An Evaluation of Atypical Antipsychotic Use, Costs and Effectiveness in the Pediatric Population

The pediatric mental health burden in the United States (US) is substantial, with more than 4 million children meeting diagnostic criteria for a mental health disorder. As of 2014, this number represented 20% of US children and adolescents. In 2010, mental health disorders are estimated to cost children and their families $247 billion dollars annually and severely impact quality of life for children and their families. From 2007 to 2010, inpatient admissions for mental health disorders in pediatric patients increased 24% and mood disorder admissions in pediatric patients increased 80% from 1997-2010. An estimated $11.6 billion was spent on pediatric mental health hospitalizations from 2006 through 2011, with public sources such as Medicaid and Medicare responsible for approximately 50% of the payments, leaving 50% to private payers. This economic and clinical concern has led pediatric medical associations and health quality agencies to increase support and funding for pediatric mental health research and treatment. Medication therapy is a common intervention in mental health treatment and atypical antipsychotics are increasing in utilization, often becoming first-line therapy. Despite available data describing the need to treat pediatric mental health conditions, the available evidence for clinical effectiveness and economic impact of atypical antipsychotics (AAPs) has many shortfalls. Most available research is derived from patients utilizing publicly-funded medical care, such as Medicaid or Medicare resources, with little data available about patients with privately-funded care. To help address this gap in the literature, we used a large, privately-insured, US population for our analysis. We examined if the increased trend in AAP utilization from previous research is also present in this pediatric population. Considering the payer perspective, we evaluated the cost of AAP medication therapy based on most recent utilization. Available studies lack information about the direct costs of pediatric mental health treatment and efficacy of psychiatric medications in the pediatric population. Most efficacy studies are based on clinical trials necessary for pediatric indication approval from regulatory agencies such as the Food and Drug Administration (FDA). Many of the AAP medications do not have pediatric clinical trial evidence available and are frequently utilized without pediatric indications. The available data suggests that off-label prescribing is not an uncommon practice in the pediatric patient population. Approximately half of atypical antipsychotics do not have pediatric indications but are increasingly used, particularly in treating behavior disorders, due to such factors as improved patient compliance and improved side effect profiles. Limited formal studies examining atypical antipsychotic use compared to other agents in the class have been conducted. Studies with direct comparisons have yet to be conducted in the pediatric population with mental health disorders. The manuscripts that comprise this dissertation aim to provide new insights into available trend and utilization patterns of atypical antipsychotic medication use in children. This research characterized the prevalence of atypical antipsychotic use in pediatric patient with mental health conditions in a large, privately insured US population, evaluating the diagnoses associated with treatment and estimate the cost of AAP medication therapy in this population. This research determined if the trends observed in publicly-insured children persist in the privately-insured, pediatric patient. The analysis evaluated annual trends in prevalent use of atypical antipsychotic medication over 6-year period in this pediatric population and evaluated the appropriate use of AAPs for mental health diagnoses. Lastly, an evaluation determined if specific antipsychotic therapy delayed time to readmission among privately-insured children following a psychiatric hospital admission. The results of this dissertation will provide new insights regarding the trends and direct medication costs of atypical antipsychotic agents when utilized in pediatric patients with mental health disorders. Manuscript 1: This analysis focused on characterizing the most recent (2015) AAP use in the pediatric population with mental health disorders, using a large, US population of privatelyinsured children. The study evaluated if the prevalence data observed among publicly insured children persists. Characterization of the prescribing trends for atypical antipsychotics and the medication costs of the use in this population were examined. Patterns of use across demographics and associated mental health diagnoses were characterized by the class of medication. This study focused on the prevalent use of AAPs in pediatric patients with a mental health diagnosis, evaluated the mental health diagnoses associated with AAPs and the direct cost burden of medication therapy associated with this use of AAP in the pediatric population to the private payer. Manuscript 2: This research evaluated the trends in the prescribing of atypical antipsychotic medications from 2010 through 2015 in this privately-insured pediatric population. The trends of AAP use in the pediatric population over six years were examined. The associated mental health diagnoses corresponding with AAP prescribing were described to examine the off-label diagnoses treatment prevalence in this population. This study hypothesizes that the prevalent use of AAPs is increasing in the privately-insured patients and off-label prescribing accounts for most clinical use in pediatric patients. Manuscript 3: This analysis examined pediatric patients who utilized oral atypical antipsychotic therapy after an inpatient admission for mental health treatment. Readmission for mental health treatment was evaluated to determine the efficacy of using oral AAP medications in pediatric mental health patients. Some oral AAP agents have shown benefit in pediatric patients compared to placebo and have an official FDA indication for pediatric use. Many clinical providers believe that this entire class of medications can demonstrate benefit in pediatric patients, regardless of FDA indication. This study hypothesized that certain oral AAP medications are associated with delayed readmission in pediatric patients with an index admission for mental health treatment.

