Psychotropic Medication Use and Clinical Outcomes Among Children and Adolescents Receiving System of Care Services

Abstract

Objective:

Few studies of psychotropic medication use among children and adolescents address the effectiveness of this medication, as it is typically used in naturalistic treatment settings. The objective of this study was to investigate psychotropic medication use among children and adolescents treated in system of care communities, to identify subject characteristics associated with psychotropic medication use, and determine whether psychotropic medication use is associated with reduced symptom severity.

Methods:

Data were collected through the National Evaluation of the Comprehensive Community Mental Health Services for Children and Their Families Program. For this evaluation, 27 system of care communities reported data on medication use and clinical ratings at intake and 6-month follow-up on children and adolescents 6–18 years of age receiving mental health services between 2006 and 2009. We used mixed-effects logistic regression to determine associations between patient characteristics and medication use, and mixed-effects linear regression to determine whether subjects taking medication and those not taking medication experienced different changes in symptom severity between intake and 6 months.

Results:

Subject characteristics associated with psychotropic medication use, when controlling for other characteristics, included having more severe emotional and behavioral symptoms at intake, having more co-occurring diagnoses, and receiving more than one type of service. Those both taking and not taking medication showed symptom reduction at 6 months, although symptom severity among subjects taking medication remained in the clinical range. When controlling for covariates, symptom reduction was associated with medication use.

Conclusion:

Taking psychotropic medication was more strongly associated with measures of illness severity—greater symptom severity at intake, more co-occurring diagnoses, and more service intensity—than with other demographic characteristics. Subjects who took medication showed more symptom reduction at 6 months than those not taking medication, although this reduction was not sufficient to normalize symptoms among those taking medication.

Introduction

R ate estimates of psychotropic medication use among children and adolescents diagnosed with serious emotional disturbance range between 50 and 90%, with variations across age, gender, race/ethnicity, poverty status, symptom severity, and diagnosis (Teich et al. 2003; Moreno et al. 2007; Visser et al. 2007; Jerrell 2008). Studies suggest that psychotropic medication is more likely to be prescribed to children and adolescents who are male, with some variations according to diagnosis (Leslie et al. 2003; Martin et al. 2003; Warner et al. 2004); white (Leslie et al. 2003; Zito et al. 2003; Olfson et al. 2006; Hudson et al. 2007; Pavkov and Walrath 2008); and from higher income families (Leslie et al. 2003; Froehlich et al. 2007; Pavkov and Walrath 2008). Those few studies investigating symptom severity suggest that those most seriously impaired are more likely to be treated with medication than those less seriously impaired (Leslie et al. 2003; Warner et al. 2005; Pavkov and Walrath 2008). More severe illness is associated with fewer recoveries, longer episodes of illness, and increased recurrence (Emslie et al. 2003), and is suggested to be a better predictor of posttreatment functioning than the type of treatment received (Molina et al. 2009). Most studies examining medication use and diagnosis investigate single diagnoses that are more likely to be treated with medication than are other diagnoses; only two studies have investigated co-occurring diagnoses and medication use. These two studies suggest that children and adolescents with co-occurring psychiatric diagnoses are more likely to be treated with psychotropic medication than those with a single diagnosis (Warner et al. 2004, 2005). The association between service intensity and medication effectiveness among children and adolescents has rarely been investigated, other than demonstrations that service intensity—both inpatient and outpatient—has decreased over time as the use of psychotropic medications has increased (Martin and Leslie 2003).

Most pharmacoepidemiological research, studying medication use and effects among large populations, typically analyzes data from medical records, pharmacy claims, and service-event surveys (Zito and Safer 1997, 2005; Martin et al. 2003; Martin and Leslie 2003; Teich et al. 2003; Zito et al. 2003; Warner et al. 2004; Olfson et al. 2006; Moreno et al. 2007; Jerrell 2008; Fontanella et al. 2009). These analyses provide accurate measures of medication use as prescriptions written or filled. However, these measures are different than measures of medication actually taken, given that as many as 31% of prescriptions written are not actually filled (National Community Pharmacists Association 2010), and that not all medication from filled prescriptions is actually taken. Estimates of nonadherence for psychotropic medication range from 20 to 72%, with variations according to diagnosis (Coletti et al. 2005; Julius et al. 2009). Very few studies on medication use report medication actually taken.

Similarly, few pharmacoepidemiological studies of psychotropic medication use among children and adolescents address the effectiveness of this medication use, as it is typically used in naturalistic treatment settings (Zito and Safer 1997; Jerrell 2008). The Report of the Surgeon General's Conference on Children's Mental Health drew attention to the substantial gaps that exist between how medications are used in ideal research settings and how they are used in “real world” practice (United States Public Health Service 2000). Studies on medication outcomes among children and adolescents typically report efficacy results from clinical trials conducted within controlled research settings. Efficacy studies typically are limited to homogeneous groups of participants with a single diagnosis taking a single medication under limited conditions for a short period of time. In contrast, effectiveness studies are conducted among clinically diverse populations in the real world of clinical care, in which often there are many complicating factors, including medications prescribed off-label to participants who have co-occurring conditions, and in polypharmacy combinations (Center for Health Services Research on Pharmacotherapy 2010; Riddle et al. 2001).

