per member per year (presence or absence of ICD-9 psychiatric diagnostic categories and codes for mental retardation [317, 318.xx, 319], pervasive developmental disorders [299.xx], attention-deficit hyperactivity disorder [ADHD] and disruptive behavior disorders [ADHD, 314.xx; conduct disorders, 312.xx; and oppositional defiant disorder, 313.81], oppositional defiant disorder [ODD; 313.81], schizophrenia and psychotic disorders [295.xx, 298.8, 293.xx, and 298.9], mood disorders [296.xx and 311], adjustment disorders [309.xx], anxiety disorders [300.xx and 308.3], and antisocial [301.7] and borderline [301.83] personality disorders and personality disorder NOS [301.9]).

A primary goal of the statistical analysis was to assess the impact of Medicaid policy changes (the “intervention”) on utilization and cost of second-generation antipsychotics. Analysis of the effects of these changes was performed initially by inspection of yearly and monthly cost and utilization trends across all of the study years (2004–2012) and then by a separate examination of data for 75 months. The 75-month period was divided into three periods, defined as preintervention (October 2005–December 2007), intervention (January 2008–March 2010), and postintervention (April 2010–December 2011). Generalized estimating equations (GEEs) were used to determine if, for the variables of interest, the mean response values differed across the considered time periods. GEE methods are universally applicable to the correlated data arising from the exponential family of distributions, for example, repeated observations of dichotomous count data. As long as the mean models are correctly specified, the assumed (working) correlation structure among the clustered observations is used as a component of the so called “sandwich-variance estimator,” which assesses uncertainty of the estimated means. The additional component is derived from the empirical covariances obtained from the original observations. In our application, we specifically used a linear model with the working exchangeable correlation structure to estimate the mean rates of utilization for the groups. All analyses were performed using SAS, version 9.2.

Overall, our data indicated a high rate of enrollment in Medicaid among Indiana children, which increased over the study period. This may have reflected the impact of the U.S. economic recession on Indiana families, a notion that is further supported by the increase in children in the TANF aid category from 2004 to 2012 (Table 1). Another notable trend was the increase in non-FFS patients among children covered in Indiana Medicaid. These numbers grew from 400,000 in 2004 to 520,000 in 2012, reflecting the increase in RBMC enrollment and emergence of new RBMC programs, such as Care Select. In contrast, FFS Medicaid enrollment remained relatively flat, decreasing from just under 300,000 members in 2004 to approximately 250,000 members in 2012.

Utilization rates for second-generation antipsychotics among all patients (FFS and non-FFS) were similar to those in the AHRQ-Rutgers FFS cohort (approximately 2% versus 1.6%, respectively). Notably, rates of utilization of second-generation antipsychotics in Indiana increased steadily between 2004 and 2008 and leveled out thereafter (2008–2012). [Figure 1 in the online supplement shows utilization of second-generation antipsychotics by FFS status. An analysis of monthly utilization indicated that there was growth in the number of patients using aripiprazole and all other second-generation antipsychotics until April 2009, followed by a leveling out thereafter [see Figure 2 in the supplement]. In addition, male children were more likely than female children to be administered an antipsychotic, another finding that was similar to the AHRQ-Rutgers data (Table 1). Although children from racial-ethnic minority groups were overrepresented in Indiana Medicaid compared with their representation in the overall state population, they were less likely than white children to receive an antipsychotic (Table 1).

There was a statistically significant difference in utilization rates between the preintervention period and the intervention period (F=118.07, df=1 and 52, p<.001) and between the preintervention period and the postintervention period (F=206.15, df=1 and 46, p<.001) [see Figure 3 in the online supplement]. There were no statistically significant differences in utilization rates between the intervention period and the postintervention period. Therefore, collectively, the OMPP program changes were associated with containment of antipsychotic utilization rates. Furthermore, the analyses provided evidence of containment of PMPM total costs for antipsychotics and of PMPM costs for antipsychotics excluding aripiprazole. In contrast, PMPM costs for aripiprazole continued to climb [see Figure 4 in the online supplement].

