Jipan Xie, MD, PhD

To evaluate the treatment patterns among commercially insured adults in the United States with attention-deficit/hyperactivity disorder (ADHD) who received long-acting (LA) combination therapy (CT) or monotherapy for ADHD. 

Adults with at least 1 ADHD diagnosis and at least 1 LA ADHD medication were identified from the MarketScan claims database (April 1, 2009, to March 31, 2014). The index date was randomly selected among LA medication initiation dates (index treatment). CT was identified if a different ADHD medication was filled within 30 days of the index date and the 2 medications overlapped by 30 days or more; otherwise, the treatment was considered monotherapy. Adherence was measured using proportion of days covered (PDC) during the 1 year post index date and was defined as a PDC of 0.8 or greater. Persistence was defined as time to discontinuation (TTD) (ie, ≥30-day supply gap). Adherence and persistence were compared between CT and monotherapy using multivariable logistic and Cox models, respectively, adjusting for baseline characteristics. 

Of 225,600 eligible patients, 7.3% received LA CT and 92.7% received LA monotherapy (mean age, 29 vs 31 years, respectively). Patients receiving LA CT had significantly lower adherence than those receiving LA monotherapy (mean PDC, 0.33 vs 0.41; adherence rate, 7% vs 16%, respectively; adjusted odds ratio, 0.38; 

<.001). They also demonstrated significantly lower persistence than patients receiving LA monotherapy (median TTD, 59 vs 79 days, respectively; adjusted hazard ratio, 1.32; 

Among US adults with ADHD treated with LA medications, LA CT was associated with significantly lower adherence and persistence compared with LA monotherapy. 

Am J Manag Care. 2018;24(Spec Issue No. 8):SP329-SP337

Attention-deficit/hyperactivity disorder (ADHD) is a chronic disorder, and treatment adherence is important for providing optimal symptom control. Some adults with ADHD on long-acting (LA) medications experience inadequate symptom control using a single medication and augment their ADHD therapy.

Our study highlights that poor adherence and persistence were pronounced issues among adult patients with ADHD receiving LA therapy overall, but those receiving combination therapy versus monotherapy had a greater risk of nonadherence and nonpersistence.

These findings suggest an unmet need for longer-acting treatments in adult patients with ADHD, which may improve adherence and ultimately prevent the long-term societal and economic consequences of undertreated and untreated ADHD.

Medication adherence is the extent to which patients follow a medication regimen as prescribed by their healthcare provider. It is an important part of the optimal management of health conditions, disorders, and diseases.

 Prior studies have demonstrated that adherence is influenced by several factors, including dose frequency.

 Issues with adherence are more prevalent and impactful among those with chronic conditions that require ongoing management compared with acute conditions.

 In particular, poor adherence is prominent among patients withpsychiatric conditions,

 including attention-deficit/hyperactivity disorder (ADHD),

which is a lifespan disorder characterized by symptoms of persistent inattention, hyperactivity, and impulsivity.

 Although ADHD is primarily diagnosed in children and adolescents, approximately 50% of patients diagnosed in childhood experience symptoms that persist into adulthood.

 The prevalence of ADHD was estimated in 2006 at up to 5% among adults in the United States.

ADHD is associated with a high burden of illness among affected adults and was estimated to account for 3.5% of all adult mental health visits from 1996 to 2003.

 If left untreated or undertreated, ADHD can cause significant personal, social, and economic burdens, which may impair patients’ quality of life during adulthood.

ADHD symptoms can often be effectively treated with pharmacotherapy. The 2 main classes of ADHD medications are stimulants and nonstimulants.

 Stimulant medications, such as methylphenidate and amphetamine (AMPH), are commonly used as the first-line treatment for adults with ADHD,

 which generally complies with the recommendations for treating ADHD in childhood and adolescence. Nonstimulant medications include atomoxetine (ATX) or off-label use of clonidine hydrochloride extended release (ER), guanfacine ER, or antihypertensive agents such as clonidine instant release (IR) and guanfacine IR.

 Depending on the average duration of action, ADHD medications are classified as long-acting (LA; up to 12-14 hours of activity) or short-acting (SA; up to 4-6 hours of activity).

The duration of action affects the dosing frequency of ADHD medications and thus may affect adherence and persistence (ie, whether patients follow prescribed treatment regimens and remain on treatment, respectively).

 As other chronic diseases do, ADHD requires long-term pharmacotherapy to control symptoms. Poor medication adherence and persistence are common challenges in ADHD treatment and can negatively impact the optimal management of ADHD and increase healthcare costs.

For example, a literature review by Gajria et al of 91 studies (1990-2013) reporting ADHD medication adherence and persistence among adults or children with ADHD found adherence to be low in both groups (medication possession ratio <0.7 in both groups).

 Adverse effects (AEs), lack of symptom control, inconvenient dosing, and social stigma associated with ADHD medications were reported to be common reasons for treatment discontinuation in that review.

Previous study findings have suggested that patients with ADHD have better adherence and persistence if they use LA versus SA stimulants.

 A recent study of ADHD medication use in European countries reported an increase in the use of LA stimulants to treat ADHD in adults.

 However, if available LA medications cannot provide sufficient symptom control throughout the day, combination therapy (LA medications in combination with other ADHD medications) may be used.

 Results from a US study using retrospective claims data among adult patients with ADHD between 2007 and 2008 showed that 8.3% of patients augmented their LA medications.

However, no studies to date have assessed how adherence and persistence with LA ADHD medications is impacted by using combination therapy compared with monotherapy. To address this gap in the literature, the present study compared the real-world treatment patterns (medication adherence and persistence) of US adults with ADHD who received either LA combination therapy or monotherapy.

Commercial administrative claims data from the Truven Health MarketScan and Medicare Supplemental databases (April 1, 2009, to March 31, 2014) were used in this study. The data represent the medical claims of insured employees and their dependents from more than 40 national employers, as well as Medicare-eligible retirees with employer-provided Medicare Supplemental plans. These databases contain eligibility information (ie, enrollment dates), medical service claims (provider and institutional), and pharmacy claims from approximately 100 insurance providers and third-party administrators in the United States, representing approximately 50 million covered lives. Data were deidentified and complied with the patient confidentiality requirements of the Health Insurance Portability and Accountability Act and the most recent amendment to the Declaration of Helsinki. Because the data were anonymized, no institutional review was required.