Although the efficacy of specific psychotropic medications has been demonstrated in clinical trials (Kapczinski et al. 2003; Wolraich 2003; Ipser et al. 2009; Molina et al. 2009), the efficacy demonstrated in these trials does not necessarily translate into effectiveness in naturalistic treatment settings (Zito and Safer 1997). Among those few studies of psychotropic medication effectiveness among children and adolescents, one study measured effectiveness using inpatient readmission rates among a sample of suicidal adolescent inpatients, and found that treatment with a single antidepressant was associated with 85% lower risk of readmission, but treatment with three or more medications from other drug classes was associated with a 2.6 times higher risk of readmission (Fontanella et al. 2009). A second study of children treated with stimulant medication found no support for the long-term effectiveness of medication treatment as monitored in community treatment settings, in contrast to the initial efficacy established during the randomized clinical trial phase of the same study (Molina et al. 2009).

The current study examines psychotropic medication use and clinical outcomes among a large number of clinically diverse children and adolescents treated in naturalistic settings. This study examines overall psychotropic medication use rather than the use of specific medications or combinations of medications. Our first research objective was to determine whether differences in subject characteristics exist between those who were and were not taking psychotropic medication. Our second research objective was to examine the association between psychotropic medication use and changes in emotional and behavioral symptoms within the first 6 months of receiving system of care services.

Methods

Participants

Data analyzed in this investigation were collected as part of the Cross-Sectional Descriptive Study and the Longitudinal Child and Family Outcome Study of the National Evaluation of the Comprehensive Community Mental Health Services for Children and Their Families Program, also known as the Children's Mental Health Initiative (CMHI). Data from these studies were collected by 30 system of care communities receiving CMHI grants in 2005 and 2006, including states, counties, and Native American tribes. CMHI system of care communities provide services to children and adolescents with serious mental health challenges and their families through a coordinated network of community-based services. The system of care communities that collected data for these studies were funded by the Substance Abuse and Mental Health Services Administration (SAMHSA) under the CMHI. Consent from caregivers and assent from youth 11–18 years of age were obtained for analysis of these data.

Children and adolescents in CMHI-funded system of care communities receive a wide range of services, including case management, assessment, therapy, support services, inpatient hospitalization, residential treatment services, and medication management. Entry into system of care services requires that a child or adolescent have at least one clinician-assigned psychiatric diagnosis, as defined by the Diagnostic and Statistical Manual of Mental Disorders, 4th ed. (DSM-IV) (American Psychiatric Association 1994).

Among these system of care communities, 27 communities reported data for 2,517 children and adolescents (6–18 years of age) receiving system of care services between October 2006 and October 2009. Data on medication status at both intake and 6-month follow-up were available for 1,295 of these children and adolescents. To focus our analysis on those subjects with reliable data on medication status, 407 with unreliable data on medication use were excluded. This analysis is based on the remaining 888 children and adolescents 6–18 years of age (M=12.3, SD=3.09). Among our study sample of 888, 61.5% were male, 50.9% were white, 25.3% were black/African American, and 15.9% were Hispanic. Approximately 70% lived at or below the federal poverty level. At intake, these subjects had mean Total Problems scale scores on the Children's Behavior Checklist (CBCL) for ages 6–18 of 68.4, with 75.7% scoring at or above the clinical level. All subjects in this sample met criteria for at least one DSM-IV diagnosis, with 43.6% having one diagnosis, 34.2% having two diagnoses, and 22.2% having three or more diagnoses. Similarly, all subjects in this sample received at least one type of system of care service for their behavioral health symptoms, with a mean number of service types of 5.3. These services may have been provided by any of the service sectors coordinated by system of care programs, including mental health, child welfare, juvenile justice, education, and family and youth organizations. These services included assessment, case management, a wide range of therapeutic services, support and training services, and inpatient and residential treatment. Among this sample, 43.8% received between one and four service types, 36.6% received between five and eight service types, and 19.6% received nine or more service types.

Measures

Data on subjects' age, gender, race/ethnicity, and annual household income were collected from case records and from the subjects' primary caregivers at intake into services. We recoded race/ethnicity into four mutually exclusive groups: white, black/African American, Hispanic, and other (including mixed race). We recoded annual household income to construct family poverty status based on the U.S. Department of Health and Human Services poverty guidelines for 2006 (US DHHS 2009). Annual household incomes at or below 1.5 times the threshold were categorized as “at/below poverty”; incomes greater than 1.5 times the threshold were categorized as “above poverty.”

