The Appropriateness of Use of 3 Classes of Psychotropic Medications in Children and Adolescents
For the majority of children and adolescents receiving a psychotropic medication there is a valid medical diagnosis.
The prevalence of mental health disorders in children and adolescents has been estimated to be 12%.1 Comorbid psychiatric conditions are common among this segment of the population.2 Psychotropic medications play an increasingly important role in the treatment of these disorders in children. The prevalence of having at least 1 prescription for a psychotropic medication in Medicaid enrollees has been estimated to be as high as 6%.3 Historically, few studies have been conducted on the effi cacy and safety of the use in children of many of these medications commonly prescribed today. Only a few have been approved by the Food and Drug Administration (FDA) for use in patients younger than 18 years, often based on trials with a relatively small number of child subjects.4 Concern by the public and from several clinical trials over a possible increase in suicidality in pediatric and adolescent users of antidepressants led to recent reviews by regulators in the United States and the United Kingdom. In the United Kingdom, these reviews resulted in a recommendation against the use of all but 1 drug (fl uoxetine) in patients younger than 18 years5 and, in the United States, in a Public Health Advisory and mandated “black box” warnings for the package inserts.6,7 Journal articles, reviews, editorials, and lay press articles have also appeared in recent years expressing concerns regarding the use of psychotropic medications, particularly of stimulants, in the preschool age group.8-15 Until late 2006, there were only small, short-term trials of 1 stimulant in the preschool age range, which included a total of only 246 subjects.16
The aim of this study was to compare the use of 3 groups of psychotropic medications by Medicaid and State Children’s Health Insurance Program (SCHIP) enrollees younger than 18 years in Arizona with age-specifi c FDA approvals, published practice guidelines, meta-analyses, or review articles, an approach which has not been previously published. In addition, we assessed the prevalence of use of these medications by enrollees.
Subjects included in this study were children who were 17 years or younger on January 1, 2006, and continuously enrolled through calendar year 2006 in either the Medicaid or SCHIP program in Arizona.
Study Design and Data
This retrospective analysis was a cross-sectional study using administrative databases. Demographically de-identifi ed medication claims and medical/facility claims data were used. Data of interest was limited to those members aged 17 years or less who were continuously enrolled for 12 months in 2006. The requirement of continuous enrollment eliminated all members less than 1 year of age. There were 160,714 subjects with at least 1 pharmacy claim and 232,284 with at least 1 medical/facility claim.
Three groups of psychotropic medications were selected for this study, based on expected substantial utilization and concern about potential inappropriate use. The medications included in each group (attention defi cit hyperactivity disorder [ADHD], antidepressants, antipsychotics) were all marketed during 2006 and are listed in Table 1. They included central nervous system stimulants and atomoxetine, the selective serotonin reuptake inhibitors (SSRIs) and bupropion, and the atypical antipsychotics. Claims for study medications were identifi ed by using all NDC (National Drug Code) codes associated with each generic ingredient name.
Defining Appropriateness of Use of the Medications
For each study medication, we identifi ed age- and diagnosis-specifi c indications for appropriate use. These were determined by reviewing the FDA-approved product package inserts. In addition, we searched for relevant practice guidelines using the Agency for Healthcare Research and Quality’s National Guideline Clearinghouse (www.guidelines.gov) and websites for major specialty societies. The PubMed Mesh database (www.pubmed .gov) was also searched for each medication name using birth to 23 months, 2 to 5 years, 6 to 12 years, and 13 to 18 years age limits. The MicroMedex database (Thomson Reuters Healthcare, New York, NY) was accessed for additional recommended indications. When a specifi c lower age limit of a “childhood” indication in a reference was not stated, it was assumed to be 6 years of age; for “adolescent,” 13 years of age; for “adult,” 18 years of age. All indication age ranges were adjusted to fi t 4 standard age groups: 0 to 2, 3 to 5, 6 to 12, and 13 to 17 years. A board-certifi ed pediatrician (MRB) and 2 pharmacists (DCM, TW) reviewed this information and concurred on the indications for appropriate use.
Diagnostic codes for the indications for appropriate medication use were determined using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) (World Health Organization as modified by the National Center for Health Statistics). In view of previous experience working with the limited validity of diagnostic coding in administrative claims data, we decided to use ICD-9-CM codes limited to 3 digits for this analysis. As a consequence, this study had reduced precision.