Patients were eligible if they had at least 1 diagnosis for ADHD (

International Classification of Diseases, Ninth Revision, Clinical Modification

 codes 314.0 or 314.9) in the databases and initiated atleast 1 LA ADHD medication within 30 days prior to or any time after the first ADHD diagnosis. The list of ADHD medications is provided in the 

). For each patient, continuous treatment episodes were constructed for each LA medication initiated. Discontinuation of a treatment episode was defined as no prescription fill of a medication within 30 days from the last day of supply of the previous fill. Each initiation date of a unique treatment episode was defined as a potential index date. Patients were further required to be 18 years or older on any potential index date and continuously enrolled in a health plan during the 6 months before (baseline period) and 12 months after the potential index date (study period). The final index date was randomly selected from among all potential index dates for each patient. The treatment prescribed on the index date was defined as the first index treatment.

Patients were classified into either the LA combination therapy cohort (LA CT cohort) or LA monotherapy cohort (LA mono cohort), as defined below. 

 Patients were classified into the LA CT cohort if they had (1) a prescription for a different LA or SA ADHD medication within 30 days of the index date and (2) 30 or more days of overlap in medication supply between the first index treatment and the additional treatment. The index treatments for patients in the LA CT cohort included the 2 treatments in combination. Drugs of the same molecule but of different brands were considered 2 different treatments. Patients who used more than 2 ADHD treatments in combination were not considered in the study.

 Patients not meeting the criteria for inclusion in the LA CT cohort were classified into the LA mono cohort. The first index treatment was the only index treatment for patients in the LA mono cohort.

Baseline patient characteristics were measured on the index date or during the 6-month baseline period. These included age, sex, US geographic region, insurance type, prior ADHD-related psychotherapies, and prior ADHD medication use at drug class level. Comorbidities included a composite assessment of physical comorbidities using the Deyo-Charlson Comorbidity Index (DCCI).

 In addition, the 5 most prevalent neurological and psychiatric comorbid conditions

 and the 5 most prevalent physical comorbidconditions based on the literature were reported.

The following outcomes assessing the treatment patterns of the 2 cohorts were measured during the 1-year study period.

 Adherence to the index treatment(s) was measured using the proportion of days covered (PDC; ie, the proportion of days with medication on hand). The PDC was calculated as the sum of days that are “covered” by the index treatment (for the LA mono cohort) or both agents in the combination therapy (for the LA CT cohort) during the study period, divided by the number of days in the study period. Patients were considered adherent if their PDC was 0.8 or greater and nonadherent if their PDC was less than 0.8, based on the threshold defined by Haynes.

Persistence was measured using treatment discontinuation, which was defined as a gap of 30 or more consecutive days between the last day of supply of a prescription fill for the index treatment and the next prescription fill date of the same drug or the end of continuous healthcare plan enrollment, whichever occurred first. For the LA CT cohort, the date of treatment discontinuation was the earlier discontinuation date of either agent.

A treatment switch was defined as initiation of a different ADHD medication from the index treatment(s) within 30 days before or after the discontinuation date. The new medication filled was the switched-to treatment. The date of a treatment switch was defined as the date of the first prescription fill of the switched-to treatment.

Patient baseline characteristics were compared between the 2 cohorts using Wilcoxon rank-sum tests for continuous variables and χ

 tests for categorical variables. Fisher’s exact tests were used if 25% or more of the cells had expected counts less than 5 for categorical variables.

For the comparison of treatment adherence, the mean PDC was summarized and compared between cohorts using univariate and multivariable linear regression models. Unadjusted and adjusted mean differences were reported. In addition, univariate and multivariable logistic regression models were constructed and unadjusted and adjusted odds ratios (ORs) of being adherent to treatment were reported. Treatment discontinuation and treatment switch were assessed using Kaplan-Meier analyses and compared between cohorts using log-rank tests. For the analysis of treatment discontinuation, patients who did not discontinue the index treatment(s) during the study period were censored at the end of the study period. For the analysis of time to treatment switch, patients were censored at 30 days after the discontinuation dates or the end of study period, whichever occurred earlier. In addition, Cox proportional hazards models were used to compare the hazards of treatment discontinuation and treatment switch between cohorts controlling for baseline covariates. Proportional hazards assumption was tested on the cohort variable. 

All multivariable models controlled for the following baseline characteristics: age, sex, US geographic region, insurance plan type,year of the index date, any prior ADHD-related psychotherapy, any prior ADHD medication use, DCCI score, and neurological, mental, and physical comorbid conditions that were significantly different (

 <.05) between the cohorts. All tests were 2-sided and the significance was assessed based on a 

 value of .05. All analyses were conducted using SAS version 9.3 (SAS Institute; Cary, North Carolina).

Of the 225,968 adults who fulfilled all of the sample selection criteria, 16,401 (7.3%) were classified into the LA CT cohort and 209,199 (92.6%) were classified into the LA mono cohort. A total of 368 (0.2%) patients using more than 2 ADHD treatments in combination were excluded from the analysis (

). Among the LA CT cohort, the majority of patients (55.9%) used dextroamphetamine and AMPH mixed salts (LA and SA) in combination (

). Among the LA mono cohort, about two-thirds of patients (65.8%) used dextroamphetamine and AMPH mixed salts. The 5 most prevalent neurological and psychiatric comorbid conditions were mood disorders, anxiety disorders, adjustment disorders, substance-related disorders, and other conditions that may be a focus of clinical attention; the 5 most prevalent physical comorbid conditions were hypertension (uncomplicated and complicated), chronic pulmonary disease, hypothyroidism, diabetes (uncomplicated), and obesity.