We included three indicators of illness severity: symptom severity at intake, co-occurring diagnoses, and service intensity. We measured symptom severity at intake using the CBCL 6–18 Total Problems scale scores. For diagnosis, as many as three DSM-IV Axis I diagnoses and as many as two Axis II diagnoses were recorded for each subject. In order to account for diagnostic complexity, we constructed an indicator of co-occurring Axis I and II diagnoses that grouped subjects into mutually exclusive categories of one diagnosis, two diagnoses, or three or more diagnoses. For service intensity, caregivers were asked during the 6 month follow-up interview whether the subject or family had received any of 25 types of services during the previous 6 months. These services included an array of outpatient services (including medication monitoring), inpatient or residential treatment services, and support services. We created a continuous service intensity measure, with subjects receiving between 1 and 16 types of service (M=5.2, SD=3.0).

We constructed our first outcome, medication status, from two questions asked during the 6 month follow-up interview. Caregivers were asked whether their child was taking medication for emotional or behavioral symptoms at the time of the interview, and whether the child had taken medication in the previous 6 months. Of note is that caregivers were asked about medications their child had taken, not medications prescribed or prescriptions filled. We identified two medication status groups: 1) subjects who were neither taking medication at the time of the 6 month follow-up interview, nor had taken medication in the previous 6 months according to their caregivers (no medication group), and 2) subjects who were taking medication at the time of the 6 month follow-up interview and had taken medication in the previous 6 months according to their caregivers (medication group). These two groups provided a clear contrast for measuring potential differences in clinical outcomes. We excluded subjects whose caregivers reported that they had taken medication in the previous 6 months, but who were not taking medication at the time of the 6 month interview, because we could not accurately establish the timing or duration of medication use, adherence, or reasons for discontinuing. Likewise, we excluded subjects whose caregivers reported that they were taking medication at the time of the 6 month interview, but had not taken medication in the previous 6 months, because the duration of medication use was unknown, and may not have been long enough to affect emotional and behavioral symptoms. Including either of these two latter groups had the potential to introduce bias, the magnitude and direction of which could not be accounted for nor hypothesized a priori. We did not control for any service use or medication experience prior to system of care intake.

As our second outcome measure, we used change in the CBCL 6-18 Total Problems scale scores between intake and 6 month follow-up to indicate change in symptom severity. The CBCL 6-18 is a standardized measure of social competence and behavioral and emotional problems among children and adolescents 6–18 years of age, with strong internal consistency, test–retest reliability, construct validity, and criterion validity (Achenbach and Rescorla 2001). Scores on the CBCL 6-18 Total Problems scale between 60 and 63 are in the borderline clinical range; scores ≥64 are in the clinical range. Change in the CBCL 6-18 Total Problems scale score was measured using a continuous reliable change index (RCI) (Jacobson and Truax 1991), constructed using the intake and 6 month follow-up CBCL 6-18 Total Problems t scores. The RCI creates an indicator of change using scores on standardized clinical instruments at two points in time. The RCI differs from a traditional change score in that it adjusts for the instrument's reliability. If an instrument's reliability is close to 1, the RCI score is similar to the observed difference score. This reliability adjustment yields a more conservative estimate of change than the observed difference score. The RCI used in the current analysis was calculated by dividing the difference between the 6 month and intake CBCL 6-18 Total Problems t scores by the standard error of measurement (SEM). The SEM is calculated using the standard error of the intake score and the reliability term for the CBCL 6-18 Total Problems scale, in this case Cronbach's α=0.95. The RCI is interpreted in the same way as a standardized test statistic (i.e., z-score), with values ≤–1.96 and values ≥1.96 indicating significant change. In the case of the CBCL 6-18 Total Problems scale, negative RCI scores indicate a reduction in symptoms between intake and 6 months, whereas positive RCI scores indicate an increase in symptom severity.

Data analysis

We explored differences in subject characteristics by medication status using t-tests and χ² tests, as appropriate. These characteristics included age, gender, race/ethnicity, family poverty status, co-occurring diagnoses, symptom severity at intake, and service intensity.

We used multilevel mixed-effects multivariate logistic regression to evaluate the association between symptom severity at intake and medication use, controlling for the demographic and service-related subject characteristics. These analyses estimated odds ratios (OR) and 95% confidence intervals for the likelihood of medication use. Findings in which the OR was >1.00 and the confidence interval did not include 1.00 indicated a greater likelihood of medication use. Findings in which the OR was <1.00 and the confidence interval did not include 1.00 indicated a lower likelihood of medication use. To examine the association between medication use and change in symptom severity between intake and 6 months, we used multilevel mixed-effects multivariate linear regression, controlling for the previously mentioned subject characteristics. For ease of interpretation, we then translated the coefficients of the linear regression model into standardized CBCL change scores by multiplying the RCI score by the SEM (SEM=3.0137).

For all regression analyses, we included a random effect for CMHI-funded system of care community, as variations in community-specific populations and service types presented the potential for confounding. Analyses were conducted using Stata, Release 11 (Stata Corporation, College Station, TX). All tests were two-tailed.