50% to private payers. This economic and clinical concern has led pediatric medical associations and health quality agencies to increase support and funding for pediatric mental health research and treatment.
Medication therapy is a common intervention in mental health treatment and atypical antipsychotics are increasing in utilization, often becoming first-line therapy.
Despite available data describing the need to treat pediatric mental health conditions, the available evidence for clinical effectiveness and economic impact of atypical antipsychotics (AAPs) has many shortfalls. Most available research is derived from patients utilizing publicly-funded medical care, such as Medicaid or Medicare resources, with little data available about patients with privately-funded care. To help address this gap in the literature, we used a large, privately-insured, US population for our analysis. We examined if the increased trend in AAP utilization from previous research is also present in this pediatric population. Considering the payer perspective, we evaluated the cost of AAP medication therapy based on most recent utilization.
Available studies lack information about the direct costs of pediatric mental health treatment and efficacy of psychiatric medications in the pediatric population.
Most efficacy studies are based on clinical trials necessary for pediatric indication approval from regulatory agencies such as the Food and Drug Administration (FDA).
Many of the AAP medications do not have pediatric clinical trial evidence available and are frequently utilized without pediatric indications. The available data suggests that off-label prescribing is not an uncommon practice in the pediatric patient population. 3,4 Approximately half of atypical antipsychotics do not have pediatric indications but are increasingly used, particularly in treating behavior disorders, due to such Stephen Kogut for his continued support and motivation while in graduate school. Dr.
Kogut was instrumental in making this program possible while in active service to the US Navy. He is a wonderful educator, mentor and friend. If he was not willing to support me in the aggressive academic timeline and research deadline, my graduate degree would not be feasible.
The mentorship and professional support that several members of the US Navy have given me from the beginning of this graduate program must be mentioned. This graduate program has been a blessing to our family and my own professional success. For anyone that is not directly mentioned, I apologize, but please know your support has been crucial to my achievements.

OBJECTIVE:
This study will determine the prevalence and costs of atypical antipsychotic medication therapy in the privately-insured pediatric patient population with a mental health disorders. This study will also identify patient and clinical characteristics that influence the use of atypical antipsychotic medication in pediatric patients with mental health diagnoses.

Introduction
The pediatric mental health burden in the United States (US) is substantial, with more than 4 million children meeting diagnostic criteria for a mental health disorder. 2 privately-insured patients. It is unclear whether the available prevalence of AAP use among publicly insured children is also comparable to that among privately-insured pediatric populations with mental health diagnoses. 8,16,18,19 The goal of this study is to characterize AAP use in the pediatric population with mental health diagnoses, using a large, US population of privately-insured children. Characterization of the prescribing prevalence for atypical antipsychotics and the medication costs of the use in this population will be examined. Examining this population for changes in prescribing over the most recent year can provide additional insight into spending trends and changes in payer spending for AAP therapy. This study will evaluate how new market entries and new generic medications have possibly changed the spending profile and may provide additional data on the medication costs differences seen in this study compared to available literature.
Patterns of use across demographics and associated mental health diagnoses will be described to better characterize the use of this class of medication in the pediatric population with mental health diagnoses. Considering the payer perspective, we evaluated the direct cost burden of AAP medication therapy for 2015, the most recent year of available data. We evaluated the overall utilization of these medications among privately-insured pediatric patients with mental health diagnoses. We discuss a comparison of the overall utilization observed in our study to available reported utilization among publicly-insured children.

Data Source and Study Design
This cross-sectional study was conducted utilizing administrative data (Optum To determine the diagnosis associated with a specific AAP paid claim, a 60-day window (60 days before and after) around the date of prescription fill was established and the closest medical visit was selected to ascertain the diagnoses. If the 60 days window fell outside the study period, then patients were excluded for not having sufficient data for complete analysis. For patients that received no AAP therapy in 2015, the diagnoses from the most recent medical provider visits in 2015 were used.
Per medical visit, only the first ten fields were used because these captured 95% of available diagnosis information in the database.
Among pediatric patients with mental health diagnoses, we determined the frequency and percentage of patients with AAP use compared to those without AAP therapy in 2015. Differences in patient characteristics and prescription claim information between the two groups were determined using the Chi-Square test. and was the model employed in the final analysis. 26 All statistical tests were two-sided and performed at a 0.05 significance level and conducted using SAS Enterprise Guide Version 7.1 (Cary, North Carolina, USA).  Results from the unadjusted and adjusted logistic regression models are displayed in Overall, demonstrated use of other psychotropic medication classes were significant predictors in the prescribing or not prescribing of AAP medication therapy. Of note, in the adjusted model, several variable associations changed directions when adjusted for other covariates. In the adjusted model, female gender, use of stimulants and use of antidepressants were associated with a reduced risk of AAP use, which was a change from an increased risk in their respective univariate models. By adding one variable to the model at time, we determined that these estimates changed direction after adjustment for age group. This indicates that the age group of the patient at time of paid claim demonstrates some unmeasured confounding affecting the other covariates, that is independent of the direct interaction of the variables.

Results
In the adjusted model, prescriber specialty was a significant predictor for AAP prescribing. Patients seen by a specialty provider had 2.5 times the increased odds of receiving a paid claim for AAP medication than patients seen by a primary care provider (aOR=2.5; 95% CI=2.3, 2.7). Patients seen by a specialist had 5 times the increased odds of receiving an AAP medication paid claim than children seen by a non-physician mental health professional (psychiatric nurse practitioner, physician assistant) (aOR= 5.0; 95% CI=4.5, 5.6). After adjusting for other patient characteristics, region was no longer a statistically significant predictor of AAP prescribing in the final model.