Patients in the LA CT cohort were younger than those in the LA mono cohort (mean age, 29.0 vs 31.0 years, respectively), and that cohort had a higher proportion of males than the LA mono cohort (51.0% vs 49.1%, respectively; both

<.001) (Table 1). Compared with the LA mono cohort, patients in the LA CT cohort had a lower baseline comorbidity burden (DCCI score 0.10 vs 0.12, respectively; 

 <.001) and higher rates of prior ADHD medication use (68.2% vs 54.5%) and prior ADHD-related psychotherapy (25.1% vs 17.1%; both

The PDC was significantly lower among patients in the LA CT cohort (mean [SD] = 0.33 [0.25]) compared with those in the LA mono cohort (mean [SD] = 0.41 [0.30]; adjusted mean difference, —0.08; P <.001) (

). In addition, patients in the LA CT cohort were less likely to be adherent to the index combination compared with those in the LA mono cohort (6.5% vs 16.2%, respectively; adjusted OR, 0.38; 

In the unadjusted analysis, the median time to treatment discontinuation was 59 days for the LA CT cohort and 79 days for the LA mono cohort (log-rank 

). The LA CT cohort had a significantly higher risk of treatment discontinuation than the LA mono cohort (adjusted hazard ratio [HR], 1.32 [95% CI, 1.30-1.33] over the entire study period; 1.49 [95% CI, 1.47-1.54] over the period from 2 months after the index date to the end of study period). However, patients in the LA CT cohort were less likely to switch to a new ADHD treatment compared with patients in the LA mono cohort (adjusted HR, 0.77 [95% CI, 0.74-0.81]) (

), although the risk of switching was similar between the 2 cohorts from 2 months after the index date to the end of study period. At the end of the study period, a higher proportion of patients in the LA CT cohort had discontinued the index treatment(s) compared with the LA mono cohort (12-month rate, 93.1% vs 83.1%, respectively), while a lower proportion switched treatments (12-month rate,12.7% vs 19.6%, respectively).

This real-world study found that approximately 7% of US adults with ADHD who received LA medications used combination therapy with another LA or SA agent. Patients using LA combination therapy had significantly lower rates of treatment adherence and persistence compared with those using LA monotherapy. Moreover, a relatively small proportion of patients on combination therapy switched to a new treatment, in contrast with the high proportion that discontinued treatment.

Previous research has primarily focused on comparing the treatment patterns of adults with ADHD by the class of LA medications, instead of comparison between combination therapy and monotherapy. The patients in the current study who were receiving combination therapy were younger and had higher rates of use of prior ADHD-related medications and psychotherapies compared with patients receiving monotherapy. Few studies have compared the characteristics of these patient populations, although Clemow et al

 similarly noted that adult patients with ADHD receiving combination ATX therapy had higher rates of prior use of ADHD medications compared with those receiving monotherapy (76% vs 28%, respectively). However, that study, unlike the present study, reported that the patients receiving monotherapy were younger than those who received combination therapy (mean age, 32 vs 36 years, respectively). Patients receiving LA combination therapy may represent a group with greater unmet needs. Improvements in our understanding of the treatment patterns in this specific group may help improve the overall treatment and outcomes in adult patients with ADHD. The proportion of patients using LA combination therapy identified in the current study (7%) is consistent with the literature. A previous claims-based analysis conducted by Hodgkins et al in 2012 reported an augmentation rate of 8.3% over 6 months among adults with ADHD.

In addition, the adherence and persistence results from our study are comparable with those of other studies using similar outcome definitions. In 3 studies by Setyawan et al, the reported mean PDC for LA monotherapy ranged from 0.30 to 0.49, which was similar to our estimates of 0.33 for LA combination therapy and 0.41 for LA monotherapy.

 Similarly, Setyawan et al reported 1-year discontinuation rates of 76% to 91% in those 3 studies, comparable with the current results (93% for LA combination therapy and 83% for LA monotherapy). Other claims-based studies used different definitions for adherence and persistence, which makes it difficult to compare them with the current study.

 For example, Hodgkins et al and Christensen et al defined persistence as the number of days that the patient remained on treatment until the date of the last filled prescription during a 1-year study period.

 Treatment adherence was measured during time on treatment. Those definitions may have led to higher observed rates of treatment persistence and adherence.

 An abundance of literature has discussed the issues of nonadherence and nonpersistence among patients with chronic psychiatric conditions. Certain ADHD symptoms, such as disorganization and difficulties with planning, may contribute to the poor treatment adherence and/or persistence rates.

 According to a 2014 systematic literature review, treatment ineffectiveness/suboptimal response and dosing inconvenience are the 2 most commonly reported reasons for discontinuation in ADHD studies of adults.

 In addition, the AEs of stimulant medications, which are dose dependent, may contribute to poor rates of adherence, especially in patients using combination therapy. Common AEs related to stimulants include insomnia, anorexia, headache, abdominal pain, nausea, decreased appetite, increases in heart rate and blood pressure, and changes in mood.

 Moreover, patients receiving LA combination therapy are likely to be nonresponders to their regimen,

 resulting in a higher rate of nonadherence compared with patients receiving LA monotherapy.

Although previous studies' results have demonstrated that, among patients with ADHD, use of LA medications is associated with higher treatment adherence compared with use of SA medications,

 the rates of treatment adherence and persistence with LA medications observed in the present study may be suboptimal. Additionally, although it is difficult to know how long patients should persist on combination therapy, this regimen is attempted for certain adult patients, and these patients were observed to have shorter persistence (by approximately 2 months) compared with patients who received LA monotherapy. Reasons for poorer persistence among adultsreceiving LA combination therapy may include inadequate symptomatic coverage, a need for temporary relief, or factors associated with pill burden (ie, requiring multiple daily doses or medications). Current LA medications provide a maximum of only 12 to 14 hours of coverage. Thus, patients in need of longer durations of coverage are required to augment their therapy. A single medication providing more than 12 to 14 hours of coverage may benefit these patients by improving their treatment adherence and persistence on monotherapy.

