Factors That Predict the Use of Psychotropics Among Children and Adolescents With PTSD: Evidence From Private Insurance Claims

Exposure to one or more major traumatic events, such as physical or sexual assault, accidents, emotional abuse, and natural disasters, may trigger the development of posttraumatic stress disorder (PTSD) among children and adolescents (1). According to DSM-5, PTSD symptoms include intrusive memories or re-experiencing of the traumatic event (nightmares and flashbacks), avoidance of reminders of the traumatic event, negative thoughts and feelings (memory problems and a distorted sense of blame), and hyperarousal (irritability, reckless behavior, difficulty concentrating, and sleep disturbance). Such broad symptomatology may adversely affect a child's age-appropriate development (2). According to point-prevalence estimates, PTSD affects between 3% and 15% of girls and between 1% and 6% of boys. The lifetime prevalence estimate is 8% for girls and 2.3% for boys (3). The latest estimates reported an overall pediatric PTSD prevalence of 15.9% (4,5).

Even though PTSD is becoming a more common pediatric diagnosis, most of the treatment evidence comes from adult trials. Indeed, the bulk of research exploring PTSD treatment options relies almost entirely on war veterans and adult disaster victims (6). This evidence suggests that psychotherapy has superior clinical effectiveness in treating PTSD compared with pharmacotherapy (7–9). For severe cases, the combination of psychotherapy and pharmacotherapy may work best (10,11). Pediatric PTSD treatment guidelines (including those from the American Academy of Child and Adolescent Psychiatry) agree on the appropriateness of psychotherapy, particularly trauma-focused cognitive-behavioral therapy (TF-CBT) as the first-line treatment (12). This recommendation was based upon multiple systematic reviews and meta-analyses that demonstrated large effect sizes (.8–.9) for psychotherapy (13–15). The favorable effects of psychotherapy for youths pertained across a wide range of traumatic exposures, including sexual and physical abuse, war, accidents, sickness, and natural disasters, and were shown to be durable for up to one year (8,13). Yet psychotherapy remains underused in primary care settings (14,15), where expertise in this treatment modality may be limited.

In contrast to psychotherapy, pharmacotherapy for pediatric PTSD lacks rigorous empirical support (16–18), and—as such—the Food and Drug Administration (FDA) has not granted approval for use of any medication among children and adolescents for this indication. The only randomized trials of a psychotropic medication (sertraline, an antidepressant) for the treatment of pediatric PTSD failed to demonstrate significant beneficial effects, either as monotherapy or as an adjunctive treatment to psychotherapy (19,20). Medications from other classes have only low-quality evidence of effectiveness for PTSD (8,13). Although these medications, including alpha-agonists (guanfacine [16,21] and clonidine [17,18]), antipsychotics (risperidone [22] and quetiapine [23]), mood stabilizers (carbamazepine [24] and divalproex [25]), an alpha-antagonist (prazosin [26]), and a beta blocker (propranolol [27,28]), have shown potential efficacy in case reports and uncontrolled trials, stronger evidence is needed to support a first-line treatment recommendation (29,30).

To date, little is known about the prevalence of psychotropic prescribing for PTSD among children and adolescents in the United States. If common, this practice would be concerning, given that it is associated with significant side effects and limited efficacy, hardly a favorable risk-benefit ratio (31,32). Similarly, little is known about the factors associated with the use of pharmacotherapy for PTSD among youths, including characteristics of patients and of practice settings, such as primary versus specialty care and geographic location (33).

For pharmaceuticals, evidence from adult veterans enrolled in Veterans Affairs clinics suggest that most (80%) veterans who receive pharmacotherapy for PTSD are treated in specialty mental health clinics, and only 13% are treated exclusively in primary care (34). Also, ruling out random noise, geographic variation in pharmacotherapy for PTSD among adults is substantial, pervasive, and persistent (35). Yet this important research agenda, identified more than 20 years ago, remains underexplored (31).

This study is one of the first that aims to determine the utilization of three classes of commonly prescribed psychotropic medications—antidepressants, benzodiazepines, and antipsychotics—in a national sample of children and adolescents with diagnosed PTSD and to identify the patient and setting factors that significantly predict the use of these medications by these youths.

Methods

Study Design

We conducted a retrospective cohort analysis with data from the MarketScan Commercial Claims and Encounter database for children and adolescents (ages six to 18) who received a new diagnosis of PTSD between January 1 and December 31, 2012. The Boston Medical Center Institutional Review Board considered this data source as de-identified, and the study was exempted from review.