Cost Model for AAP Medications Use in Pediatric Patients
The total 2015 annual expenditure for AAP prescriptions in the pediatric The median cost per paid claim in 2015 is displayed by medication in Table 4.
Generic risperidone was the most commonly prescribed atypical antipsychotic medication (25%) followed by generic aripiprazole (22%) and name brand Abilify® The cost data was then analyzed using a generalized linear model (GLM) to determine any covariates that were a significant predictor of PMPM costs. Table 5 presents the results of the log-gamma regression of the per-member per-month (PMPM) costs during the 12-month study period adjusted for patient demographics.
Using a gamma regression model with an identity link function, age group, gender, mental health diagnostic category and provider specialty were statistically significant predictors of total annual expenditure for AAP medication therapy. Children aged 6-12 years had overall adjusted mean spending for AAPs that was $90. 22

Discussion
In privately-insured children and adolescents, the prevalence of atypical antipsychotic medication therapy was 67.5 per 1000 patients (6.75%; 95% CI=6.6%, 6.9%) with a mental health diagnosis of interest present in 2015. In our study, gender was associated with differences in prescribing AAP therapy, which aligned with previous research in private-and publicly-insured children. 11,15,27 These previous studies found that patients of male gender had increased odds of receiving AAP medication therapy. Our study found similar increased odds in male patients. The children receiving AAP therapy were significantly older (13.6 vs. 11.8 years) and older age was an important predictor of a patient receiving AAP therapy.
In our analysis, the 2015 prevalence of atypical antipsychotic use was higher than determined in previously studied research of privately-insured children and adolescents across the US. 15 Previous studies have found that publicly-insured youth have consistently lower AAP prevalence to that found in our study, at 1.9% in 2005 and 1.7% in 2010. 11 Since 2005, the AAP medication therapy options have doubled, as AAP medication approvals have increased dramatically in the US. Furthermore, previous studies in publicly insured population included all children, not only children with mental health diagnoses present in medical records for their analysis. 8,11,19 Including all children in the analysis could increase the population that is considered at risk for AAP medication use, leading to a possible underestimation of the proportion of study participants that received AAP medication therapy. These differences in study population could explain some of these observed differences. Private insurance payers have different formulary practices than public payer systems. Formulary approval and reimbursement practices could change the utilization and diversity of a medication class and represent the difference between our study and the results from studies analyzing publicly funded patients. Combination therapy with multiple AAP medications or therapy switching was not examined in this analysis. Combination therapy or medication switching is common in mental health treatment recommendations and represents a future direction that should be explored. Future research should also examine overlapping medication classes with AAP therapy or switching therapy to AAP as a significant factor in AAP medication use in pediatric patient with mental health diagnoses.
Our study found that 29% of pediatric patients treated with an atypical antipsychotic have no mental health diagnosis in an associated claim for medical visit within 60 days of the paid prescription claim. found that 72% of subjects analyzed were missing a diagnostic code associated with paid claim and this issue was only resolved after 2006 once Medicaid rules required an appropriate code before paid claim would be fulfilled. 10 Other previous trend studies in public-and privately-insured pediatric patients categorized missing diagnoses as "other" or excluded patients with no diagnosis available completely. 8,28 The rate of missing mental health diagnosis found in our study was lower than previously published literature. A 2015 study found no mental health diagnosis present at an associated medication visit in 60% of pediatric patients treated with AAPs. 12 Previous research also found that in 75% of cases, all children with MH diagnoses of interest treated with AAP medications has multiple psychiatric diagnoses. 15 Similarly, we found that in the original cohort, 44% of children had multiple mental health diagnoses present during the study period.
We found that specialty providers were the leading prescribers (41%) associated with paid claims for AAP therapy. Olfson et al. 12 noted that specialists were the provider associated with AAP therapy in approximately 69% of paid claims in 2010. The differences in associated prescriber characteristics could be related to coding differences between private insurers and their claims process. Prescribing physician requirements can differ between private and public payers. Formulary requirements for certain payers require specialist prescribing for certain populations or medication classes that may not be required of practices in our privatively-insured population.
Concomitant medication therapy was a significant predictor of a pediatric In 2015, we estimated that $12.5 million was spent on atypical antipsychotic medication therapy among this privately-insured pediatric population. In a similar study of children enrolled in Florida Medicaid, researchers found that in Fiscal Year 2005 (FY2005) , $151 million was spent on AAP medication therapy. 30 This drastic difference compared to our study findings can most likely be explained by the peak utilization that was seen for AAP medication therapy in 2005 and lack of generic formulations available on the US market. Their study adjusted dollar spending amounts to align with the medical care component of the consumer price index for the region during the FY2005. 30 Our study took direct costs paid by the payer from prescription claims data. Most of the available cost research focuses on publiclyinsured children and our study is one of the first to explore the direct medication costs to a private, national payer. In 2004, the FDA issued advisory committee findings that recommended more conservative use of atypical antipsychotics in children and Pamer et. al examined a corresponding decrease in AAP medication use. 28,31 This research group observed a decline in AAP medication prescribing, but this decline did not achieve statistical significance nor did they examine overall spending or changes in average cost. 31

Limitations
The analysis was conducted using insurance claims data; therefore, limited clinical information was available for patients that were included in the final analyses.

Conclusion
The prevalence of using atypical antipsychotic medications in pediatric patients with mental health disorders is significant in the privately insured population. The prevalence in the privately insured population was 6.75% (CI=6.6%, 6.9%) or 67.