Results

Among all children and adolescents 6–18 years of age in our sample, 62.0% had taken psychotropic medication at any time during their first 6 months of system of care services. After limiting our sample to include only our two medication status groups—1) those who did not take medication, and 2) those who were taking medication both at 6 months and during the previous 6 months—44.6% took psychotropic medication.

Results are presented in Tables 1 –4. Sample sizes in Tables 2 and 4 were reduced from 888 to 617 because of listwise deletion of cases missing any of the variables included in the regression models. The sample size in Table 3 was reduced from 888 to 850 because of cases missing the CBCL 6-18 either at intake or at 6-month follow-up.

Table 1.

Psychotropic Medication Status by Subject Characteristics

		Medication status
	Total sample n (%)	Took no medication n (%)	Took medication n (%)	p
Total	888	492 (55.4)	396 (44.6)	<0.001
Age
Years, M(SD)	12.3 (3.09)	12.7 (3.10)	11.8 (3.00)	<0.001
6–9 Years	184 (20.7)	87 (47.3)	97 (52.7)	<0.001
10–14 Years	444 (50.0)	230 (51.8)	214 (48.2)
15–18 Years	260 (29.3)	175 (67.3)	85 (32.7)
Gender
Male	546 (61.5)	283 (51.8)	263 (48.2)	<0.001
Female	342 (38.5)	209 (61.1)	133 (38.9)
Race/Ethnicity
White	449 (50.9)	206 (45.9)	243 (54.1)	<0.001
Black/African American	223 (25.3)	143 (64.1)	80 (35.9)
Hispanic/Latino	140 (15.9)	83 (59.3)	57 (40.7)
Other	70 (7.9)	59 (84.3)	11 (15.7)
Poverty status
At/below poverty	571 (69.6)	340 (59.5)	231 (40.5)	<0.001
Above poverty	249 (30.4)	112 (45.0)	137 (55.0)
Symptom severity at intake
CBCL Total Problems score, M(SD)	68.4 (9.77)	65.0 (10.1)	72.6 (7.3)	<0.001
Below clinical level	212 (24.3)	179 (84.4)	33 (15.6)	<0.001
At/above clinical level	661 (75.7)	302 (45.7)	359 (54.3)
Co-occurring Diagnoses
1 diagnosis	333 (43.6)	221 (66.4)	112 (33.6)	<0.001
2 diagnoses	261 (34.2)	128 (49.0)	133 (51.0)
3+ diagnoses	170 (22.2)	64 (37.7)	106 (62.3)
Service intensity
Number of service types, M(SD)	5.3 (3.0)	4.1 (2.5)	6.4 (3.0)	<0.001
1–4 service types	370 (43.8)	261 (70.5)	109 (29.5)	<0.001
5–8 service types	309 (36.6)	117 (37.9)	192 (62.1)
9+ service types	166 (19.6)	19 (19.4)	79 (80.6)

M, mean; SD, standard deviation.

Table 2.

Multilevel Mixed-Effects Logistic Regression Estimates of Medication Use on Subject Characteristics (n =617)

	B	SE(B)	95% CI	OR	p<0.001
Age (years)	−0.067	0.040	(−0.145, −0.011)	0.936	0.094
Gender (reference=male)	−0.429	0.220	(−0.860, 0.001)	0.651	0.051
Race (reference=white)
Black/African American	−0.530	0.272	(−1.063, 0.003)	0.589	0.051
Hispanic	−0.585	0.339	(−1.249, 0.079)	0.557	0.084
Other	−0.416	0.598	(−1.589, 0.757)	0.660	0.487
Above poverty	0.539	0.236	(0.076, 1.003)	1.714	0.023
Symptom severity at intake	0.069	0.014	(0.043, 0.096)	1.072	<0.001
Co-occurring diagnoses (reference=1 diagnosis)
2 diagnoses	0.867	0.242	(0.392, 1.341)	2.379	<0.001
3+ diagnoses	1.292	0.294	(0.716, 1.868)	3.639	<0.001
Service intensity	0.299	0.042	(0.217, 0.382)	1.349

B, beta coefficient; SE, standard error; CI, confidence interval; OR, odds ratio.

Table 3.

Comparison of Mean CBCL ^a 6-18 Total Problems Scores at Intake and 6 Months

	n	Mean (SD)	p
Total sample
Intake	850	68.54 (9.69)	<0.001
6 months	850	66.22 (10.58)
Subjects who took no psychotropic medication
Intake	467	65.21 (10.09)	<0.001
6 months	467	62.32 (11.01)
Subjects who took psychotropic medication
Intake	383	72.62 (7.38)	<0.001
6 months	383	70.96 (7.74)
Mean Total Problems Score change, Intake-6 Months
Subjects who took no psychotropic medication	467	−2.88 (7.45)	0.01
Subjects who took psychotropic medication	383	−1.65 (6.45)

Child Behavior Checklist. Scores from the CBCL 6–18 Total Problems scale between 60 and 63 are in the borderline clinical range; scores of at least 64 are in the clinical range.