Data Source

The Truven MarketScan claims database offers information on more than 200 million unique patients (including employees, spouses, and dependents) since 1995. It allows the creation of a nationally representative data sample of Americans with employer-provided health insurance. The commercial claims are collected from small and large employers and health plans. Data comprise service-level claims for inpatient and outpatient services and outpatient prescription drugs. All claims have been paid and adjudicated.

Study Population

The index date was defined as the date of the first PTSD-related outpatient medication claim during the index period (January 1, 2012, to December 31, 2012). The preindex period included the six months prior to the index date.

The inclusion criteria included age between six and 18 years at the index date; a new diagnosis of PTSD (ICD-9 code 309.81) at the index date, with no PTSD claims during the preindex period (36); at least two PTSD-related outpatient claims or one PTSD-related outpatient claim and one PTSD-related medication claim during the index period; and insurance coverage for at least 12 months after the index date and during the preindex period (18 months total). Individuals were excluded if they had a comorbidity that would legitimize the use of antidepressants, benzodiazepines, or antipsychotics (bipolar disorder [296.0x, 296.4x], schizophrenia [295.6x], depressive disorders [296.2x or 296.3x and 300.4 or 311], and mixed psychotic disorders [e.g., depressive-type psychosis]) or a PTSD diagnosis or a PTSD-related medication claim during the preindex period. This approach allowed us to capture new PTSD episodes.

Measurement and Outcomes

Patient characteristics such as sex, age, region, insurance arrangement, provider specialty, hospitalization history, patient complexity, and history of psychotherapy were included. Patients were divided into two groups by age: children ages six to 11 and adolescents ages 12 to 17. Four geographic regions of the Unites States were analyzed (Northeast, Midwest, South, and West). Insurance arrangements were categorized as follows: health maintenance organization (HMO) and point-of-service (POS) with capitation, noncapitated POS, high-deductible plan, noncapitated preferred provider organization (PPO), and comprehensive plan.

Provider specialties were divided in two groups: mental health providers (MHPs) (psychiatrists, psychiatric nurses, and psychologists) and primary care providers (PCPs) (internists, pediatricians, family physicians, general practitioners, and medical nurses). History of hospitalization was defined as having an inpatient claim during the 12 months prior to the index date. PTSD-related psychotherapy was identified as receiving psychotherapy within 30 days of the index date. We used this approach to capture the most common clinical practice (psychotherapy is provided after the hospitalization [32]) rather than searching for the evidence-based recommendation (psychotherapy is the first line of treatment). Having multiple mental conditions has been associated with poorer outcomes and higher mortality and morbidity rates (37,38). Hence, to assess complexity, we constructed a complexity index grouping four psychiatric comorbidities common with PTSD (anxiety, disruptive behavior, dissociative conditions, and suicidal ideation). To account for general medical comorbidities, we utilized the Elixhauser Index, developed for use with large administrative databases (39). The general medical index identifies 31 unweighted comorbidity indicators by scanning diagnosis-related groups and secondary ICD-9 codes (40). Medication data were extracted from the outpatient pharmacy claims by using the National Drug Code. Medication was categorized in three mutually exclusive groups (antidepressants, benzodiazepines, and antipsychotics).

Data Analysis

We calculated frequency of prescriptions in the three medication classes for all individuals with at least one prescription. For univariate analysis, we compared continuous data by using t tests and categorical variables by using chi-square tests. Logistic regression models were developed to predict two dependent variables: odds of receiving an antidepressant (versus no medication) and odds of receiving an antipsychotic (versus no medication). Each dependent variable was regressed on sex, age, region, insurance arrangement, provider specialty, hospitalization history, patient complexity, and history of psychotherapy. The weighted estimates were calculated based on a probabilistic nationally representative sample. All analyses were performed with person-level national weights constructed with the Medical Expenditure Panel Survey (MEPS), which provides estimates of the number of people with employer-sponsored private health insurance in the United States.

The person-level weights reflect the ratio of MEPS-based estimates in four categories: sex, age, region, and insurance arrangement. We used the SAS raking macro, developed by Izrael et al. (41), to adjust the Truven database marginal totals. Hence our results are representative of the privately insured population in the United States in terms of sex, age, region, and insurance arrangement. We inspected collinearity by the variance inflation factor criteria (<10). The models were evaluated by using c statistics for discrimination, Akaike information criterion and the Hosmer-Lemeshow test for goodness of fit, and bootstrapping to assess the internal validity of the coefficients. All statistical analysis was performed with SAS, version 9.3.

