Use of psychotropic medications among youths in foster care has been on the national child health agenda for several reasons. The prevalence is considerably higher among youths in foster care compared with other children covered by Medicaid or private insurance (
1). Moreover, children in foster care are at increased risk of receiving multiple classes of psychotropic medications (
2). Finally, concerns have been raised about inadequate monitoring of psychotropic treatment for youths in foster care (
3,
4).
National trends in psychotropic use overall or by specific therapeutic class for youths in foster care have been consistent across several studies. A single-state study of Medicaid-enrolled preschool-age children reported a nearly twofold greater likelihood of receipt of psychotropic medication among youths in foster care compared with children living in poverty (
5). From 2002 through 2008, an increase in use of second-generation antipsychotics and stimulant medications among preschool-age children was observed (
5,
6). Among Medicaid-enrolled foster care children age three to five years, a 30% increase in antipsychotic use was observed from 2002 (1.7%) to 2007 (2.2%) (
6). Stimulant use in this age group remained stable, with a prevalence of 4.6% in 2002 and 5.2% in 2004, decreasing to 4.2% in 2007 (
6). The U.S. Food and Drug Administration’s approvals of antipsychotics for irritability and autism have considerably lowered the age range for indicated use of antipsychotics, which may explain the rise in use of antipsychotics relative to other psychotropic medications. Few other studies have focused specifically on very young children in foster care. Moreover, the literature has summarized trends in cross-sectional analyses, but no studies have examined longitudinal patterns over time to assess whether psychotropic use among preschoolers is periodic, short-term use (that is, for an acute crisis) or whether, once initiated, these agents are used long term as part of ongoing scheduled treatment.
This study sought to better understand age-related trajectories in psychotropic use among very young children in foster care, in particular among preschool-age children. The specific objectives were to investigate changes in prevalence of psychotropic drug use for each year increase in age among children ages six and younger in foster care and to examine the time-varying odds of longer duration of use for each year increase in age. It was hypothesized that each year increase in age would be associated with a significantly greater likelihood of using psychotropic medication for longer periods of time.
Methods
Study design and sample selection
This retrospective study examined use of mental health services and psychotropic medications among children six years old or younger (as of January 2010) from one large, mid-Atlantic city who were involved in foster care in 2010. The inclusion criteria were that the child was in foster care for at least 365 days over the three-year study period (2009–2011) and had a pharmacy claim for a psychotropic medication anytime during the study period. Of the 1,491 youths in foster care for 365 days or more, the analysis focused on the 178 youths with at least one psychotropic medication prescription during the study period. The study focused on 2010 because this is the cohort selection year, and it was certain that the entire population of youths involved in foster care at any time in the year would be identified. The state mental health and child welfare agencies and the University of Maryland institutional review boards approved this study.
Data sources
Data for this study were obtained from child welfare administrative records and state department of mental health claims data. Identification of children in foster care was based on the child welfare administrative database. The date a child is removed from the home and all placements thereafter are recorded in the database.
Information about mental health services and psychotropic treatment was derived from Medicaid administrative service and pharmacy claims. Mental health service claims represented all services provided by the state mental health carve-out program, in which mental health benefits are detached from general medical benefits. Outpatient mental health–related visits were used to identify psychiatric diagnoses and psychotherapy visits at any point during the study period. ICD-9 codes 290 through 319 were used to categorize major diagnostic groups. Individual diagnoses were grouped into the following broad categories: any disruptive disorder (attention-deficit hyperactivity disorder [ADHD], conduct disorder, oppositional defiant disorder, and impulse control disorder), any internalizing disorder (anxiety, depression, and posttraumatic stress disorder), any mood disorder (mood or bipolar disorder), and any developmental disorder (pervasive developmental disorder, mental retardation, and all other developmental disorders). Mood and bipolar disorders were not categorized as internalizing disorders because it was important to be able to distinguish antipsychotic use by youths with anxiety or depression from use by those with mood or bipolar disorders for which antipsychotics are typically used. Psychotherapy services were identified from procedure codes for individual (90804–90808, 90816, 90817, 90819, 90821, 90824, and 90827), family (90832–90834, 90836, 90837, 90846, 90847, and 90849), and group (90853) psychotherapy.
