A recent sharp increase in the number of youths diagnosed as having bipolar disorder has focused attention on community practice patterns (1,2,3,4). Concern exists that a diagnosis of bipolar disorder may be excessively used by health care professionals who treat children and adolescents (5,6,7). A diagnosis of bipolar disorder may lead to the prescription of antipsychotic medications, and increasing evidence of adverse metabolic risks associated with these medications in youths (8,9), especially in combination with mood stabilizers (10), has only heightened public concern. Yet surprisingly little is known about the circumstances under which clinicians diagnose young people as having bipolar disorder and its implications for medication management. A better understanding of these issues would help provide a context to assess prevailing concerns over recent growth in the diagnosis of bipolar disorder among young people in community settings.

In carefully characterized clinical samples, youths with bipolar disorder commonly have a complex constellation of psychiatric symptoms. Young people, especially children (11,12), diagnosed as having bipolar disorder frequently also receive diagnoses of disruptive behavior disorders. Anxiety disorders (13), other mood disorders (12,13,14,15), and psychotic symptoms (16) are also frequently present. Patients treated for bipolar disorder commonly receive complex pharmacological regimens (3,17). Temporal relationships between comorbid clinical diagnoses and psychotropic medication prescriptions, however, remain poorly described in the community treatment of young people who have been diagnosed as having bipolar disorder.

Treatment guidelines for youths with bipolar disorder recommend that pharmacological treatment continue for at least one to two years after symptom remission (18). This recommendation is supported by the high relapse rates among youths with narrowly defined mania (11) and bipolar spectrum disorders (16). Yet the duration of actual treatment episodes for juvenile bipolar disorder in community practice is not known.

In this study we examined service patterns surrounding new clinical diagnoses of bipolar disorder among young people. Pharmacological treatments and clinical diagnoses given during the year before and the year after a new diagnosis of bipolar disorder are described, and variations among patients are characterized according to the frequency of bipolar diagnoses in the year after the index diagnosis. These analyses were guided by four hypotheses. First, a clinical diagnosis of bipolar disorder will be more common among adolescents than among children. Second, a diagnosis of bipolar disorder will be associated with a prior diagnosis of depressive disorder among adolescents and a prior diagnosis of disruptive behavior disorder among children. Third, youths who receive more frequent diagnoses of bipolar disorder after the index diagnosis will have had more intensive mental health service use during the year before the index diagnosis. Finally, a clinical diagnosis of bipolar disorder will be followed by increased use of mood stabilizers and antipsychotic medications, decreased use of antidepressants and stimulants or atomoxetine, and a decrease in diagnoses of depressive disorder and disruptive behavior disorder.

Data source

Service and pharmacy claims were examined from the MarketScan Research Databases (2003–2006). They include information from over 27 million privately insured individuals and their family members from over 150 employers (19). Deidentified data were used for this study and have been determined to be exempt from human subjects review by the institutional review boards of New York State Psychiatric Institute and Rutgers University.

Patient selection

Selection of patients with a new diagnosis of bipolar disorder occurred in several stages. Patients with a claim for a new diagnosis of bipolar disorder (ICD-9-CM: 296.0 or 296.4–296.8) between April 1, 2004, and March 31, 2005, were first selected. Selected patients had not received any listed diagnosis of bipolar disorder during the year before the index visit for bipolar disorder (prediagnosis period) and were continuously eligible for services for at least one year before and after the new diagnosis of bipolar disorder. The sample was limited to patients who were aged 17 years and younger.

A comparison group without a diagnosis of bipolar disorder was then selected from the same population during the same period; the same selection criteria were used for eligibility and age. These patients had no claims for bipolar disorder during the year before and after the date of their first service claim within the recruitment period.

Rates and predictors of a new diagnosis of bipolar disorder

For the analysis of rates and predictors of a new diagnosis of bipolar disorder, patients in the study were characterized by age (younger than seven years, seven to 12 years, and 13 to 17 years), sex, and treatment for mental disorders (ICD-9-CM: 290–319) during the year before the index diagnosis. Claims for other mental disorders were categorized as being for disruptive behavior disorder (312, 312.9, 313.81, and 314,), pervasive developmental disorder and psychotic disorder (295, 297, 298, and 299), depressive disorder (293.83, 296.2, 296.3, 296.90, 300.4, 311, and 301.13), anxiety disorder (293.84, 309.81, 300.0, 300.2, 300.3, 308.3, and 309.21), adjustment disorder (309, except 309.21 and 309.81), or other mental disorder (290–319, not classified above). These groups were not mutually exclusive. Among patients with a diagnosis of disruptive behavior disorder, a subgroup was defined for those with at least one diagnosis of attention-deficit hyperactivity disorder (ADHD) (ICD-9-CM: 314) but no other diagnoses of disruptive behavior disorders.