Introduction
The overall trend of use for AAP in pediatric patients with mental health disorders has increased over the last 20 years. of incident users were being treated for mood or behavior disorders, rather than traditional psychiatric conditions. 1,4,7,9 This treatment is often "off-label" or not for a specific FDA-approved condition in children or adolescents. 1,2,4,5 Furthermore, 20% of pediatric patients prescribed an atypical antipsychotic medication had no FDA approved diagnosis associated with treatment. 6 The medications of interest and the corresponding FDA approval are listed in Appendix B. 10 As of 2010, most of the trend studies focused on publicly-insured children, such as Medicaid enrollees, with few studies including large, privately-insured populations. The available data suggests a growing trend in atypical antipsychotic use in pediatric patients; however, there have been few studies that evaluated this medication class use in privately-insured patients.
More data in privately-insured children is unavailable and evaluation of the trend among privately-insured children has not been characterized nor compared to a population of publicly-insured children. 4,5,7,9,11 This study evaluated the overall utilization of these medications among privately-insured pediatric patients and discussed comparisons to publicly-insured children.
Clinical and demographic characteristics of youth receiving atypical antipsychotic medications are not fully understood. Available analyses of commercial and Medicaid prescription claims indicated that AAP treatment was significantly more common in boys than girls. 5,9,12,13 According to several state Medicaid studies, treatment of mood disorders, attention-deficit/hyperactivity disorder or disruptive and aggressive behavior disorders accounted for the majority of antipsychotic use. 4,5,14 One sample population of psychiatric outpatient visits found 77% of children had no diagnosis of any psychotic disorder associated with AAP medication therapy. 15  criticism and concern, leading to Medicaid earning reimbursement for spending for off-label prescribing. 16,17 Policy makers anticipate similar off-label utilization in the pediatric population. The proportion of prescriptions authorized for off-label use in pediatric patients has not yet been evaluated in the privately-insured population.
Prior research that indicated increased use of atypical antipsychotic medication in children, coupled with the potential increase in off-label use, have led to public and professional uncertainty regarding recommended treatment regimens. The goal of this study was to evaluate annual trends in prevalent use of AAP medication over 6-year period from 2010 to 2015 in a large, privately-insured pediatric population and evaluate the appropriate use of AAPs for a given mental health diagnoses.
Appropriate use was determined by labeled FDA indications for AAP medication referenced in the paid claim. We hypothesized that AAP medication utilization increased over the 6-year available study period and off-label prescribing of AAP medication represented the predominant use in pediatric patients.

Methods
This retrospective cohort study identified all paid prescription claims for AAP medications used in pediatric patients from 2010 through 2015. For each calendar year of the study period, patients were identified that were 2 to <18 years of age (as of the start of the year) and filled at least 1 prescription for an AAP agent during the year.
We reported the number of patients who used any AAP overall and stratified by specific AAP agents. The utilization of AAPs was quantified as the prevalence of AAPs use; that is, the proportion of the pediatric population on AAPs during each year. Incidence of AAP therapy was also evaluated at the first year of follow-up to assess patients newly prescribed. All pediatric patients present for analysis in the administrative database (Optum Clinformatics ® Data Mart; OptumInsight, Eden Prairie, MN) were considered in the prevalence calculation.

Data Source and Study Design
This retrospective longitudinal study used the commercial data set (Optum Clinformatics ® Data Mart; OptumInsight, Eden Prairie, MN) from January 2010 to December 2015. This data includes commercial health insurance claims (inpatient and outpatient medical records, laboratory data, facility information, and outpatient pharmacy) and enrollment data from large, private insurer across the United States. 18 This dataset provides healthcare information on 36 million beneficiaries and encompasses 1.2 billion individual medical records.

Sample Selection
Patients aged 2 to 17 years of aged (at the start of the year with paid claim) that associated with the prescription. For both analyses, patients and paid claims were excluded if they had no medical visit within the reference window (+ 60 days around fill date) to provide clinical information for analysis. Ten diagnosis fields were queried for the associated reason for the medical visit. Medical visit data was carried forward for up to one year if it was missing at a particular medical visit. A one-year follow-up was used to ascertain all relevant clinical information for patients that were likely to be stabilized on AAP medication therapy and no longer presented to a provider for monthly medication refills. This ascertainment may explain repeated prescription paid claims for AAP medication therapy without a more recent medical appointment associated with the paid claim, since stable patients may be provided refills on an AAP prescription that do not require repeat monitoring by a provider. Pharmacy claims were recorded for all outpatient pharmacy plan claims and were coded with National Drug Codes (NDCs), with detailed information that included medication name, fill data, days' supply, quantity and drug strength. All pediatric patients were included in this analysis regardless of presence of associated mental health diagnosis at the associated medical visit.

