The use of antipsychotic medications appears to be increasing among very young children enrolled in Medicaid or in private insurance plans. Zito and colleagues (1) reported a fivefold increase in antipsychotic prevalence among children up to four years old when rates observed in two Medicaid programs in 1995 were compared with the rates observed in seven Medicaid programs in 2001. Patel and colleagues (2) observed increases of over 100% in annual prevalence of antipsychotic use per 1,000 enrollees among children ages two to four years in three of four Medicaid programs from 1996 to 2002. In a study of privately insured children ages two to five years, Olfson and colleagues (3) reported that the accumulated rate of antipsychotic use per 1,000 children up to four years old increased from .78 in 1999–2000 to 1.59 in 2007.

Some data suggest that antipsychotic use can be effective in treating very young children exhibiting symptoms of irritability in the context of autism (4), bipolar disorder (5), and disruptive behavior disorders (6), and antipsychotics are generally well tolerated in the short term. At the same time concerns have been raised about the potential effects of long-term use on cardiovascular and metabolic health and on the developing brain in pediatric populations (7,8). Children appear to have a high liability to experience the metabolic and endocrine side effects of second-generation antipsychotic medications (9,10). In addition, in a study in which follow-up extended to one year, exposure to antipsychotic medications was associated with significantly more frequent movement disorders compared with children who were exposed for only one month (10). This heightened sensitivity suggests significant risks of long-term health complications associated with antipsychotic exposure.

In the fiscal year (FY) ending 2007, an annual report on the use of antipsychotic medications among children in Florida’s fee-for-service Medicaid program revealed that the antipsychotic utilization rate among children under six years old was 1.6 per 1,000 enrollees, representing 931 unique antipsychotic recipients (11). In order to better understand this utilization and the potential risks involved, the Florida Agency for Health Care Administration (AHCA), the state’s Medicaid agency, commissioned several studies of antipsychotic use among children under age six. A number of findings from these studies raised concerns about the long-term health of children who begin antipsychotic treatment at very young ages. First, many of the children who started antipsychotic treatment were extraordinarily young. Twenty-nine percent of the children were four years old and 13% were three years old or younger when their first antipsychotic prescription was paid for by Medicaid (12). Second, exposure to antipsychotic medications among these children was extensive. The mean±SD and median days of exposure over the four years after the initiation of antipsychotic treatment in FY 2004 were, respectively, 828.9±431.9 and 888 days (12).

These results, together with the information in the literature on the metabolic, cardiovascular, and movement side effects of antipsychotic medications on children, heightened concerns in the AHCA about the long-term health of the children involved. One response to these concerns was the implementation of a prior-authorization process that provided expert review of proposals to prescribe antipsychotics to children younger than six years. Rather than arranging for a contractor with already established criteria and review processes, the AHCA initiated a collaborative process with Florida’s psychiatric community. A group of child psychiatrists representing academic medicine as well as psychiatrists working in private and group practice settings was convened by the Florida Mental Health Institute at University of South Florida under contract with the AHCA to formulate a Florida-specific prior-authorization policy as well as processes for adjudicating requests to prescribe antipsychotic medications to children under age six.

This report presents the changes that occurred over the 11 quarters after the implementation of this policy in the numbers of applications submitted as well as in the characteristics of the children for whom applications were submitted and of the professionals who submitted them.

Each month beginning in July 2008, all new applications (N=1,424 applications and unique children) for antipsychotic treatment of children less than six years old were reviewed and either approved or denied by the reviewing child psychiatrists. The medical specialty of the requesting professional and the gender, age, diagnosis, and target symptoms of each child who was the subject of an application were recorded. Children were grouped into three age groups: five-year-olds, four-year-olds, and children younger than four years. The diagnoses available on the application form included attention-deficit hyperactivity disorder (ADHD), autism and pervasive developmental disorders, psychotic disorders, conduct disorders, bipolar or mood disorders, and other psychiatric diagnoses. The instructions for the completion of the form allowed more than one of these diagnoses to be checked.

The symptoms that are the targets of antipsychotic treatment available on the form included severe aggression, self-injurious behavior, extreme impulsivity, extreme irritability, and other target symptoms. Treating physicians identified their specialty in an open-ended question on the application form. Responses were grouped into the following categories: child psychiatrist, general psychiatrist, general neurologist, pediatric neurologist, general pediatrician, developmental pediatrician, family practice physician, and advanced registered nurse practitioner (ARNP).

We used multivariate generalized estimating equations (GEEs) to examine the association between each study variable and change in the odds of receiving new requests over the 11 quarters of the study. GEE belongs to the generalized linear model family that can account for repeated observations via the use of a robust variance estimator (13). It can take the form of advanced logistic regression for clustered data. Each model contained all study variables plus one variable × time interaction, the analysis of which was our main goal. For example, a model testing the change in the odds of new requests in relation to diagnosis included all study variables plus terms for time × diagnosis interactions, with autism and pervasive developmental disorders used as the reference category.

From July 2008 to March 2011, new applications for antipsychotic use were submitted for 1,424 children younger than six years. Child psychiatrists submitted the largest percentage of applications (33%, N=475), followed by pediatric neurologists (20%, N=290), pediatricians (14%, N=200), developmental pediatricians (8%, N=117), general neurologists (11%, N=151), adult psychiatrists (6%, N=87), ARNPs (6%, N=87), and other professionals (1%, N=17). Patients for whom applications were submitted were primarily male (77%, N=1,099). Half of the children (N=714) were five years old, 30% (N=426) were four years old, and 20% (N=281) were under age four when the application was submitted.