Medicaid pharmacy claims were used to identify children who received at least one psychotropic medication during the study period. All psychotropic treatment reflected use after entry into foster care. Psychotropic medications were selected by individual agent and categorized by major therapeutic class. The therapeutic classes included ADHD medications (stimulants and atomoxetine), antidepressants, antipsychotics, and mood stabilizers (lithium and anticonvulsants: carbamazepine, valproic acid, gabapentin, lamotrigine, and oxcarbazepine), alpha-agonists (clonidine and guanfacine), and sedatives (hydroxyzine and benzodiazepine). Exclusions were made if a youth received a mood-stabilizer anticonvulsant and also had a seizure disorder diagnosis.
Psychotropic utilization measures
Psychotropic prevalence.
Psychotropic prevalence was estimated at the population level for use overall, by therapeutic class, and by year of age. Prevalence was calculated as the number of users (that is, youths who received at least one prescription claim for a psychotropic medication) per 100 youths in foster care, overall and by year of age.
Psychotropic duration.
Duration of psychotropic use was estimated at the individual level among the subset of those who received treatment. Duration was measured as days per month (continuous), three or more weeks per month (binary: yes or no), and average days across the 36-month study period from January 2009 through December 2011. Average duration was estimated for each therapeutic class as well as for concomitant regimens of three or more therapeutic classes. Because a prescription dispensing may cross two months, a monthly use file was created on the basis of the dispensing date and days’ supply from consecutive refills to calculate the days of use per month—that is, a 30-day supply may have had 25 days in one month and five days in the next month. The monthly use file was a longitudinal file with indicator variables for year (2009, 2010, and 2011), month (1–36), and therapeutic class days (0–30/31) for ADHD medications, antipsychotics, antidepressants, mood stabilizers, alpha-agonists, and three or more classes. To account for variable time in foster care during the study period, the medication possession ratio (MPR) was estimated as the days of medication use divided by the total days in foster care.
Statistical analyses
Descriptive frequencies, means, and standard deviations were used to characterize demographic characteristics, use of mental health services, and use of psychotropic medication. Bivariate chi square tests and one-way analyses of variance were used to compare key demographic and placement characteristics (that is, gender, race-ethnicity, days in foster care, psychiatric diagnoses, psychotropic classes, days of medication use, and use of psychotherapy) by year of age.
Longitudinal regression was used to assess the association between age and days of antipsychotic use per month, and a longitudinal logistic regression was used to examine the association between age and the likelihood of receiving antipsychotic medication for three weeks or more per month, adjusting for gender, race-ethnicity, days in foster care, psychiatric diagnoses, and use of another major psychotropic therapeutic class. Other psychotropic medication classes were included in the set of predictors as a proxy for an underlying condition or symptom that would not be detected from the ICD-9 codes. Furthermore, receipt of more than one psychotropic medication may represent prescriptions for children with more severe or impairing conditions, and this would thus be accounted for in the analysis. The binary dependent measure (yes or no) of three weeks or more per month was a proxy for scheduled use (of longer duration) as opposed to as-needed use (short-term use) in acute crises. Because entry into foster care differed across youths, the index date for study entry was either the date of entry into foster care or January 2009 (earliest data available), whichever date came later. Assessment of antipsychotic use began after a 30-day lead-in time from the index date to allow for administrative processing that may occur on entry or that may be an artifact of our access to historical data. Children were followed until December 31, 2011, or data were censored if a child left foster care before the end of the study. [A figure illustrating the study design is available in an online data supplement to this article.]
Covariates were entered sequentially, and the independent contribution of each was assessed. A covariate was retained if it improved model fit. Generalized estimating equations were used to account for correlated data from multiple observations within subjects. The quasi-likelihood under independence model criterion was used to assess goodness of model fit. Data were analyzed with SAS, version 9.3.