Statistical Analysis
Overall AAP use prevalence was presented as a proportion of the pediatric population with mental health diagnoses prescribed AAPs in the cohort in the given year (no. of users per 1,000 children). 19 The total number of paid claims, unique patients and prevalence (represented as a percentage) is described in Table 1. The total number of children available in dataset was determined by examining all patients with at least one paid claim in each year for patients age 2 to 17 years. The number of paid claims from 2010 to 2015 for AAP medication in pediatric patients were examined overall and by AAP medication (Figure 1). Among patients with at least one AAP paid claim and an associated medical visit available for analysis, we determined the frequency and percentage of baseline patient and clinical characteristics ( Table 2).
We conducted a longitudinal analysis to evaluate the annual rates of AAP use, both overall and by medication class. Total counts of AAP paid claims per year was determined per patient. The unit of analysis for this section of the study was the medical visit for each patient associated with the AAP paid claim, which was then aggregated to a yearly count per patient. We examined a yearly count to better capture any market changes that may affect AAP prescribing, such as new drug approvals and generic formulations. Previous prescribing trends indicated that these market changes usually influence prescribing patterns over six to twelve months after the change is in effect. 3,4,11,12 The outcome of interest was the count of AAP claims per patients in each year. As mentioned above, 40,750 patients were included in the study cohort, representing 378,007 paid claims. The yearly claim count variable was determined by summing the individual paid claims for AAP medications for a given patient for each year during the study period. Mental health diagnosis associated with each paid claim was retained as the primary diagnosis for analysis unless a more recent paid claim was available with this information. Similarly, the provider details and specialty information were retained for each patient until a newer paid claim occurred with upto-date information available.
Because there are multiple visits per patient, a generalized estimating equation (GEE) model was used to estimate the prevalence of children prescribed AAPs over time accounting for correlation within patient. 20 A GEE model with Poisson variance and a log-link was used to evaluate the association of covariates with annual claim count of AAP paid claims per patient over the study period. 20 No interaction terms were included in the final model due to lack of statistical significance of these terms.
Collinearity between independent variables was tested using Variance Inflation   Table 1 and Figure 1 below outline the yearly AAP medication claim count over the study period.

Trends in Atypical Antipsychotic Prescribing
A full description of baseline characteristics for the study population is presented in Table 2. The mean age of pediatric patients receiving AAP therapy was Based on the GEE model among the aggregated annual data, the rate ratios of claim count are presented in Patients with developmental disorders and psychotic disorders did not have a significantly different rate of paid claims for AAP medications over the study period, when compared to patients with no mental health diagnosis. Overall, the category of prescribing provider responsible for the AAP paid claim did not have a significant effect on the rate of AAP paid claims during the study period.

Off-Label Diagnostic Prescribing of Atypical Antipsychotic Medications
Off-label diagnostic prescribing of atypical antipsychotics was common in pediatric patients in our study. During the study period, 62% (95% CI=62%, 63%) of paid claims for atypical antipsychotics in pediatric patients were classified as off-label diagnostic use. Much of the off-label diagnostic use was due to the lack of mental health diagnosis present in the medical visit (35%) associated with AAP paid claim.
No diagnostic code for a mental health condition at the medical visit associated with the paid claim was classified as off-label diagnostic use.
All covariates demonstrated significances as predictor in the univariate analysis ( Table 4). In the final multivariable model, age group (P value=0.05), gender (P value =0.002), mental health diagnosis (P value<0.001), provider category (P value=0.08), and US region (P value<0.001) were significant variables in the likelihood of off-label diagnostic prescribing of AAPs. The adjusted odds ratios are presented in Table 4. In the adjusted model, children 2-5 years old were 15% more likely (aOR)=1.15; 95% CI=1.0, 1.3) than children 13-17 years old to be prescribed atypical antipsychotics for off-label diagnostic indications. Children aged 6-12 years old were 2% less likely to have off-label diagnostic (aOR=0.98; 95% CI=0.93, 1.0) use compared to adolescents (ages 13-17 years). Female pediatric patients were 10% times more likely (aOR=1.1; 95% CI=1.0, 1.2) to be prescribed an AAP in an off-label diagnostic manner compared to male children. In the adjusted model, children located in the Midwest were 13% less likely (aOR=0.87; 95% CI=0.8, 0.97) to have an offlabel diagnostic AAP paid claim, compared to children located in the Northeast.
Similarly, children located in the South US were 16% less likely (aOR=0.84; 95% CI=0.76, 0.92) to have off-label diagnostic AAP use, compared to children located in the Northeast. The type of provider that a child received their AAP prescription from was not a significant predictor of off-label use by diagnosis in the adjusted model, when compared to prescriptions written by a specialty provider. After adjusting for other covariates, patients with a documented Mood or Anxiety Disorder were 95% less likely to receive an AAP medication for an off-label diagnosis (aOR=0.05; 95% CI=0.048, 0.053) compared to patient with psychotic, other or no mental health diagnosis present. Also, patients with a documented DAB or developmental disorder were 97% less likely to receive an AAP medication for an off-label diagnosis (aOR=0.03; 95% CI=0.03, 0.04) compared to patients with psychotic, other or no mental health diagnosis present after adjusting for other covariates.