A total of 814 children were assigned a single psychiatric diagnosis, with autism or pervasive developmental disorder appearing most frequently (46%, N=374), followed by ADHD (31%, N=249), bipolar and mood disorders (11%, N=90), and conduct disorders (6%, N=49). Among the 610 children with more than one diagnosis, the most frequently appearing combinations of diagnoses included autism with ADHD (32%, N=192), ADHD with a conduct disorder (19%, N=117), bipolar disorder with ADHD (16%, N=98), and autism with a behavior disorder (4%, N=27). A total of 150 of the applications had no diagnosis checked on the form. Seventy-four percent (N=1,059) of the applications mentioned that the symptom targeted by proposed antipsychotic treatment was severe aggression, 59% (N=835) extreme impulsivity, 32% (N=462) self-injurious behavior, and 24% (N=344) extreme irritability.

During the 11 quarters after implementation of the prior-authorization policy, the numbers of applications to treat children under age six with antipsychotic medications declined by 35%, from 124 during the July–September 2008 quarter to 81 during the January–March 2011 quarter. The decline was not completely linear. Although seven of the ten quarter-to-quarter comparisons had declines of 2% to 10%, there was a 21% decrease in applications from the October–December 2009 quarter to the January–March 2010 quarter and a 29% increase from the January–March 2010 quarter to the April–June 2010 quarter.

The analysis of the interaction between each variable and time (quarter), which controlled for all other study variables (Table 1), revealed no significant differences in the odds of generating a new request to prescribe antipsychotic medications based on age or gender. There were also no significant differences in the odds of an application based on diagnosis, with the exception that the “unknown diagnosis” category declined significantly over time. There were significant interactions between time and provider type and between time and target symptoms. Compared with the application rate of child psychiatrists, the odds of submitting an application increased over the study period among general neurologists (odds ratio [OR]=1.15, p<.001), pediatric neurologists (OR=1.09, p<.001), general pediatricians (OR=1.07, p<.02), and adult psychiatrists (OR=1.47, p<.001). Finally, the odds of new application submissions decreased for children with target symptoms of extreme impulsivity (OR=.98, p<.05), whereas they increased for children with extreme irritability (OR=1.41, p<.001).

The prior-authorization process appears to have reduced the number of applications to prescribe antipsychotics to children under six years old. With one exception, there were no significant changes over time in the demographic and diagnostic characteristics of children who were the subjects of applications to prescribe antipsychotic treatment. The exception was the decline in the number of applications with a missing diagnosis, as efforts to get the prescribing community to complete all items on the form continued over time. The age distribution of children indicated that consistently 20% were under four years old when the applications were submitted. The lack of consistent reductions over time in the odds of ADHD as the primary diagnosis is somewhat surprising in view of the fact that the U.S. Food and Drug Administration (FDA) has not approved the use of antipsychotics in the treatment of ADHD for children of any age.

Autism and pervasive developmental disorders were consistently the most common diagnoses over the entire study period, as might have been expected in view of existing FDA indications for antipsychotic use with very young children (14). The target symptoms of antipsychotic treatment changed over time. It is not clear what motivated the significant increase in the odds of identifying extreme irritability as a target symptom and the small decrease in the odds of extreme impulsivity as a target symptom. It may be that the prescribers gravitated toward the precise wording of the FDA indications for the use of aripiprazole among children six to 17 years old and for use of risperidone among children five to 16 years old; these indications specifically mention irritability in the context of autism.

The results regarding the medical specialties of the professionals who submitted applications suggest that clinicians with less specific training and experience in treating children with severe emotional disturbances played an increasing role over time. It can be argued that neurologists, and especially pediatric neurologists, are equipped to treat children with autism spectrum disorders. However, the increasing role of these professionals cannot be explained by a corresponding increase in the number of children with these diagnoses in that the data do not suggest significant changes in the diagnoses of children who were the subjects of applications.

The shift toward less specialized care is clearer; general pediatricians and general psychiatrists were more likely than child psychiatrists to submit applications. We do not know whether the declining role of child psychiatrists relative to several other medical specialties represented a change in the practices of the professionals involved, including the choice of other medications or nonmedication therapies, or a decline in the numbers of child psychiatrists treating young children in Florida’s Medicaid fee-for-service program. In any case, applications came increasingly from clinicians who appeared to have less specific training in the treatment of severe emotional disturbances of very young children. The changes in the training and education of provider-applicants raise questions about the impact of the prior-authorization process on access to appropriate specialists. Because the use of antipsychotics with very young children is an unusual practice with limited data on the associated risks and benefits, a reduction in the role of child psychiatry specialists may have an impact on the quality of care.

There are several limitations to this study, including the fact that the analysis was limited to Florida Medicaid fee-for-service children and may not apply to children enrolled in health maintenance organizations. In addition, the professionals who submitted applications to use antipsychotics with children under six years old identified their medical specialties on the prior-authorization form. These entries may not always comport with information on their actual training and board certifications.

The experience in Florida’s Medicaid program suggests that a prior-authorization program can reduce the submission of applications to prescribe antipsychotic medications to children less than six years old. However, part of this reduction appears to be the result of a trend toward increasing roles of clinicians with less specific training in the treatment of severe emotional disturbances of very young children. It is unclear what impact the prior-authorization review process had on the outcomes of the children involved. A study is under way to address this question.

The study that is the basis for this article was funded by the Florida Agency for Health Care Administration, the State Medicaid agency.

Most of Dr. Constantine’s salary expenses have been paid for by a contract with this state agency. The Medicaid Drug Therapy Management Program funds a variety of activities that are designed to improve the quality of prescribing of psychotherapeutic medications in Florida’s Medicaid program. Dr. Murphy has received funding from Otsuka, Transcept, and Shire. The other authors report no competing interests.