This study was subject to the common limitations inherent to retrospective observational studies using claims data. First, the treatment patterns defined in this study may not be completely accurate. For example, combination therapy was identified using prespecified algorithms in claims databases, which may not reflect physicians’ prescribed dosing regimens for how patients should take their medication. Although using a 30-day overlap to define combination therapy is consistent with the literature,

 it may misclassify some “switchers” as “combination therapy patients.” Second, adherence was measured based on days of supply information in claims, assuming patients take the medication as long as it is supplied. This may not reflect the actual drug-taking behavior of patients. Third, reasons underlying treatment patterns, including the use of combination therapy and treatment nonadherence, nonpersistence, and switching, were not available in the claims databases. Moreover, as the analysis was not stratified by specific medications, it is unknown whether poor adherence or persistence were more pronounced within any particular class of ADHD medication. Fourth, the current study estimated ADHD medication use among the commercially insured adult population in the United States. The present results apply to commercially insured US adults diagnosed with ADHD and who had ADHD medication usage. They may not be generalizable to the general US adult population with ADHD, such as those without commercial insurance coverage.

Articles

Adults with attention-deficit/hyperactivity disorder (ADHD) who received long-acting combination therapy had significantly lower adherence and persistence compared with those who received long-acting monotherapy.

Treatment Patterns Among Adults With ADHD Receiving Long-Acting Therapy

To compare treatment patterns, resource utilization, and costs to US third-party payers of stimulant-treated adolescent attentiondeficit/ hyperactivity disorder (ADHD) patients who switched to or augmented with atypical antipsychotics (AAPs; not FDA-indicated for ADHD) with those who switched to or augmented with nonantipsychotic medications.

Retrospective cohort study conducted using a US commercial medical/pharmacy claims database.

Adolescent patients with an ADHD diagnosis and ≥1 stimulant medication claim between January 2005 and December 2009 were identified. Patients were classified into the AAP or non-antipsychotic cohorts based on subsequent claims for AAPs or nonantipsychotic medications, respectively. Patients with psychiatric diagnoses for which AAPs are often prescribed were excluded. Patients were matched 1:1 from the AAP to the non-antipsychotic cohort using propensity score matching. Treatment patterns, resource utilization, and costs in the 12 months after AAP or non-antipsychotic initiation were compared using Cox models, Poisson regression, and Wilcoxon signed-rank tests, respectively.

After propensity score matching, a total of 849 adolescents were included in each of the matched cohorts. Patients in the AAP cohort had a significantly higher rate of medication augmentation (27.7% vs 15.5%; hazard ratio = 2.56; 95% confidence interval [CI], 1.90-3.46; 

 <.001) than patients in the non-antipsychotic cohort. The AAP cohort also had significantly higher incidences of inpatient admissions (0.13 vs 0.05; incidence rate ratio [IRR] = 2.45; 95% CI, 1.73-3.48; 

 <.001), emergency department visits (0.39 vs 0.31; IRR = 1.27; 95% CI, 1.08-1.49; 

 = .004), and outpatient visits (14.82 vs 13.19; IRR = 1.12; 95% CI, 1.10-1.15; 

 <.001), and incurred significantly higher mean annual medical ($3622 vs $3311; 

 <.001), and total healthcare ($7936 vs $6195;

Stimulant-treated adolescents with ADHD who switched to or augmented with AAPs had significantly greater drug augmentation, healthcare resource utilization, and costs compared with the non-antipsychotic cohort.

• This study was the first to examine the treatment patterns and economic outcomes associated with off-label atypical antipsychotic (AAP) use in the treatment of adolescent attention-deficit/hyperactivity disorder using real-world data.

• After controlling for potential confounding variables using a propensity score match, treatment with AAPs was associated with significantly greater subsequent drug augmentation; higher rates of inpatient, emergency department, and outpatient visits; and higher all-cause medical, prescription drug, total healthcare, and mental health—related costs, compared with treatment with non-antipsychotic medications.

Atypical antipsychotics (AAPs) are one of the most common and costly classes of prescription drugs, with annual expenditures exceeding $13 billion, representing nearly 5% of all drug expenditures in the US.

 AAPs are approved by the FDA for the treatment of schizophrenia, behavioral symptoms in autism, and mixed or manic bipolar disorder, and the benefits and risks of AAPs are well documented for these indications.

 However, AAP use for off-label indications has rapidly increased, and now accounts for the majority of AAP utilization.

 A recent study found that AAP use in children grew by 62% from 2002 to 2007.

 Due to the potential side effects of AAP use and limited clinical evidence regarding the efficacy and safety of such off-label uses, the utilization of atypical antipsychotics for off-label indications is controversial.

 Among studies of off-label AAP use, heightened attention has been paid to attention-deficit/hyperactivity disorder (ADHD), as almost a third of off-label AAP use is related to this condition.

ADHD can pose a significant barrier to personal development and cause substantial psychological difficulties for patients and their families if left untreated.

 There are many pharmacologic treatment options for ADHD, including stimulants and non-stimulants, which have well-established efficacy and safety profiles.

Conversely, the risks and benefits of AAP use in current clinical practice for ADHD are largely unknown.

 The few clinical studies that investigated AAP use in ADHD patients are confounded by patient comorbidities for which AAPs may be appropriate, and are therefore difficult to interpret.

 However, the pediatric population appears to be at higher risk than adults for AAPinduced adverse events including weight gain, elevation in prolactin levels, extrapyramidal symptoms, sedation, and cardiac events.

 Additionally, there is limited evidence regarding the real-world economic outcomes of AAP use for ADHD, and recent literature has called for further investigation of the health outcomes in pediatric populations.

 A companion study examined this effect in children aged 6 to 12 years with ADHD, finding that patients who utilized AAPs had higher rates of drug switching and augmentation, greater medical resource utilization, and higher total healthcare costs compared with patients who used nonantipsychotic therapies.

 However, no peer-reviewed publications have investigated the economic costs of AAP use specifically in adolescents with ADHD. There is evidence that both the rates of ADHD diagnosis and AAP prescription for ADHD can vary by age category, which indicates the need for specific attention to the adolescent subpopulation.