Discussion
The proportion of children receiving AAP medication therapy in a large private payer was small (<1%) but still meaningful. Previous studies that included all children available for AAP prescribing, not only ones with documented mental health disorders, found similar rates of AAP medication therapy (<1% for children ages 2 to 17). 11 This low percentage is still meaningful, because it represents thousands (N=51,699) of children over the six year study period that are exposed to medications that have documented metabolic and cardiac long-term effects in adult patients. 21,22 This study included all pediatric patients available in the study dataset during the study period for the denominator, because all of these patients were at risk for receiving AAP therapy for any reason, off-label or on-label. an increase in the annual rate of paid claims over the study period, compared to older children (ages [13][14][15][16][17]. This may indicate that over the last six years provider began utilizing AAP medications in a young patient population as familiarity with the medication class grows. Geographic location of patient showed an association with the rate of AAP paid claims over the study period. These minor differences could be due to local treatment practices and clinical preferences. Mental health diagnosis associated with the paid claim for an AAP demonstrated a significant association with the rate of AAP paid claims over the study period. Patients with Developmental Disorders and Psychotic Disorders did not have a significantly different rate of paid claims for AAP medications over the study period, when compared to patients with no mental health diagnosis. This is finding is surprising, because previous trend studies have shown that use of atypical antipsychotics for Developmental Disorders was increasing overall and represented the highest rate of utilization compared to other mental health disorders. Our study population had a much lower percentage of patients with Developmental Disorder (4%) compared to previous literature (53%). 4 All clinical categories of mental health diagnoses were compared to the absence of mental health diagnosis in the medical visit around the paid claim. In our study, 35% of paid claims did not have an associated mental health diagnosis. Providers may withhold the documentation of mental health diagnosis due to potential stigma that could follow a pediatric patient through to adulthood. 11,23 Some antipsychotics could be utilized for treatment for other conditions (insomnia, agitation) that do not meet clinical criteria as a mental health disorder. 11,24 Finally, provider specialty or category did not demonstrate an association with the rate of paid AAP claims over the study period. Our study found that whether a patient is seen by a primary care provider or a mental health specialist, the rate of AAP prescribing is comparable.
Off-label prescribing can describe the use of medication therapy for indications that are not officially approved by the Food and Drug Administration. Off-label use also includes using medications for unapproved age groups and at unapproved dosing levels for certain populations. This study defined off-label diagnostic use as prescribing of an AAP medication with no documented mental health disorder or an unapproved mental health disorder. Future studies should explore dosing levels of AAP paid claims and differentiated age groups to examine all types of off-label prescribing. Off-label use of atypical antipsychotic agents in pediatric patients is heavily debated. 17,25 Many AAP medications have limited or no official FDA indication in children due to lack of research evidence in pediatric patients. Our study found that off-label diagnostic prescribing of AAP medications occurred in 62% (95% CI=62%, 63%) of all paid claims. This means that providers and patients were using AAP medications for other mental health diagnoses that have not been formally studied and approved by the FDA. Our study found that off-label prescribing of atypical antipsychotics is common in the pediatric population.

Limitations
Our study assumed that a paid claim for an AAP represents therapy adhered to by the patient. This assumes that the patient is exposed to a given medication because the paid claim was processed and therefore the patient adhered to the regimen. This could overestimate the actual exposure to AAP medication therapy because patients may have been prescribed the AAP medication, but never actually consume the prescription. For the purposes of this study, only prescribing trends were evaluated, and outcomes based on patient exposure were not examined. Future research that explores outcomes related to exposure of AAP could perform patient surveys or pill count methods to confirm the exposure to AAP medication therapy.
A sizable percentage (35%) of paid claims for AAP medications was not associated with a mental health diagnosis of interest. All patients included in the original cohort had a mental health diagnosis of interest to warrant inclusion in the cohort. With so many patients missing a mental health diagnosis at associated medical visit, the rate of other categories of mental health disorder might be underrepresented.
Many patients could have a diagnosis in one of the categories, but it is not documented and recorded in the "missing" category. This can underestimate the true rate of the mental health diagnostic categories that are used for off-label analysis. The lack of a mental health diagnosis associated with an AAP paid claim constituted off-label prescribing for the purposes of this study. This could be overestimating the rate of off-label diagnostic use of AAP medications in this study because provider could have simply failed to properly document the reason for AAP use and this undocumented reason could align with an approved FDA indication. Providers could justify this to protect a patient from the bias or stigma of mental health disease by not documenting a mental health diagnosis at medical visits.

Conclusions
Overall, the proportion of the pediatric population in a large privately-insured cohort receiving AAP medication therapy was small 0.28% (2015). with no pediatric indication at all were still found to be used in the study population (6.7%). Patients ages 2 to 5 years old were at an increased risk for using atypical antipsychotic medications for off-label diagnoses. Female patients were at increased risk for using atypical antipsychotic medications for off-label diagnoses. With these new insights, providers should consider more stringent use of atypical antipsychotic agents based on diagnosis until further safety studies are available specific to pediatric patients.

Characteristic Pediatric Patients with an AAP Paid Claim (N=40,750)
Age, y (mean, SD) 12    can determine if these agents should be considered for increased use in clinical practice for in pediatric patients to reduce the risk of readmission for mental health treatment.