Results

A total of 13,818 pediatric patients with PTSD were identified. Among them, 7,726 met our inclusion and exclusion criteria. Overall, 5.9% of youths received pharmacotherapy, 59.0% received psychotherapy alone, 2.7% received combination treatment (psychotherapy and pharmacotherapy), and 35.2% received neither psychotherapy nor pharmacotherapy. Of those receiving pharmacotherapy, 71.3% received antidepressants, 21.6% received antipsychotics, and 7.1% received benzodiazepines (Table 1).

The baseline characteristics of youths who received pharmacotherapy were significantly different from those who did not. Youths in the pharmacotherapy group tended to be older, to be female, to live in the South, to reside in rural areas, to be insured under capitated arrangements, and to be treated by MHPs. They also tended to have been previously hospitalized and, on average, had a larger number of comorbidities (data available upon request).

Table 2 summarizes the demographic characteristics of youths who received a medication and the results of univariate assessment comparing patient characteristics by medication class. The mean±SD age of medication recipients was 14.0±2.8. Those who received benzodiazepines were older (mean age=15.2±1.8) than those who received antipsychotics (13.1±3.0) and antidepressants (14.1±2.7). Children received antipsychotics more often (27.6%) than antidepressants or benzodiazepines (17.7% and 6.3% respectively). Almost all use of benzodiazepines was by adolescents (93.8%). In terms of gender, females accounted for about three-fourths of benzodiazepine and antidepressant users, compared with 59.2% of users of antipsychotics. For insurance type, use of antidepressants was highest among individuals with PPO coverage (54.6%), whereas use of benzodiazepines was highest among people covered by an HMO (31.3%). Other predictors such as area of residence, region, use of psychotherapy, and provider type were evenly distributed across medication types.

Table 3 reports the adjusted independent predictors of receiving a prescription for antidepressants. First, provider specialty had the strongest association with antidepressant use (OR=3.13). For age, children under age 12 were less likely than adolescents to receive antidepressants (OR=.49). Geographically, providers in Northeastern and Western regions were less likely than those located in the South to prescribe antidepressants. For insurance arrangements, persons at PPOs and high-deductible plans were less likely to receive antidepressants compared with those under capitated HMO/POS plans. Last, the number of medical comorbidities, as measured by the Elixhauser severity index, was significantly higher for youths receiving antidepressants (OR=2.34).

Table 4 reports the adjusted independent predictors of receiving a prescription for antipsychotics. A history of hospitalization was the strongest predictor of antipsychotic use (OR=2.99). As was the case with antidepressants, the Northeast and West regions were less likely to use antipsychotics compared with the South. The Elixhauser severity index (more medical comorbidities) and the complexity index (more psychiatric comorbidities) were significantly higher among youths receiving antipsychotics. Last, the odds of receiving psychotherapy were lower for those receiving antipsychotics (OR=.49).

The discrimination and goodness of fit were good for both models. After 100 bootstrapping iterations, the direction and magnitude of the estimates did not vary significantly, which suggested good internal validity for both models (data available upon request).

Discussion

To our knowledge, this may be the first large population-based study to identify frequency and predictors of various treatments for children and adolescents with PTSD. We noted several important findings. First, less than 10% of youths were treated with pharmacotherapy. Second, two-fifths of youths with PTSD did not receive psychotherapy, and more than one-third did not receive any treatment, which highlights an important treatment gap. As mentioned earlier, there is strong and consistent empirical support for the safety and effectiveness of psychotherapy, particularly TF-CBT, for PTSD (9,12–15,19). As such, psychotherapy is considered the first-line treatment for PTSD. On average, clinicians followed guidelines for treating PTSD among youths, given that few patients with PTSD were treated with off-label medications and over 60% received some sort of psychotherapy.

Second, more than one of every 20 youths with PTSD received pharmacotherapy, predominantly antidepressants (three-quarters of the medicated youths) and antipsychotics (one-fifth). Even though the FDA has not approved the use of any medication for pediatric PTSD (9,16–19,22–35), PTSD experts suggest that pharmacotherapy might be considered for more severe or complex presentations (12), a recommendation predominantly based upon experience with adults. The tendency to prescribe medication for more complex cases was confirmed in our findings. Youths receiving pharmacotherapy had, on average, more medical and psychiatric comorbidities. The main predictor of antidepressant use was MHP specialty type and for antipsychotics, the second strongest factor was a previous hospitalization history. These findings support a relationship between PTSD severity and complexity and psychotropic medication use.

Most providers who prescribed psychotropic medications chose antidepressants, which may reflect the symptom overlap between clinical depression and PTSD (e.g., for both disorders, negative mood and loss of pleasure or interest in usual activities are included in the diagnostic criteria). Hence, to opt for an antidepressant may reflect symptom-guided medication selection (42). Another reason for this decision may derive from the lack of alternative medications with a similar record of safety and effectiveness among youths. This is particularly relevant for antipsychotics and benzodiazepines, both of which are associated with significant side effects (19,28,29) and, for benzodiazepines, a risk of dependency (43).