Therefore, the purpose of this study was to compare, from a US third-party payer perspective, treatment patterns, resource utilization, and healthcare costs of stimulant- treated adolescents with ADHD who switched to or augmented their stimulant treatment with AAPs, versus those who did the same with non-antipsychotic medications.

This retrospective cohort study was conducted utilizing data from the Truven Health MarketScan Commercial Claims and Encounters database. These data include commercial health insurance claims and enrollment data from large employers and health plans across the United States. Such plans provide private healthcare coverage for more than 45 million employees and their spouses and dependents. This administrative claims database includes a variety of fee-for-service, preferred provider organization, and capitated health plans.

Patients were required to have at least 1 medical claim associated with a primary diagnosis of ADHD (

Classification of Diseases, Ninth Revision, Clinical Modification 

] codes 314.00 and 314.01) and 1 pharmacy claim for a stimulant 

during the period of January 1, 2005, to December 31, 2009, to be included in the study sample. The date of the first stimulant claim during this period was defined as the initial stimulant date.

Patients with a pharmacy claim for an AAP (Appendix) after the initial stimulant date were selected into the AAP cohort. The date of the first AAP pharmacy claim after the initial stimulant date was defined as the index date, and the AAP filled was defined as the index drug. Patients with no pharmacy claims for AAPs after the initial stimulant date, but at least 1 claim for atomoxetine, guanfacine immediaterelease (IR), clonidine IR, or a stimulant of a different class than the initial stimulant were selected into the non-antipsychotic cohort. For patients who had pharmacy claims for more than 1 non-antipsychotic medication after the initial stimulant date (eg, having claims for both atomoxetine and clonidine IR after the initial stimulant), the index drug was randomly selected from the eligible non-antipsychotic medications. This random selection method was employed to match the treatment history of the AAP cohort (which may have one or more non-antipsychotic medications prior to the index date) in an unbiased manner.

Patients in both cohorts were required to be aged 13 to 17 years on the index date. Individuals were included in the study if they had at least 30 days of supply of a stimulant before the index date and at least 18 months (6 months pre- and 12 months post index date) of continuous eligibility for their health plan. Furthermore, patients were required to have at least one primary diagnosis of ADHD during this 18-month period. Patients were only included if they switched from the initial stimulant to the index drug or augmented a stimulant with the index drug. Patients were considered to have switched if they initiated the index drug within a period defined as beginning 30 days before and ending 30 days after the last day with stimulant supply. Patients were considered to have augmented with the index drug if they had at least 30 consecutive days of overlap between the stimulant and the index drug.

To increase the likelihood that patients’ pharmacy claims were for the treatment of ADHD and not for other comorbid conditions that are often treated with AAPs, patients were excluded if they had a medical claim associated with any of the following diagnoses during the 18-month study period: bipolar disorder, delusions/hallucinations, paranoia, psychosis, tics, or dementia. These conditions were identified by group of medical experts as indications for which AAPs are approved by the FDA or commonly prescribed.

To control for observable confounding factors, propensity score matching was used to match patients in the non-antipsychotic cohort 1-to-1 with patients in the AAP cohort. The propensity score was estimated using an unconditional logistic regression including patient characteristics during the 6-month pre-index period. In addition, patients were exactly matched on whether the patient switched to or augmented with the index drug.

All outcomes were measured during the 12-month period following the index date. Outcome categories included treatment patterns, healthcare utilization, and costs.

Treatment pattern outcomes included discontinuation, switching, and augmentation. Discontinuation was defined as a gap in the usage of the index therapy greater than 30 days. A switch was defined as the initiation of a new ADHD medication (either an AAP or a non-antipsychotic other than the index therapy and the current stimulant) within 30 days before or after the index therapy discontinuation date. Augmentation was defined as initiation of a new ADHD medication, in which the supply of the newly initiated medication had at least 30 days of supply overlap with the index therapy.

Healthcare utilization outcomes included 3 mutually exclusive categories: inpatient, emergency department (ED), and outpatient services. In each category, both all-cause utilization and mental health (MH)-related utilization (

: codes 290.xx-319.xx) were calculated. Healthcare costs included medical costs and prescription drug costs. Within each category, both all-cause and MH-related costs were evaluated. Cost analyses were conducted from a third-party payer perspective, where costs were defined as the total amount paid without including out-of-pocket costs to patients. All costs were inflated to 2010 US dollars using the medical component of the Consumer Price Index.

During the 6-month pre-index baseline period, baseline characteristics of patients in the AAP cohort were compared with those of the non-antipsychotic cohort before and after propensity score matching. Comparisons employed c

 tests and Wilcoxon rank-sum tests before matching, and McNemar’s tests and Wilcoxon signed-rank tests after matching.

Treatment patterns were estimated using the Kaplan- Meier (KM) survival estimator; the corresponding hazard ratios (HRs) were estimated using Cox proportional hazard models. Rates of all-cause and MH-related medical utilization were compared between patients in the 2 matched cohorts using McNemar’s test, and the corresponding incidence rate ratios (IRRs) were estimated using Poisson regression. Since cost outcomes are typically not normally distributed and highly skewed to the right, cost comparisons were conducted using Wilcoxon signed-rank tests.

All analyses were performed using SAS Version 9.2 (SAS Institute, Cary, North Carolina), and statistical significance was evaluated at the .05 level (2-sided).

RESULTS Between January 1, 2005, and December 31, 2009, a total of 263,772 adolescents with at least 1 primary diagnosis of ADHD were identified from the database (

). Of these, 156,291 (59.3%) filled at least 1 prescription for a stimulant medication. Among those with at least 1 stimulant prescription, 17,268 (11.0%) had at least 1 prescription for an AAP medication after the initial stimulant. Similarly, 51,333 (32.8%) had at least 1 prescription for a non-antipsychotic medication (and no fills for an AAP medication) after the initial stimulant date. Among the AAP cohort, 9511 (55.1%) patients did not have any of the psychiatric diagnoses for which AAPs are commonly used. A total of 1151 patients in the AAP cohort and 6528 patients in the non-antipsychotic cohort met all additional eligibility criteria and were included in the study. A total of 1698 patients were included after matching, with 849 patients (187 who switched to the index drug, and 662 who augmented with it) included in each cohort. Inspection of the matched pairs within the propensity score histogram demonstrated that there was sufficient numeric representation and appropriate overlap between the two cohorts across the full range of propensity scores.

compares baseline characteristics between the 2 cohorts before and after matching. Prior to matching, several baseline characteristics were significantly different between the 2 cohorts. After matching, the differences in baseline demographic characteristics, resource utilization, and costs (with the exception of drug costs and total costs per patient) were not significantly different, indicating balanced baseline characteristics between the 2 cohorts.