The primary outcome for this study was the percent of use of study medications meeting appropriateness criteria. The numerator for appropriate use was defi ned as the number of unduplicated pharmacy claims for a medication for which there was a medical/facility claim for the same subject with an ICD-9-CM code for an age group—specific recommended indication. The denominator was the total number of unduplicated pharmacy claims for each medication of interest. We report the results aggregated at the medication class level. A secondary outcome was the prevalence of psychotropic medication use and was defined as the number of subjects with at least 1 prescription claim for a study medication during the study year times 1000 divided by the total number of subjects. Both prevalence and appropriate use measures were stratified by total number of subjects, male and female subjects, subjects by race (white, black, Hispanic, Native American, Asian/Pacifi c Islander/Cuba /Haitian,other, and unspecified), and subjects by age group (1 to 2 years, 3 to 5 years, 6 to 12 years, and 13 to 17 years). Data analysis was performed using SAS version 9.1 (SAS Institute, Cary, North Carolina).
Of the 274,569 subjects, 14,481 had at least 1 prescription for any study medication during 2006, resulting in a prevalence of use of 52.7 per 1000 subjects (95% confidence interval [CI]: 51.9-53.5). As seen in Table 2, the prevalence of ADHD medications was 39.2 per 1000 subjects (95% CI: 38.5-39.9), for antipsychotics it was 21.2 per 1000 subjects (95% CI: 20.7-21.7), and for antidepressants it was 13.3 per 1000 subjects (95% CI: 12.9-13.7). The total number of subjects in the 3 cohorts of subjects using the different medication groups exceeds the number of subjects using any study medication because of a considerable degree of co-prescribing of these medications.
The prevalence of use of any study medication by males was more than twice that of females (males 71.3 per 1000 subjects [95% CI: 70.0-72.6]; females 33.1 per 1000 subjects [95% CI: 32.1-34.1]; P <.001). The sex difference was greatest for ADHD medications (males 56.7 per 1000 subjects [95% CI: 55.5-57.9]; females 20.6 per 1000 subjects [95% CI: 19.8-21.4]; P <.001) and least for antidepressants (males 14.6 per 1000 subjects [95% CI: 14.0- 15.2]; females 11.9 per 1000 subjects [95% CI: 11.3-12.5]; P <.001). The prevalence of use of any study medication in the 1 to 2 year age group was very low (0.7 per 1000 subjects [95% CI: 0.5-0.9]). The 3 to 5 year age group was substantially higher for any medication use, 19.6 per 1000 subjects (95% CI: 18.4-20.8; P <.001), with the prevalence for ADHD medications being 15.9 per 1000 subjects (95% CI: 14.9-16.9) and for antidepressants only 1.9 per 1000 subjects (95% CI: 1.4-2.0). The prevalence of use for these medications was still higher in the 6 to 12 and 13 to 17 year age groups. The prevalence of use of ADHD medications was highest in the 6 to 12 year age group (62.9 per 1000 subjects [95% CI: 61.5-64.3]), but the prevalence of use of both other medication groups was highest in the 13 to 17 year age group (antipsychotics 39.9 per 1000 subjects [95% CI: 38.3-41.5]; antidepressants 34.6 per 1000 subjects [95% CI: 33.1-36.1]).
With respect to racial differences, the prevalence of medication use was highest (171.9 per 1000 subjects [95% CI: 163.9-179.9]) among the 8534 subjects for whom the racial category was listed as unspecifi ed. Otherwise, the prevalence of use of any study medication was highest for whites (111.5 per 1000 subjects [95% CI: 109.0-114.0]), then progressively lower for blacks (75.1 per 1000 subjects [95% CI: 71.0-79.2]), Hispanics (26.9 per 1000 subjects [95% CI: 26.1-27.7]), and Native Americans (15.2 per 1000 subjects [95% CI: 13.7-16.7]). The prevalence of use of the 3 medication groups had a similar pattern. As seen in Table 3, 2 other racial categories, Asian/Pacifi c Islander + Cuban/Haitian and other, each had less than 0.7% of subjects.