Introduction
Several randomized, controlled trials have demonstrated that atypical antipsychotic medications, such as risperidone, olanzapine, aripiprazole and quetiapine, produce fewer adverse effects and offer better psychotic symptom relief in a short course than other agents in pediatric patients with mental health disorders. [1][2][3][4][5][6] However, there is limited information available about the comparative effectiveness of these medications in clinical practice settings, specifically in pediatric patients. 7 A major indication of drug effectiveness in clinical practice is relapse. In regards to mental health disorders, this relapse is characterized by worsening symptoms or changes in behavior that become harmful to the patient and/or society. 8 Time to readmission for inpatient mental health treatment is a commonly used measure for assessing relapse and effectiveness of mental health therapies. 8,9 Available follow-up studies in adults indicate that up to 50% of patients with schizophrenia and other psychotic disorders are readmitted within one year post discharge. 10,11 This high rate of readmission is particularly concerning because a higher rate of relapse is associated with worse long-term prognosis in adult mental health patients. 11 . Poor adherence to antipsychotic therapy has been shown to increase risk of relapse and hospitalization with a related increase in related healthcare resource utilization and costs. 12 16 These patients had an average length of stay of 11.1 days and average cost per admission was $7,500. 16,17 Patients who experienced a recent relapse (within previous 6 months) were found to have four times higher costs compared to patients without a recent mental health relapse. 17 This study focused primarily on adult patients and only included patients diagnosed with schizophrenia 12,13,16,17 A 2014 report analyzing admissions for mental health treatment in pediatric patients estimated the cost of hospital visits (inpatient and emergency department) to be $11.6 million from 2006 to 2011, based on HCUP data. 18 In 2014, 10% of all hospitalizations in children over the age of 3 years were for a primary mental health diagnosis. 19 Previous research followed adult schizophrenic patients for two years and found statistically significant differences between atypical antipsychotic agents in regards to risk of increased readmission rates. 9,20 To the best of our knowledge, no study has yet examined a direct comparison of oral atypical antipsychotic agents in privately-insured pediatric patients with mental health conditions to delay hospital readmission for mental health treatment.
This study focused on pediatric patients who utilized oral atypical antipsychotic therapy after an initial admission for mental health treatment.
Readmission for mental health treatment was evaluated to determine the efficacy of using specific atypical antipsychotics in pediatric mental health patients. Many randomized controlled trials and post-marketing trials demonstrated the efficacy of individual oral agents in the reduction in readmission in adults patients, compared to placebo or first generation antipsychotics. [21][22][23] Furthermore, clinical providers often extrapolate the demonstrated benefit of these agents in adults to pediatric patients with limited direct evaluation among children. 24 This study evaluated the effectiveness of specific oral AAP agents in delaying readmission in pediatric patients.

Study Design
The study was a retrospective cohort study utilizing the administrative dataset The inpatient file contains up to five diagnoses associated with an admission or encounter available for evaluation. Pharmacy claims data included medication information such as days' supply, quantity, prescribing physician and cost data. This dataset represents approximately 36 million covered patients across the United States.
The index date was the date of the first hospital admission for a mental health diagnosis during the study period. A look-back period of 90 days from index admission date was examined to ensure no prior admission for mental health treatment was present. Patients were followed for up to one year from the discharge date of the index hospitalization. According to studies evaluating inpatient mental health treatment in adults, the highest risk of readmission is in the first-year post-discharge, so this same follow-up period was chosen.

Inclusion Criteria
The study included all patients with an inpatient admission for a mental health

Exclusion Criteria
Patients will be excluded from the study if they were older than 17 years at index hospitalization and no paid claim within 14 days for atypical antipsychotic medication therapy. 9, 24 The 14-day post discharge date window was used to identify a paid claim as representing a discharge prescription from the index admission. This window was defined based on clinical practice parameters from the American Academy of Child and Adolescent Psychiatry that recommend follow-up appointments post hospitalization occur within 7-14 days to provide continuity of care across levels of mental health treatment. 26 Patients who were hospitalized for mental health treatment during the study period in the recent months preceding the index admission were excluded. A look back period of 90-days was examined for any recent admissions for mental health treatment. Patients were included in study cohort if their index admission was for a mental health diagnosis. A 90-day look back period from index admission was performed to examine prior exposure to AAP therapy and patients were classified as having no prior exposure, exposure to same AAP as discharge agent or exposure to different AAP as discharge medication. Figure 1 describes the study cohort with relevant exclusion or inclusion criteria.

Exposures and Outcomes
The exposure of interest was the use of atypical antipsychotic therapy at time of hospital discharge. 20 Exposure to specific AAP agents was evaluated and each agent was compared to risperidone. Risperidone was chosen as the reference agent because it was the first atypical antipsychotic to be awarded an Food and Drug Administration (FDA) indication for use in pediatric patients. The primary outcome evaluated was time from index hospital discharge to readmission for any mental-health related diagnosis (Appendix A).
Covariates considered sufficient to adjust for confounding included the following at baseline: age, gender, admission diagnosis, length of hospital stay, Charlson Comorbidity Index (CCI) and AAP exposure (same AAP as discharge, different AAP as discharge or no AAP therapy) prior to index hospitalization.
Admission diagnosis was categorized in to groups listed in Appendix A and compared to diagnosis codes in the "other" category. The other category was used as the reference group because it contains mental health diagnoses that AAP agents do not have a FDA approved indication and should have the lowest exposure risk to AAPs since no official FDA indication is present. The CCI score was used to evaluate the severity of illness among patients at their index hosptialization. 27 The CCI score

Statistical Analysis
We reported descriptive statistics to characterize each outcome group of interest (readmission for mental health treatment or no readmission). Baseline characteristics were also examined by discharge AAP (exposure group) and presented in and (Discharge AAP, P Value=0.14)]. The proportional hazard assumption for the covariate age group was not satisfied (P value = 0.02); therefore, the model was stratified by age group to allow for separate baseline hazards for each age group. 30 All pairwise interactions between covariates were not statistically significant in a single contrast (P value=0.56). These values indicated that none of the interactions of covariates were significant, so interaction terms were not included in the final model.
Collinearity between independent variables was tested using Variance Inflation Factors (VIF) test and no significant collinearity was found; thus, no adjustments for collinearity were made in the final model. Covariates associated with the outcome with P value less than 0.20 in the univariate analysis were included in the final adjusted model. 28 Gender (P value<0.001), prior AAP mediation exposure (P value <0.001), mental health diagnosis (P value=0.11) and discharge AAP medication (P value=0.14) demonstrated a significant effect on risk of readmission during the study period and all of these covariates were included in the adjusted model. CCI score was non-significant and was not included in the adjusted analysis. The length of stay (LOS) was not found to have a significant effect on risk of increased readmission and LOS was not included in the adjusted model. We used Efron's method to handle tied event times. 29,31 Cumulative incidence curves were generated using inverse probability weights to adjust for the baseline covariates. All statistical tests were twosided and performed at the 0.05 significance level. Analyses were performed using SAS Enterprise Guide Version 7.1 (Cary, North Carolina, USA).