In the post matching AAP cohort, 42.4% used risperidone followed by aripiprazole (27.2%), quetiapine (22.0%), olanzapine (4.7%), ziprasidone (3.3%), and paliperidone (0.4%) as the respective index drug; this distribution was similar to that observed in the pre-matching AAP cohort. In the pre-matching non-antipsychotic cohort, 72.5% of all index drugs were stimulants. After matching, 89.3% of all index drugs were non-stimulants (atomoxetine, 52.2%; clonidine IR, 31.1%; guanfacine IR, 6.0%).

During the 12-month period, persistence to the index drug was low in both the AAP and non-antipsychotic cohorts, with the non-antipsychotic cohort being more likely to discontinue the index drug (12-month KM: 74.9% vs 81.2%; HR = 0.85; 95% CI, 0.76-0.94) (

). Overall, patients were more likely to augment the index drug in the AAP cohort compared with the nonantipsychotic cohort (12-month KM: 27.7% vs 15.5%; HR = 2.56; 95% CI, 1.90 3.46) (

). The rate of switching was not statistically different between the 2 cohorts (12-month KM: 11.6% vs 9.7%; HR = 1.40; 95% CI, 0.90-2.20) (

In the 12-month study period, a significantly higher proportion of patients in the AAP cohort had at least 1 inpatient visit compared with the non-antipsychotic cohort (8.6% vs 3.3%, 

). Patients in the AAP cohort experienced an average of 0.13 inpatient visits compared with 0.05 inpatient visits in the non-antipsychotic cohort (IRR = 2.45; 95% CI, 1.73-3.48). The majority of inpatient visits (84.3% in the AAP cohort and 65.9% in the non-antipsychotic cohort) were MH related. Patients in the AAP cohort experienced an average of 0.11 MH-related inpatient visits compared with 0.03 inpatient visits in the nonantipsychotic cohort (IRR = 3.14; 95% CI, 2.07-4.77).

The AAP cohort had a significantly higher proportion of patients having at least 1 ED visit—both all-cause (28.2% vs 19.9%, 

= .004)—compared with the non-antipsychotic cohort. The average number of all-cause ED events per patient was also significantly higher in the AAP cohort (0.39 vs 0.31, IRR = 1.27; 95% CI, 1.08-1.49). The average number of MH-related ED visits per patient was not statistically different between the 2 cohorts (0.095 vs 0.069, IRR = 1.37; 95% CI, 0.98-1.92).

Almost all patients had at least 1 outpatient visit (98.4% in both cohorts), while the number of patients with at least 1 MH-related outpatient visit was higher in the AAP cohort (94.7% vs 90.8%, 

= .002). Patients in the AAP cohort also had more outpatient visits during the study period compared with non-antipsychotic patients. (All-cause: 14.82 vs 13.19; IRR = 1.12; 95% CI, 1.10-1.15. MH-related: 9.34 vs 7.51; IRR = 1.24; 95% CI, 1.20-1.29.)

Overall, the AAP cohort had higher mean healthcare costs compared with the non-antipsychotic cohort for both all-cause ($7936 vs $6195, 

<.001) and MH-related services ($5739 vs $3216, 

). Mean drug costs were significantly higher in the AAP cohort for both all-cause ($4314 vs $2884, 

<.001) pharmacy claims. The total cost for the index drug was also higher in the AAP cohort ($1718 vs $632, 

Patients in the AAP cohort had higher all-cause and MH-related medical costs compared with the non-antipsychotic cohort (all-cause: $3622 vs $3311, 

<.001). Medical costs across all categories (inpatient, outpatient, and ED) were higher in the AAP cohort, with the differences in inpatient costs (all-cause: $853 vs $707, 

<.001) and ED costs (all-cause: $239 vs $159, 

= .010) being statistically different. The majority of the medical costs for both cohorts were associated with outpatient visits but only MH-related costs were significantly different (all-cause: $2530 vs $2444, 

To our knowledge, this is the first study to evaluate the association between AAP treatment and medical resource utilization and costs for adolescents with ADHD. Our study found that 55% of adolescent ADHD patients with at least 1 prescription for an AAP did not have any psychiatric diagnoses for which AAPs are commonly prescribed. This result is consistent with previous studies of prescribing patterns for ADHD, and highlights the fact that the high prevalence of AAP use in this population warrants a careful assessment of the clinical, medical, and economic impact of such utilization.

 This study showed that even after controlling for baseline differences between the 2 treatment cohorts, adolescents with ADHD using AAPs were more likely to have greater resource utilization and healthcare costs compared with those in the non-AAP cohort.

Although there are documented differences between children and adolescents in the developmental presentation and trajectory of ADHD, the study findings are consistent with a companion investigation that studied a sample of children (aged 6-12 years) with ADHD. These 2 studies demonstrate that AAPs prescribed for ADHD present a substantial incremental burden to US payers compared with non-antipsychotics.

 Much of the incremental burden was related to higher MH-related utilization and costs. Although matched to MH-related 

codes, the exact reasons for the increased resource utilization and higher medical costs among AAP users cannot be elucidated from the administrative claims data; it is possible that they may be related to inadequate ADHD symptom control, the known side effects of AAPs, or other unknown causes.

 Additional research is warranted to determine the specific causes of treatment pattern differences, higher utilization, and higher costs among this population of ADHD patients.