Table 4.

Multilevel Mixed-Effects Linear Regression Estimates of Relationship Between Change in CBCL ^a 6–18 Total Problems Scores Between Intake and 6 Months and Subject Characteristics (n =617)

	B	SE(B)	95% CI	p
Taking medication	−0.765	0.202	(−1.161, −0.368)	<0.001
Age (years)	0.014	0.088	(0.078, 0.118)	0.620
Gender (reference=male)	−0.154	.0181	(−0.509, 0.202)	0.397
Race (reference=white)
Black/African American	−0.315	0.207	(−0.721, 0.091)	0.129
Hispanic	−0.343	0.275	(−0.882, 0.196)	0.212
Other	0.428	0.366	(−0.289, 1.144)	0.242
Above poverty	−0.284	0.194	(−0.664, 0.095)	0.142
Symptom severity at intake	0.098	0.010	(0.078, 0.118)	<0.001
Co-occurring diagnosis (reference=1 diagnosis)
2 diagnoses	−0.445	0.202	(−0.840, −0.503)	0.027
3+ diagnoses	−0.550	0.232	(−1.005, −0.094)	0.018
Service intensity	−0.040	0.031	(−0.101, 0.020)	0.189

Child Behavior Checklist.

B, beta coefficient; SE, standard error; CI, confidence interval.

Table 1 presents findings related to our first research question regarding differences in subject characteristics by medication status. Younger children, males, whites, and subjects living above poverty were more likely to take psychotropic medication than their counterparts. Additionally, subjects with greater symptom severity at intake, with more co-occurring diagnoses, and receiving more services were more likely to take psychotropic medication than their counterparts. Subjects with symptom severity in the clinical range at intake were more than three times as likely to take medication (54.3%) than those whose symptoms were below the clinical level at intake (15.6%). Subjects with three or more co-occurring diagnoses (62.3%) were almost twice as likely to take medication between intake and 6 months as those with a single diagnosis (33.6%).

Subjects who took medication received more types of services, on average, than those who did not take medication. Subjects receiving five to eight types of services were more than twice as likely to take medication as those receiving one to four types of services (62.1% compared with 29.5%).

Also related to our first research question, Table 2 shows the multivariate associations between medication use and subject characteristics. Living above the poverty level was associated with medication use; the odds of subjects living above poverty taking medication was 1.71 times the odds of subjects living at or below poverty taking medication (OR=1.71, p<0.05). Level of emotional and behavioral symptoms and having multiple co-occurring diagnoses were also associated with medication use. For every one point increase in CBCL 6-18 Total Problems score at intake, there was a 7% increase in the odds of medication use (OR=1.07, p<0.001). Subjects with two diagnoses were more than twice as likely (OR=2.38, p<0.001), and subjects with three or more diagnoses were more than three times as likely (OR=3.64, p<0.001), to take medication as were subjects with one diagnosis. Service intensity (measured as the total number of different types of services received) was related to medication use; for each additional type of service accessed, there was a 35% increase in odds of medication use (OR=1.35, p<0.001).

For our second research question, we examined the association between medication use and change in emotional and behavioral symptoms within the first 6 months of receiving services. We first compared mean CBCL Total Problems scores at intake with those at 6 months using a series of paired t-tests (Table 3). Both those who took no medication and those who took medication showed significant symptom reduction on average between intake and 6 months. On average, subjects who took no medication scored within the clinical range (M=65.2, SD=10.09) at intake and, after a CBCL score reduction of 2.9 points, fell to within the borderline clinical range at 6 months (M=62.3, SD=11.01); this indicates a statistically significant change in mean CBCL Total Problems scores between intake and 6 months (t[466]=8.37, p<0.001). Subjects who took medication scored, on average, within the clinical range at intake (M=72.6, SD=7.38) and, despite the mean CBCL score reduction of 1.7 points, remained in the clinical range at 6 months (M=71.0, SD=7.74); this signifies a statistically significant change in mean CBCL Total Problem scores between intake and 6 months (t[382]=5.01, p<0.001). We also tested whether those changes in mean CBCL Total Problems scores for subjects who did and did not take medication were statistically different from each other. In fact, subjects who took medication experienced significantly less change in mean CBCL Total Problem score than did those who did not take medication (t[848]=−2.55, p=0.011).

Also related to our second research question, Table 4 illustrates multivariate associations between change in CBCL 6-18 Total Problems score and medication use, controlling for other subject characteristics. We converted the regression coefficients into standardized CBCL change scores by multiplying the coefficients by the SEM (SEM=3.0137). Medication use was significantly associated with greater decreases in intake-to-6-month CBCL 6-18 Total Problems scores. The RCI for subjects who took medication (RCI=−0.77) translated into a 2.31 point decrease in CBCL score between intake and 6 months, a significant indicator of symptom reduction, compared with subjects not taking medication.