Discussion
The study found that use of psychotropic medication began among early preschool-age children and became more prevalent by age six. Concomitant use of three or more psychotropic classes began among children as young as age four. For most psychotropic classes, duration of use was significantly associated with each year increase in age. By age six, the MPR for antipsychotic medications exceeded the MPR for ADHD medications, antidepressants, and mood stabilizers.
This study corroborates other reports that document psychotropic treatment initiation among young children. Most studies have examined antipsychotic use specifically. Initiating antipsychotics before the age of six was correlated with continued use over a four-year period, averaging 822 days when used singly and 624 days when used concomitantly with other psychotropic classes (
7). The proportion treated with an antipsychotic was found to nearly double from 13% among children less than four years old, to 29% among four-year-olds, and to 58% among five-year-olds, with 77% receiving concomitant psychotropic classes (
8). Others have noted differences in second-generation antipsychotic use between children less than six compared with children age six to 12 (
9).
In addition, our study contributes new information because we examined time-varying changes in duration of psychotropic use by year of age. Differences in the duration of use per month have clinically meaningful implications with regard to how the medications are being used and the overall exposure. The higher likelihood of using antipsychotic, antidepressant, and ADHD medications for three or more weeks per month with each increase in year of age suggests a trend toward chronic use. Such use occurred over a very short period of tremendous developmental growth.
The findings of a relationship between early exposure and a trajectory for increased duration with age raise important safety concerns that warrant further discussion. Concerns have been expressed about the effect of early, chronic antipsychotic exposure among young children on brain development given data showing effects on brain size, neuronal circuitry, and brain volume (
10–
12). Another concern is over metabolic adverse events with second-generation antipsychotics. Antipsychotic-naïve users may be more susceptible to metabolic side effects (
13,
14), and very young children are likely to be first-time users. A comprehensive review noted that more than half of all youths who take these medications experience this side effect (
15). Furthermore, several studies involving privately insured and Medicaid-insured youths have demonstrated a two- to threefold increased risk of incident diabetes (
15,
16). Notwithstanding these risks, few patients receive routine metabolic monitoring in outpatient clinical care (
17). The literature is scant for young children, but the evidence warrants further investigation given the potential long-term risks.
Few youths in this study had a psychotherapy visit, not to mention evidence-based treatments. Other investigators have reported low rates of behavioral health services among three- and four-year-olds who were identified as having a social-emotional problem (
6,
18). Nationally, only 24% of antipsychotic-related visits in 2005–2009 among youths age 20 or younger involved psychotherapy (
19). However, it was not possible to isolate patterns for very young children. The Preschool Psychopharmacology Working Group, which is affiliated with the American Academy of Child and Adolescent Psychiatry, recommends use of a nonpharmacologic intervention for at least 12 weeks; if this fails, then a second nonpharmacologic intervention should be tried (
20). This study did not capture psychotherapy services that may have been delivered in school.
This study had several limitations. It was limited to one geographic location and may not represent the population of preschool-age children in foster care. Causation cannot be inferred from results of this observational study. It is not possible to know whether children who received psychotropic medication had failed to respond to psychosocial interventions. Data on the severity of the psychiatric disorder were not available. The foster care population in the study may reflect young children with greater psychiatric impairment compared with young children who spent less time in foster care or who were adopted. Nonetheless, this subgroup is of greatest concern. Because it was not possible to detect psychotropic use that occurred before entry into foster care, the study may have underestimated duration of use. However, given the young age of the sample, this should not have compromised the findings because few youths would have received a psychotropic medication before the age of three. Finally, the small numbers prohibited analysis by individual antipsychotic or a more in-depth analysis of concomitant treatment with three or more psychotropic classes.
The limitations are balanced by the strengths of this study. The child welfare administrative records provided precise information on foster care placement. The most recent data available were used to examine longitudinal treatment patterns. Finally, no other study has compared longitudinal trajectories of antipsychotic treatment by year of age among young children.