Previous studies have noted gender differences in the presentation and treatment of ADHD, with males being diagnosed at a rate approximately 3 times that of females.

 In the present study, significantly fewer females were treated with AAPs compared with non-antipsychotic treatments before matching (27.2% vs 32.7%, 

<.001) (Table 1). These differences were controlled for by including gender within the propensity score match. Future analyses may provide more insight into the role gender plays regarding off-label treatment with AAPs for ADHD.

In addition, this study found that a significant number of patients in the pre-match AAP cohort used the index drug to augment their existing drug therapy (81.0% vs 36.4% in the non-antipsychotic cohort, 

 <.001) (Table 1). Recent studies have outlined concerns that arise with such polypharmacy involving stimulants and AAPs, due to potential interactions that may affect dopamine regulation or induce metabolic syndrome and its sequelae.

The off-label use of treatments not supported by adequate clinical and economic evidence may pose a substantial economic burden to payers. The need for such evidence is especially pronounced in regard to ADHD, for which several FDA-approved therapies with demonstrated effectiveness exist, including treatments indicated as mono- and combination therapy. A third-party payer could initiate a drug utilization review for off-label treatment of ADHD, which could help managed care organizations better understand the reasons that AAPs are being prescribed, and potentially improve the quality of care for adolescents with ADHD.

There are inherent limitations to using administrative claims data to determine whether AAP prescriptions were used in the treatment of ADHD. Specifically, clinical disease severity measures are not available, and the precise intent of the treating physician cannot be inferred. To minimize these limitations, stringent selection criteria were imposed, and the common indications for the use of AAPs were excluded. However, AAPs may have been used for psychiatric disorders not recorded in the database, possibly due to stigma associated with psychiatric diagnoses or the lack of available codes. Conversely, the stringency of the comorbidities list could have excluded patients with comorbid psychiatric disorders who were prescribed the AAP as a treatment for ADHD (and not their comorbid disorder), which may cause these results to be less generalizable to the entire ADHD population.

 Additionally, it is possible that AAPs were being prescribed for other off-label indications such as depression, conduct disorders, or oppositional defiant disorder. However, these comorbidities were included in the propensity score matching, and there were no statistically significant differences between the 2 cohorts in the frequency of these diagnoses. While all relevant available variables were included in the propensity score model, potential unobserved confounders may exist that are related to the prescribing of the AAP but are not available in the data set. It is also possible that the decision to prescribe AAPs on-label versus off-label may have been influenced by formulary status or other policies.

Propensity score matching was used to create balanced and comparable cohorts. However, after the propensity score match, differences in drug costs and total costs at baseline remained. The results of the propensity score match are only generalizable to the population included in the match. The final match excluded 302 AAP patients who could not be matched with representative non-antipsychotic patients. These patients had higher rates of medical resource utilization and costs at baseline, thus the results of the current study may underestimate the association between off-label AAP use and medical resource utilization and costs.

Finally, because these analyses covered only the first 12 months following the index date, the results of this investigation cannot be extrapolated to provide insight on the longer-term economic implications of off-label AAP treatment of ADHD. Indeed, the results may understate the actual effect of such treatments, as many of the complications associated with AAPs are chronic and the entirety of their effects on costs and utilization may not be observed within the study period. Future research of the long-term health outcomes and economic implications of off-label AAP use is justified.

Compared with adolescent ADHD patients treated with stimulants who switched to or augmented with a non-antipsychotic medication, those who switched to or augmented with an AAP had significantly greater medical resource utilization, and higher total healthcare costs, and they were more likely to augment their treatment in the 12 months after initiating the AAP. These patients may present a greater economic burden to payers. Further outcomes research and/or drug utilization reviews may be necessary to fully evaluate clinical and economic outcomes in adolescents with ADHD who are receiving AAPs.

 Shire, Wayne, PA (VS, MHE); University of Texas Health Science Center at San Antonio, San Antonio, TX (SRP); Analysis Group Inc, Boston, MA (KAB, TMS, JX, EQW); and formerly at Shire, Wayne, PA (BR, PH, RSD).

This research was funded by Shire Development LLC, Wayne, PA.

 Drs Sikirica and Erder are employees of Shire and hold stock or stock options with Shire. Drs Hodgkins, Robertson and Dammerman were employees of Shire during the time of this study. Drs Betts, Xie, and Wu, and Mr Samuelson are employees of Analysis Group, Inc, which received funding from Shire Development LLC for this study. Dr Pliszka reports consulting work for Shire and has received grants from Shire and Ortho-McNeil-Janssen Pharmaceuticals, Inc. He has also provided expert testimony for Eli Lilly and Company

A retrospective study of the treatment patterns and economic outcomes associated with off-label atypical antipsychotic use in the treatment of adolescents with attention-deficit/hyperactivity disorder.

Impact of Atypical Antipsychotic Use Among Adolescents With Attention-Deficit/Hyperactivity Disorder

 A hub-and-spoke telestroke network is an effective way to extend quality emergency stroke care to remote hospitals and improve patient outcomes.

To evaluate the cost utility of a telestroke network in the management of acute ischemic stroke from the societal perspective.

 A lifetime Markov model was developed to compare the incremental costs and effectiveness of a telestroke network. One-year transition probabilities between the 3 health states based on the modified Rankin scale—minimal-to-no disability, moderate-to-severe disability, and death&mdash;were derived from literature. Costs included telemedicine setup and maintenance, initial and recurrent stroke treatment, rehabilitation, long-term care, and caregiver costs. Effectiveness was defined as quality-adjus ted life-years (QALYs). Model inputs were obtained from the literature supplemented by data from Georgia Health Sciences University and Mayo Clinic. The base case network included 1 hub and 7 spokes, and assumed no survival benefits from acute treatment in a network. One-way sensitivity analyses were conducted.

 Compared with no network, patients treated in a telestroke network incurred $1436 lower costs and gained 0.02 QALYs over a lifetime. Incremental costs decreased from $444 for the first year to —$1436 over a lifetime; incremental QALYs increased from 0.002 for the first year to 0.02 over a lifetime. Overall, results were robust in the 1-way sensitivity analyses. A telestroke network became less cost-effective with increasing spoke-to-hub transfer rates.