For every one point increase in CBCL score at intake, RCI (RCI=0.10) increased, equivalent to a 0.30 point increase in CBCL score. The positive coefficient for this variable signified that greater symptom severity at intake was significantly associated with an increase in symptoms between intake and 6 months. Subjects with more than one DSM-IV diagnosis were significantly more likely to have a reduction in symptoms at 6 months than subjects with one diagnosis. Compared with subjects with one diagnosis, those with two diagnoses experienced a 1.3 point greater reduction in CBCL score (0.45 greater change in RCI), and subjects with three diagnoses experienced a 1.7 point greater reduction in CBCL score (0.55 greater change in RCI).

Discussion

This study examined psychotropic medication use and clinical outcomes among children and adolescents 6–18 years of age receiving mental health services in naturalistic treatment settings within CMHI system of care communities. Although this group is not nationally representative, it does include children and adolescents living in urban and rural settings and Native American tribes in 21 states. Therefore, the CMHI system of care population may provide a more representative perspective than the state-based or regional pharmacy claims data sets or small clinical trial populations that are more typically analyzed. This study is among the few reports of medication use among children and adolescents that are based on actual medication taken, rather than prescriptions written or filled. It is also among only a few studies reporting the association between psychotropic medication use and clinical outcomes among children and adolescents treated in naturalistic treatment settings.

In addition to examining associations between psychotropic medication use and the typically investigated demographic characteristics, this analysis included three measures of illness severity: symptom severity at intake, co-occurring diagnoses, and service intensity. Few previously reported studies of psychotropic medication use among children and adolescents examine associations other than with age, gender, race/ethnicity, and poverty status. By including these measures of illness severity in addition to demographic characteristics, this investigation broadens the understanding of the relative association between medication use and a wider array of characteristics. When controlling for both demographic characteristics and illness severity measures as covariates, the only demographic characteristic that showed a significant association with psychotropic medication use was living above the federal poverty level. This association is consistent with other studies (Leslie et al. 2003; Froehlich et al. 2007; Pavkov and Walrath 2008). However, our analysis found that psychotropic medication use was more strongly associated with measures of illness severity than with demographic characteristics, including poverty status. By including these illness severity measures, the relative influence of demographic characteristics diminished compared with the influence of illness severity on the likelihood of medication use.

Similarly, only a few previously reported studies have examined the association between psychotropic medication use among children and adolescents and symptom reduction within naturalistic treatment settings (Jerrell 2008; Fontanella et al. 2009; Molina et al. 2009). Studies measuring symptom change over time have traditionally been limited to clinical trials (Kapczinski et al. 2003; Wolraich 2003; Ipser et al. 2009; Molina et al. 2009). In this study of naturalistic treatment, our unadjusted comparison of CBCL scores found that children and adolescents taking medication showed less symptom improvement at 6 months than those not taking medication. In contrast, when controlling for our covariates, particularly symptom severity at intake, we found that subjects taking medication experienced more symptom improvement at 6 months than those not taking medication. One possible explanation for this difference in findings is an interaction effect between symptom severity at intake and medication benefit, with those with more severe symptoms at intake more likely to take medication, but also less likely to experience symptom improvement over time.

Our analysis also found that, overall, medication use was not associated with sufficient symptom improvement to reduce symptom severity below clinical level. Similar to findings of Molina et al. (2009), medication use in community treatment settings did not, on average, normalize symptoms among subjects in this population. Because this clinical change was not sufficient to reduce symptoms below the clinical level on average, subjects who took psychotropic medication were simply “better but not well” (Frank and Glied 2006).

Limitations

This study has several limitations. Our data are based on children and adolescents receiving services within system of care communities that specifically treat children and adolescents with serious emotional disturbance and serve primarily Medicaid recipients. Thus, our sample does not represent all children taking medication. Our data are based on caregiver report, rather than observable medication behavior. These data rely on the caregiver's ability to accurately assess whether the subject actually was taking his or her medication, ability to recall medication use over 6 months, and willingness to accurately provide this potentially sensitive information during the data collection interviews. In addition, the CBCL 6–18 ratings were completed by respondents designated only as caregivers. Because the actual identity of the caregiver was not revealed to protect confidentiality, the possibility exists that the person designated as caregiver may have changed between intake and the 6-month follow-up, particularly among caregivers of children or adolescents in foster care (7.9%). In those few cases in which child-caregiver dyads may have changed, bias may have been increased due to recall inconsistency.

Our data on medication use do not specify when medication was actually started, the duration, dosage, or medication adherence. Likewise, our data on symptom severity as reported on the CBCL 6–18 are based on caregiver report. However, subjective determination of symptom severity based on either the caregiver or clinician perspective is standard procedure, given the current lack of biological measures on which to determine illness severity.