The overall rate of appropriate use was 74.5% (95% CI: 73.9-75.0), but varied considerably by medication class. As seen in Table 4, the percentage of appropriate use for ADHD medications was 82.1% (95% CI: 81.5-82.8), but was lower for antipsychotics (67.4% [95% CI: 66.2-68.5]) and antidepressants (62.7% [95% CI: 61.2-64.2]). There was little difference in appropriate use by sex (male 75.1% [95% CI: 74.5-75.8], female 72.9% [95% CI: 71.8- 73.9]; P = .0003). Based on the evidence review, there was no appropriate use in the 1 to 2 year age group overall or in the 3 to 5 year age group for antipsychotics or antidepressants, as there were no recommended indications for those groups. Only a total of 2.8% of psychotropic medication use was in those groups. In the 6 to 12 and 13 to 17 year age groups, the medication class with the highest percent appropriate use was ADHD medications (83.7% [95% CI: 82.9-84.5] and 82.0% [95% CI: 80.7- 83.3]) and lowest was antidepressants (61.7% [95% CI: 59.3-64.0] and 66.3% [95% CI: 64.4-68.3]). There was little difference in percent appropriate use by race of recipient of any medication or of each medication group.
Concerns about appropriate psychopharmaceutical use in children abound in the media and scientifi c literature. The findings of this study provide evidence that use of psychotropic medications in children in a Medicaid/ SCHIP health system is largely in accordance with published guidelines and/or manufacturer labeling. This study also adds to the limited public knowledge base regarding the prevalence of use of psychotropic medications in child and adolescent enrollees of the Medicaid and SCHIP programs, high-risk groups based on socioeconomic status and rate of disabilities.
Appropriateness of Use
We were interested in assessing how closely the use of psychotropic medications adhered to the recommended indications for their use. It is important to note that many of these medications do not have an FDA-approved indication for use in children and adolescents. So-called “off-label” use, though, is legal and frequently supported for specifi c indications by practice guidelines and the professional literature. However, recommendations are often based on few studies involving small numbers of subjects. Nonetheless, providers seeing young patients in their offi ces with significant mental health symptoms often consider the use of these medications in spite of the limited evidence base.
Within the limits of our methodology, we determined that the proportion of appropriate use of any study medication was 74.5% (95% CI: 73.9-75.0). The proportion of documented appropriate use differed by medication class. It was 82.1% (95% CI: 81.5-82.8) for ADHD medication, 67.4% (95% CI: 66.2-68.5) for anti-psychotics, and 62.7% (95% CI: 61.2-64.2) for antidepressants. We did not find similar assessments for appropriateness of use in the literature. However, 2 studies used different methodologies to suggest a lack of appropriate use. The fi rst study indicated that 57% of children receiving stimulant medications never had parent-reported ADHD symptoms.17 In the second study, 57% of children in outpatient treatment for depression received a prescription for an antidepressant, though the American Academy of Child and Adolescent Psychiatry practice guideline suggests restricting their use to only more severe or resistant cases.18,19
One potential explanation for the lower level of appropriateness of use for both the antidepressant and anti-psychotic medication groups, compared with ADHD medications, was the high prevalence of the ADHD diagnosis among the recipients of each of these medication groups (46.8% and 58.3%, respectively). ADHD has common mental health co-morbidities,20,21 including depression, a major indication for use of antidepressants, and conduct disorder/oppositional defi ant disorder, indications for anti-psychotics.19,22 Inaccurate coding pract ices could result in the use of the ICD-9-CM code for ADHD rather than the correct diagnostic code for the comorbid condition actually being treated. In addition, incomplete diagnostic coding is a well-known problem with medical claims data.23,24
Prevalence of Use
The prevalence of use of any psychotropic medication (52.7 per 1000 subjects [95% CI: 51.9-53.5]) in Arizona was similar to that found in previous studies, though population characteristics, particularly racial composition, and medications differ among the studies. Studies of 4 state Medicaid programs and 1 SCHIP program showed annual prevalence ranging from 45 to 62.6 per 1000 subjects. These estimates are remarkably similar to each other, and to this study, despite the differences in the populations and the medications included. This also suggests that the medications we selected to include adequately represented psychotropic medications in general use.