As expected from the AHRQ-Rutgers work, we observed consistently high rates of antipsychotic utilization (13%–18%) among foster children enrolled in Medicaid (Table 1). To better understand the much higher use of antipsychotics among foster children, we surveyed psychiatric diagnosis rates annually (2007–2011) among foster and nonfoster children. Among nonfoster children, the distribution of psychiatric diagnoses was as follows: ADHD and disruptive behavior disorders, including ODD (8%−9%), mood disorders (4%), anxiety disorders (3%−4%), pervasive developmental disorders (1%), mental retardation (<1%), and schizophrenia (<1%). These rates are similar to those reported in U.S. community surveys of psychiatric diagnosis prevalence (2,3). In contrast, psychiatric diagnosis rates for foster children were two to three times higher compared with results for U.S. community surveys across all categories: ADHD and disruptive disorders (23%−28%), mood disorders (15%−21%), anxiety disorders (12%−14%), mental retardation (2%−4%), pervasive developmental disorders (2%−3%), and schizophrenia (3%). Among foster children with a diagnosis of ADHD or disruptive disorders, 23%−29% were on an antipsychotic compared with 10% to 13% of nonfoster children with one of these diagnoses. This use pattern was also consistent across the other major diagnostic categories studied. For example, 36% to 42% of foster children with a mood disorder were on an antipsychotic compared with 23% to 26% of nonfoster children.

Data concerning use of second-generation antipsychotics among children enrolled in Indiana Medicaid were consistent with the earlier findings of the Rutgers consortium. Important similarities between the studies included similar overall rates of antipsychotic utilization, with both FFS and non-FFS patients included in the Indiana study. As noted previously, males tended to have higher antipsychotic exposure rates, likely due to the higher rates of ADHD and disruptive disorders among boys (2). Children from racial-ethnic minority groups were overrepresented in Indiana Medicaid as a whole, but antipsychotic use rates were lower (less than half) among children from racial-ethnic minority categories (black, Hispanic, or other) compared with white children. This finding is in contrast to recently published data from other states, where growth in use of antipsychotics was greater among African-American youths compared with other youths (4,5).

OMPP policy changes occurring after October 2007 were associated with constrained antipsychotic utilization in 2008–2012, an effect that was highly statistically significant. The same trend was true for growth in overall PMPM cost estimates for antipsychotics. These results were all the more striking given the recent approval by the U.S. Food and Drug Administration (FDA) of newer antipsychotics—or new indications for older antipsychotics—for treatment of autism among children and teens, bipolar disorder among teens, and depression among adults who have not responded to selective serotonin reuptake inhibitors.

As expected from the AHRQ-Rutgers findings and other surveys (6,7), we observed high rates of antipsychotic use (13%–15% per year) among foster children enrolled in Medicaid. This finding was not explained by higher rates of psychopathology among foster children, given that utilization among foster children remained higher across most of the major diagnostic groupings after accounting for psychiatric diagnoses among foster and nonfoster children. Also, large numbers of children received antipsychotics for disorders for which there often was no clear FDA approval for pediatric use, for example, ADHD and disruptive disorders. The scientific case for the systematic use of second-generation antipsychotics for pediatric conditions, such as externalizing disorders and aggressive behaviors, is by no means clear, although this practice is clinically widespread (8–11). Future development of an antipsychotic prescribing algorithm for externalizing disorders is indicated.

We wish to draw attention to several important limitations of this study. The retrospective nature of this analysis carried inherent limitations. The administrative claims database did not provide patient- or provider-level information about the status of individual patients, which is needed to assess the clinical appropriateness of antipsychotic prescribing. Finally, the analytic approach that we employed did not permit us to determine the impact of individual program changes on cost and utilization of second-generation antipsychotics, was not specifically adjusted for FFS status, did not cover all classes of antipsychotics, and did not evaluate brief versus longer-term antipsychotic use.

Our survey of prescribing of second-generation antipsychotics among youths enrolled in Indiana Medicaid was generally consistent with observations initially reported by the AHRQ-Rutgers consortium. A series of policy and programmatic changes within Indiana Medicaid, some specifically designed to stimulate appropriate use of and inhibit inappropriate use of antipsychotics, appeared effective in containing utilization and costs.

The authors thank the MMAC contributors for their feedback on this work as well as Roberta Ambuehl, B.S., for data extraction.