By limiting our medication group only to subjects who were taking medication at their 6 month follow-up interview and had also taken medication during the previous 6 months, it is possible that our results are less representative of all system of care children and adolescents who take medication. None of the subjects in this population took medication only, in the absence of other services for their emotional and behavioral symptoms. It is not possible, therefore, to identify the effect that these other services may have had on the measure of symptom severity at 6 months.

This investigation is not a study comparing psychotropic medication use with non-medication therapeutic service interventions. It is also important to keep in mind that these data do not imply a causal relationship between taking medication and change in symptoms. Because this investigation studied children and adolescents 6–18 years of age, findings cannot be generalized to younger children or adults.

Conclusion

These findings demonstrated that children and adolescents who take psychotropic medication in naturalistic settings tended to have more severe illness than those who did not take medication, and that illness severity was more important in predicting medication use than other demographic characteristics. This suggests that future studies associating subject characteristics with medication use should encompass the broader clinical picture, rather than restricting this association to the traditional demographic characteristics of age, gender, race/ethnicity, and poverty status.

Controlling for illness severity, these findings demonstrated that there was more symptom improvement with psychotropic medication use than with no medication use. However, this improvement was not sufficient to reduce symptom severity below clinical level within 6 months, even among these subjects receiving additional therapeutic and supportive services. These findings acknowledge the important contribution that psychotropic medication has in the treatment of children and adolescents with serious mental health challenges, but also illustrates that this contribution is relatively limited. Even after 6 months of treatment, emotional and behavioral symptoms of subjects taking psychotropic medication remained well above the clinical level. Psychotropic medication may have some effect, but these findings suggest that medication effectiveness may not be fully successful in normalizing serious emotional and behavioral symptoms. This demonstration of the limited effectiveness of psychotropic medications, even in combination with therapeutic and supportive services, emphasizes the importance of continuing to search for more effective treatment options.

Several research questions remain to be explored. Further examination of additional subject characteristics and symptom change may continue to help identify which subjects respond well to medication treatment and which do not. Specifically, further research exploring the interaction effect between symptom severity and medication effectiveness may help determine whether subjects with more severe symptoms experience more or less medication benefit than those with less severe symptoms. Similarly, further research controlling for medication use prior to intake may help determine whether these results are maintained among children and adolescents with no prior medication use. Further examination of these associations over 12 months, rather than 6 months, could indicate whether these findings are sustained. Because this investigation looked only at overall medication use, further analysis of the relative association between specific medication classes and change in symptom severity is warranted, particularly among subjects with specific diagnoses.

Clinical Significance

Because existing literature on psychotropic medication use rarely reports clinical outcomes outside of clinical trials, this investigation provides relatively unique data on psychotropic medication effectiveness among children and adolescents treated in naturalistic treatment settings. Our finding that indicators of illness severity—symptom severity at intake, co-occurring diagnoses, and service intensity— are more strongly associated with psychotropic medication use and clinical outcomes than any of the demographic characteristics serves to broaden the current understanding of factors associated with psychotropic medication use and outcomes. Our analysis also demonstrated that although subjects taking psychotropic medication showed more symptom reduction than those not taking medication, their symptom severity remained in the clinical range. This finding that medication provided some effect on clinical symptoms, but not a sufficient effect to normalize symptoms, contributes to the understanding of medication effectiveness as administered in community treatment settings, and the limitations of this effectiveness.

Footnotes

Disclosures

No competing financial interests exist.

References

Achenbach

. Manual for the ASEBA School-age Forms and Profiles. Burlington, VT: University of Vermont, Research Center for Children, Youth & Families, 2001.

American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders 4thedi. (DSM-IV)Washington, DC: American Psychiatric Association, 1994.

Center for Health Services Research on Pharmacotherapy, Chronic Disease Management and Outcomes Website, Rutgers University. www.ihhcpar.rutgers.edu/org_units/default.asp?v=2&o=27. 2010 May 17.

Coletti

, Leigh

, Gallelli

, Kafantaris

. Patterns of adherence to treatment in adolescents with bipolar disorder. J Child Adolesc Psychopharmacol, 15:913–917. 2005.

Emslie

, Mayes

, Laptook

, Batt

. Predictors of response to treatment in children and adolescents with mood disorders. Psychiatr Clin North Am, 26:435–456. 2003.

Fontanella

, Bridge

, Campo

. Psychotropic medication changes, polypharmacy, and the risk of early readmission in suicidal adolescent inpatients. Ann Pharmacother, 43:1939–1947. 2009.

Frank

, Glied

. Better but Not Well: Mental Health Policy in the United States since 1950. Baltimore: Johns Hopkins University Press, 2006.

Froehlich

, Lamphear

, Epstein

, Barbaresi

, Katusic

, Kahn

. Prevalence, recognition, and treatment of attention-deficit/hyperactivity disorder in a national sample of US children. Arch Pediatr Adolesc Med, 161:857–864. 2007.

Hudson

, Miller

, Kirby

. Explaining racial and ethnic differences in children's use of stimulant medications. Med Care, 45:1068–1075. 2007.

10.