The prevalence of use of ADHD medications in this study, 39.2 per 1000 subjects (95% CI: 38.5-39.9), was similar to that found for stimulants in 3 Medicaid programs (37.2 to 46.3 per 1000 subjects).3,25 The prevalence of use of antipsychotic medications was 21.2 per 1000 subjects (95% CI: 20.7-21.7). In 5 Medicaid programs it ranged from 5.4 to 14.9 per 1000 subjects,3,26 suggesting an upward trend over time from the 1996-2001 period of those studies to 2006. The prevalence of use of antidepressant medications in this study was 13.3 per 1000 subjects (95% CI: 12.9-13.7), similar to the 8.8 to 20.5
per 1000 subjects found in 3 Medicaid programs.3,27
Our analysis found disparities by race in use of these medications, as have previous studies of Medicaid programs. Whites in this study had a prevalence of use of any medication 1.5 times higher than for blacks, 4.1 time higher than for Hispanics, and 7.3 times higher than for Native Americans. One should keep in mind that our data did not include information on prescriptions paid by other sources, such as the Indian Health Service. Other studies have shown the prevalence for whites as high as 2.4 times that for blacks and 1.7 times that for Hispanics. In addition, there was an apparent association in this study between use of psychotropic medications and not providing a racial identifi cation at enrollment, which was not noted in the other studies referenced above.
This study also indicates a sex difference in the use of these medications. Prevalence of use of any medication was 2.2 times higher for males than for females. This difference has been reported in earlier studies of Medicaid enrollees, ranging from 1.6 to 2.4 times higher for males.3,28,29 One reason for this difference is the known higher prevalence of diagnosing ADHD for males (2.8 times higher than for females in this study).30 The prevalence of use of ADHD medications was, indeed, 2.8 times higher for males in our analysis, and has ranged from 3.0 to 3.5 times higher for males in other studies involving Medicaid enrollees.3,25,27,29
Concern has been expressed in the literature and lay press about the increasing use, and safety, of psychotropic medications in infants and preschool children.8-15 The prevalence of use of any medication in this study for a combined 1 through 5 year age group was 10.4 per 1000 subjects (95% CI: 9.8-11.0). The prevalence in 2 other Medicaid programs in 1995 for subjects aged 0 through 4 years was similar at 9.8 and 15.3 per 1000.3 The prevalence in these same 2 programs plus 5 others in 2001 for ages 2 through 4 years was 23 per 1000 subjects.31 The prevalence of use of ADHD medications in this study for the 3 to 5 year age group was 15.9 (95% CI: 14.9-16.9), and for a combined 1 to 5 year age group 8.3 per 1000 subjects (95% CI: 7.7-8.8). The prevalence of use of stimulant medication for subjects aged 2 to 4 years in 3 Medicaid programs was 8.9, 11.2, and 12.3 per 1000.11,28 For subjects aged 0 to 4 years in 2 Medicaid programs, the prevalence of stimulants was 4.2 and 6.8 per 1000.3 In another Medicaid program, the prevalence was 13 per 1000 for subjects aged 1 to 5 years and in a second it was 6.2 for subjects aged 0 to 5 years.25,27
Limitations of this study include the use of administrative data, originally collected for reasons unassociated with this study. There was a risk of misclassification, particularly related to errors in age and eligibility determination. However, those variables were also important data elements required for business use, reducing the likelihood of such misclassifi cation. The decision to limit the study population to those members continuously eligible during the study period might have made the sample population not representative of the source population. The determination of appropriateness of use was based on our evaluation of the review literature and practice guidelines. It is possible that our literature searches failed to identify relevant sources that would have impacted our evaluation. Our measure of appropriateness was also subject to coding biases. In addition to possible miscoding and incomplete coding of diagnoses, physicians can be reluctant to assign a diagnosis to a patient in certain circumstances. Also, as discussed previously in the Methods section, our use of diagnostic codes limited to 3 digits caused decreased precision.
The prevalence of use of psychotropic medications by children and adolescents in Arizona was generally consistent with that previously reported from Medicaid programs in other states. There was evidence of racial disparity in use of these medications. The majority of use was consistent with recommended indications. This was, to our knowledge, the first study to assess the appropriateness of use of psychotropic medications using an evidence-based methodology. These findings need to be replicated in further studies on other populations.