A telestroke network is cost savingand more effective compared with no network from the societal perspective in most modeled scenarios.

This health economic study demonstrated that compared with a rurally located patient receiving routine stroke care at a community hospital, a patient treated in a telestroke network incurred $1436 lower costs and gained 0.02 quality-adjusted life-years over a lifetime.

		The telestroke network was economically dominant over routine care. n The results serve to inform government organizations, insurers, healthcare institutions, practitioners, patients, and the general public that an up-front investment in telemedicine and stroke network personnel can be justified in our health system.

		Insurance plans should reimburse telestroke consultations in the same fashion as face-to-face clinical encounters.

Stroke is a leading cause of severe disability, and the fourthleading cause of death in the adult population in the United States.

Acute ischemic stroke (AIS) accounts for more than 80% of all strokes.

 Administration of intravenous (IV) thrombolysis within a 3-hour window of the onset of AIS reduces disability.

A recent clinical trial has demonstrated that IV thrombolysis administered between 3 and 4.5 hours after the onset of AIS could also significantly improve clinical outcome.

 Extending the 3-hour traditional   time window by 1.5 hours is likely to increase the demand for IV thrombolysis because more AIS patients would present to the hospitals in time for treatment.

Despite the effectiveness of IV thrombolysis, fewer than 5% of patients who suffer AIS in the United States receive IV thrombolysis.

 One of the major barriers to IV thrombolysis administration is that most hospitals do not have the resources required to enable timely diagnosis and treatment of AIS patients. Hospitals with access to qualified personnel (eg, 24/7 on-call vascular neurologists) are most likely to provide care due to the time limitation. However, it is not feasible for every hospital to meet such resource and personnel requirements, especially those in remote rural areas. To serve communities without a stroke center, telestroke referral networks have been set up to connect a stroke center (the hub hospital) with local hospitals (spoke hospitals) and thus extend the reach of the acute stroke care team of the hub hospital to spoke hospitals. Through this establishment, AIS patients can receive  more accurate diagnoses, more correct thrombolysis eligibility determination, and more appropriate emergency treatment, all resulting in potentially better prognoses. For instance, research studies revealed that telestroke results in correct diagnosis and thrombolysis decision making in 96% of instances.

A hub-and-spoke telestroke network is an effective way to extend quality emergency stroke care to remote hospitals and improve  patient outcomes. The costs associated with these networks, as well as the health-related outcomes, must be considered. We previously conducted a cost-effectiveness analysis (CEA) from the hospitals’ perspective.

 the current study aimed to assess the cost utility of a hub-and-spoke telestroke network compared with no network (absence of regional stroke system of care and telemedicine) in the management of AIS from the societal perspective. Two such studies were published and subsequently independently appraised and scored by the Tufts Medical Center Cost Effectiveness Analysis Registry.9,10 However, our study was the first to develop a cost decision model that included the possibility of endovascular therapy in addition to IV thrombolysis and the first to include an expanded array of cost inputs associated with dedicated network program managers and personnel, higher estimates of inpatient care, inter-hospital transfer, rehabilitation, long-term care, and caregiver costs, as well as a wider range of spoke-to-hub transfer rates, using 2 independent telestroke network data input sets.

A Markov model was developed to estimate the incremental costs and effectiveness with and without a telestroke network for the management of AIS over a lifetime. The model had 3 health states defined by the modified Rankin scale (mRS)

: (1) minimal-to-no disability (mRS 0-2); (2) moderate-to-severe disability (mRS 3-5); and (3) death (mRS 6). A hypothetical cohort of AIS patients with a mean age of 68 years were assumed to receive acute care in a telestroke network versus a no-network setting and to transition between these health states at the beginning of each cycle (1 year). The mean age of 68 years was obtained from a US epidemiology study of patients with first-time stroke.

 illustrates the decision and treatment process for AIS and subsequent transitions. Building on the previous model,

 the base case of this study modeled a network with 1 hub and 7 spokes, with a total of 1112 unique AIS patients presenting to emergency departments in the network hospitals per year. The number of spoke hospitals in the network was based on a recent survey of active telestroke networks by Silva and colleagues,

 and the average number of AIS patients for spoke hospitals was obtained from the Georgia Health Sciences University and the Mayo Clinic telestroke networks. The number of AIS patients for the hub hospital was assumed to be 400 , based on the typical size of a hub hospital. Data on network characteristics were obtained from the Georgia Health Sciences University and the Mayo Clinic telestroke networks (see 

available at www.ajmc.com); these data were described in a study based on the same network.

 The following assumptions were made in the model estimation:

		Acute ischemic stroke patients could only transition from a less severe to a more severe health state or remain in the same health state at each cycle.

		Stroke treatments between a telestroke network and no network differed only during the initial hospitalization for AIS, not after discharge from acute care.

		Incremental effectiveness associated with treatments in a telestroke network only resulted from IV thrombolysis or endovascular stroke therapy during the initial hospitalization for the first-time AIS.

		There was no difference in stroke-related mortality between patients with and without IV thrombolysis,and between patients with and without endovascular stroke therapy during hospitalization. We made these assumptions because clinical trials on IV thrombolysis did not show a significant difference in mortality between patients who received IV thrombolysis and patients who did not, though the former group had a numerically lower rate.

 In addition, there was also a lack of randomized controlled trial data comparing the efficacy of endovascular stroke therapy with no such therapy.

 Therefore, we assumed no mortality difference in mortality in the base case, which could be a conservative assumption.

A cost-utility analysis of a hub-and-spoke telestroke network showed that it was economically dominant over routine care.

Jipan Xie, MD, PhD

Articles

Treatment Patterns Among Adults With ADHD Receiving Long-Acting Therapy

Impact of Atypical Antipsychotic Use Among Adolescents With Attention-Deficit/Hyperactivity Disorder

Cost Utility of Hub-and-Spoke Telestroke Networks From Societal Perspective