Ipser

, Stein

, Hawkridge

, Hoppe

. Pharmacotherapy for anxiety disorders in children, adolescents. Cochrane Database Syst Rev, 3CD0051702009.

11.

Jacobson

, Truax

. Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. J Consult Clin Psychol, 59:12–19. 1991.

12.

Jerrell

. Pharmacotherapy in the community-based treatment of children with bipolar I disorder. Hum Psychopharmacol, 23:53–59. 2008.

13.

Julius

, Novitsky

Jr , Dubin

. Medication adherence: A review of the literature and implications for clinical practice. J Psychiatr Pract, 15:34–44. 2009.

14.

Kapczinski

FFK

, Silva de Lima

, dos Santos Souza

JJSS

, Batista Miralha da Cunha

AABC

, Schmitt

RRS

. Antidepressants for generalized anxiety disorder. Cochrane Database of Syst Rev, 2CD0035922003.

15.

Leslie

, Weckerly

, Landsverk

, Hough

, Hurlburt

, Wood

. Racial/ethnic differences in the use of psychotropic medication in high-risk children and adolescents. J Am Acad Child Adolesc Psychiatry, 42:1433–1442. 2003.

16.

Martin

, Leslie

. Psychiatric inpatient, outpatient, and medication utilization and costs among privately insured youths, 1997–2000. Am J Psychiatry, 160:757–764. 2003.

17.

Martin

, Van Hoof

, Stubbe

, Sherwin

, Scahill

. Multiple psychotropic pharmacotherapy among child and adolescent enrollees in Connecticut Medicaid managed care. Psychiatr Serv, 54:72–77. 2003.

18.

Molina

BSG

, Hinshaw

, Swanson

, Arnold

, Vitiello

, Jensen

, Epstein

, Hoza

, Hechtman

, Abvikoff

, Elliott

, Greenhill

, Newcorn

, Wells

, Wigal

, Gibbons

, Hur

, Houck

. MTA Cooperative Group. The MTA at 8 years: Prospective follow-up of children treated for combined-type ADHD in a multisite study. J Am Acad Child Adolesc Psychiatry, 48:484–500. 2009.

19.

Moreno

, Laje

, Blanco

, Jiang

, Schmidt

, Olfson

. National trends in the outpatient diagnosis and treatment of bipolar disorder in youth. Arch Gen Psychiatry, 64:1032–1039. 2007.

20.

National Community Pharmacists Association: Take as directed: a prescription not followedPress release.www.ncpanet.org/media/releases/2006/take_as_directed.php. 2010 May 14.

21.

Olfson

, Blanco

, Liu

, Moreno

, Laje

. National trends in the outpatient treatment of children and adolescents with antipsychotic drugs. Arch Gen Psychiatry, 63:679–685. 2006.

22.

Pavkov

, Walrath

. Clinical and non-clinical characteristics associated with medication use among children with serious emotional disturbance. J Child Fam Stud, 17:839–852. 2008.

23.

Riddle

, Kastelic

, Frosch

. Pediatric psychopharmacology. J Child Psychol Psychiatry, 42:73–90. 2001.

24.

Teich

, Buck

, Graver

, Schroeder

, Zheng

. Utilization of public mental health services by children with serious emotional disturbances. Adm Policy Ment Health, 30:523–534. 2003.

25.

United States Department of Health and Human Services (US DHHS): HHS Poverty Guidelines. aspe.hhs.gov/poverty/figures-fed-reg.shtml. 2009 December 12.

26.

United States Public Health Service: Report of the Surgeon General's Conference on Children's Mental Health: A National Action Agenda. Washington, DC: Department of Health and Human Services, 2000.

27.

Visser

, Lesesne

, Perou

. National estimates and factors associated with medication treatment for childhood attention-deficit/hyperactivity disorder. Pediatrics, 119:S99–S106. 2007.

28.

Warner

, Pottick

, Bilder

. Clinical and organizational correlates of medication for youths in U.S. mental health services. Soc Serv Rev, 454–481. 2005.

29.

Warner

, Pottick

, Mukherjee

. Use of psychotropic medications by youths with psychiatric diagnoses in the U.S. mental health system. Psychiatr Serv, 55:309–311. 2004.

30.

Wolraich

. Annotation: The use of psychotropic medications in children: an American view. J Child Psychol Psychiatry, 44:150–168. 2003.

31.

Zito

, Safer

. Sources of data for pharmacoepidemiological studies of child and adolescent psychiatric disorders. J Child Adolesc Psychopharmacol, 7:237–253. 1997.

32.

Zito

, Safer

. Recent child pharmacoepidemiological findings. J Child Adolesc Psychopharmacol, 15:5–9. 2005.

33.

Zito

, Safer

, dosReis

, Gardner

, Magder

, Soeken

, Boles

, Lynch

, Riddle

. Psychotropic practice patterns for youth: A 10-year perspective. Arch Pediatr Adolesc Med, 157:17–25. 2003.