In terms of provider specialty, more than 83% of youths who received medication were seen by an MHP. This finding matches adult use patterns (34). Interestingly, this finding contravenes patterns of use for other mental disorders, such as depression, given that most youths who are prescribed antidepressant medication receive the prescription from a PCP (44). Consequently, PCPs may be more empowered to deal with certain mental illnesses compared with others. With the migration of mental health care to primary care settings and the escalating shortages of MHPs around the country (45), this finding stands out. It suggests that there is a need to promote training and support to deal with PTSD more frequently in primary care settings.

Although the role of PCPs has evolved to include recognizing and addressing trauma exposure in families, there has been only a modest shift in practice to proactively identify and treat trauma exposure (46). The American Academy of Pediatrics suggests that PCPs can play several key roles in this regard (43). First, PCPs can familiarize themselves with screening instruments designed to facilitate the identification of PTSD. Several of these instruments assess the psychological response to traumatic exposure (47–49), whereas others identify risk factors for the development of PTSD (50,51). Early identification is important, given that studies suggest that early psychological intervention may avert the development of full-blown PTSD (52).

Second, PCPs can provide self-management tools to patients and families to enhance their ability to adaptively cope with stressful or traumatic circumstances. Even very brief educational interventions that are feasible to deliver in the primary care setting have been demonstrated to reduce the development of PTSD among youths (53). Third, PCPs can become knowledgeable about trauma-focused resources in their community and track referrals of youths with clinically significant symptoms of PTSD to ensure completion. These and other simple primary care interventions have been shown to favorably affect both clinical practices and patient outcomes (54,55).

Finally, PCPs considering the use of psychotropic medications to treat PTSD among children and adolescents would be well advised to seek specialist consultation. Payer-blind child and adolescent psychiatry (CAP) consultation programs have been launched in 27 states and the District of Columbia to facilitate on-demand PCP access to telephone and in-person CAP consultation, with emerging evidence of the ability of these programs to foster safe and effective psychotropic prescribing practices by PCPs (www.nncpap.org).

Finally, geographic variation in use of pharmacotherapy for PTSD remained significant after multivariable adjustment. The southern region utilized antipsychotics and antidepressants more often than northeast and western regions of the country. This effect may be explained by the large rural and urban poverty rate gap. In the South, historically, this gap has been the largest, averaging 5.1 percentage points over the past two decades (56). Also, rural poverty is deepest in certain parts of the south. This resource constraint acts as a disincentive to practice in these settings, greatly limiting the number of specialized mental health providers capable of offering effective psychosocial services.

Several limitations should be noted. First, our findings may not generalize to other types of insurance coverage. Second, we used billing records that may have underidentified patients with PTSD by excluding those who had an outpatient visit in which a PTSD diagnosis was not recorded. In other words, our estimates may be biased toward the null hypothesis and are more likely conservative. Third, we may have overestimated the use of psychotropics by not excluding youths with anxiety and disruptive behavior disorders. Even though previous research showed these comorbidities are likely to coexist, in our sample less than 7% of individuals reported them (data available per request). Therefore, we included these two disorders and other psychiatric comorbidities from the complexity index and accounted for their effects in the multivariable analysis. Fourth, the database captures only the formal version of psychotherapy. Hence, we may underreport other less intensive versions (e.g., brief supportive therapy) not captured in the data.

Fifth, a fraction of our sample may not represent a new episode of PTSD, given that some youths may have had a previous diagnosis prior to our six-month clean period. However, given routine care practices, a six-month period should be adequate to uncover new episodes. Sixth, although we tried to control for case mix by using a severity and complexity index, the effects of unmeasured confounders on treatment decisions may be unknown. Also, because of the cross-sectional design, we could not identify temporal changes in medication use (such as switching or augmenting drugs) that are clinically relevant. For benzodiazepines, the small sample size did not permit the calculation of stable coefficients. Finally, we did not have information to assess whether providers monitored metabolic profiles when prescribing antipsychotics.

Conclusions

Psychotherapy was the treatment option used most often among children for a new episode of PTSD. Two-fifths of youths with a formal PTSD diagnosis did not receive psychotherapy. Even though less than 10% of youths used off-label pharmaceuticals, quality-of-care concerns remain valid. Further research is needed to understand the rationale behind the prescription of pharmacotherapy for PTSD among children and develop initiatives to expand psychotherapy access, especially in the primary care setting.