Results
During the study period 2010 to 2015, 3,215 pediatric patients were admitted with a mental health diagnosis documented as the reason for admission. After applying inclusion criteria of receiving an AAP medication upon discharge (within 14-day window), 3,084 patients had a qualifying index admission for mental health treatment during the study period. Of those subjects, 313 (10%) children had a readmission within one year of the index admission discharge date for a mental health diagnosis or readmission for mental health treatment. The study sample is presented in Figure 1.

Study Cohort disposition and characteristics
For the entire cohort, the mean age of the study cohort was 14.  The unadjusted cumulative incidence of readmission is presented in Figure 2 and the inverse probability weighted cumulative incidence curves are presented in

Discussion
In the adjusted model, patients exposed to quetiapine and ziprasidone demonstrated a lower risk (aHR=0.55; 95% CI=0.37, 0.81; aHR=0.55; 95% CI=0.29, 1.0, respectively) of readmission, compared to risperidone. As represented in Figure   3, choice of discharge atypical antipsychotic does have a significant association with the risk of being readmitted for mental health treatment within the follow-up period in pediatric patients after adjusting for baseline covariates. Pediatric patients receiving quetiapine or ziprasidone also displayed a lower cumulative incidence of readmission ( Figure 3) over study period compared to patients receiving risperidone at discharge, after adjusting for baseline covariates Risperidone is one of the most frequently prescribed in the pediatric population for mental health treatment. This analysis suggests that patients might be at a lower risk of relapse when treated with quetiapine and ziprasidone and alternative AAP therapy may be more effective than risperidone.
A longer follow-up period is needed to compare effectiveness of atypical antipsychotics. 8  Our study focused on pediatric patients and many of the disease states evaluated using the CCI are chronic in nature and more prevalent as age progresses, so a majority of this pediatric study population (92%) demonstrated a CCI score of zero.
This was expected since the components of the CCI score are primarily chronic illnesses and these conditions are usually present in higher frequencies as a population ages. Most of the disease states analyzed in the CCI are chronic in nature (diabetes complications, congestive heart failure, etc.) and do not occur frequently in children.
No specific comorbidity index is available and sufficiently validated for use in pediatric patients, though there is forthcoming work for a pediatric-specific index. 34 Disease states that are more prevalent in children, such as asthma, childhood leukemia or autism, might be present in this cohort. However, the CCI index does not identify these diagnoses and they are not factored into the overall score that is intended to represent health status. Therefore, some underlying confounding by indication might be present if the patients that experience a readmission are sicker at baseline, but the disease severity is not fully captured by the CCI score. This study did not expressly evaluate cost of admission or treatment, but length of hospitalization was included as a covariate. Overall length of hospitalization can represent higher costs for the admission and poorer long-term clinical outcomes for mental health treatment, so overall length of stay (LOS) was examined as a covariate. 20  The association between discharge AAP agent and risk of readmission was somewhat attenuated within each gender.

Limitations
This study only evaluated AAP medication therapy received upon discharge from a mental health hospital admission. The permanence of this therapy or switches in treatment was not evaluated. This study only evaluated the exposure to an agent at the time of discharge and other therapies within the follow up time were not evaluated.
This could lead to exposure misclassification. Therapy switching and therapy permanence (PDC) between discharge and readmission should be analyzed to determine if certain oral AAP therapies are more effective. This study examined the difference in one-year hazard of readmission after prescribed an atypical antipsychotic agent at discharge. The "other" category trended toward a lower risk of readmission, but the use of these newer agents was low, and the determination of efficacy warrants further study. Once these new agents are utilized in clinical practice, future studies can evaluate evidence in administrative claims databases and determine if these agents are effective at lowering the risk of readmission. Some unmeasured confounding might be present for variables that we were unable to capture or did not examine in this study. Combination therapy with multiple AAPs or compliance with counseling or behavioral therapy has been documented to improve clinical outcomes and prevent relapse. 15,17 This study focused on analyzing the impact of discharge medication therapy with AAPs on risk of readmission, so switching therapy or combination therapy was not evaluated at this time. Mental health treatment often includes counseling services and other behavioral therapy interventions. This study analyzed the impact of medication therapy interventions on readmission outcomes specifically and did not explore the impact of other treatment modalities. Mental health treatment is often multifaceted and patient success is dependent on many treatment modalities.
Therapy services and group counseling provide support to the patient and play a vital role, along with medications, to treatment success. These treatment options and combinations of therapy with medications were not examined in this study but should be included in future research for their impact on mental health treatment success.

Conclusions
Patients receiving quetiapine and ziprasidone had a lower risk of readmission, compared to risperidone when used at discharge in pediatric patients.