Patients were also classified by filled prescriptions for stimulants or atomoxetine, antidepressants, antipsychotics, anxiolytics or hypnotics, or mood stabilizers. Mood stabilizers included lithium, carbamazepine, divalproex sodium, valproic acid, valproate sodium, gabapentin, lamotrigine, oxcarbazepine, and topiramate in the absence of a seizure disorder (ICD-9-CM: 345).

Rates of bipolar disorder diagnosis per 100 patients from the combined study and comparison groups with surrounding 99% confidence intervals were determined overall and stratified by sex, age group, presence of a comorbid mental disorder, and psychotropic medication prescribed in the year before the index diagnosis. One logistic regression model was fit to predict a new diagnosis of bipolar disorder, controlling for all covariates. Separate chi square analyses compared the proportion of males and females and the proportion of children (0 to 12 years) and adolescents (13 to 17 years) treated for each mental disorder and psychotropic medication class in the year before the index diagnosis.

Background characteristics of bipolar diagnosis groups

Three subgroups were defined by number of insurance claims for bipolar disorder starting with the index diagnosis. The incidental diagnosis group included patients having only the index claim, the sporadic diagnosis group had one to three additional claims, and the consistent diagnosis group had four or more additional bipolar claims.

Separate variables indicated whether patients received any inpatient or emergency treatment for mental disorders (primary or first listed ICD-9-CM: 290–319) or psychotherapy visits (CPT: 90804–90829, 90841–90847, 90849, 90853, 90855, 90857, 90875, and 90876) during the year before the index diagnosis.

The three subgroups (incidental, sporadic, and consistent) were compared with respect to age, sex, mental disorder diagnosis, and psychotropic prescriptions during the year before the new diagnosis of bipolar disorder. In these analyses, one or more prescription or diagnosis claims during this period were used to classify medication use and clinical diagnoses, respectively.

Mental health services before and after the new diagnosis

Filled psychotropic prescriptions and claims for mental disorders were compared across the incidental, sporadic, and consistent diagnosis groups. Using a criterion of two filled prescriptions claims during each one-year period to focus on substantial medication treatment episodes, each case was classified as no medication use in either year, use in the year before the index diagnosis only, use in the year after the index diagnosis only, or use in both years. Similar classifications were used for claims for mental disorders during the pre- and postdiagnosis periods. Chi square analyses were used to measure associations involving categorical dependent variables across bipolar diagnosis groups. Given the large samples, alpha was set at .01 (two-tailed).

Rates of a new diagnosis of bipolar disorder

As shown in Table 1, the annual rate of new clinical bipolar diagnosis was .23% (2,907 of 1,274,726 youths). Males and females had similar rates, although females had significantly greater odds of a new diagnosis of bipolar disorder in the multivariate model. Rates of a new diagnosis of bipolar disorder increased with patient age, and among children younger than 13 years, a new diagnosis was significantly more common among boys (.13%, 99% confidence interval [CI]=.11%–.14%) than among girls (.08%, CI=.07%–.09%). Among adolescents, however, a new diagnosis of bipolar disorder was significantly more common among girls (.50%, CI=.46%–.54%) than among boys (.39%, CI=.36%–.42%) (data not shown).

After the analysis controlled for all of the covariates, patients diagnosed as having a depressive disorder or treated with a mood stabilizer in the year before the index diagnosis were more likely than those not so diagnosed or treated to receive a new diagnosis of bipolar disorder. Other significant, although weaker, predictors included a diagnosis of disruptive behavior, anxiety, adjustment, and "other mental" disorders and prescriptions of antipsychotics, antidepressants, and anxiolytics or hypnotics (Table 1). Nearly 2.0% of youths treated for ADHD received a new diagnosis of bipolar disorder. By contrast, nearly one-half (44.8%) of youths treated with lithium received a new diagnosis of bipolar disorder during that period.

Treatment of a new bipolar disorder, by patient sex and age

Among youths with a new diagnosis of bipolar disorder, males were significantly more likely than females to have been treated for disruptive behavior and pervasive developmental or psychotic disorders and to have filled prescriptions for stimulants or atomoxetine, antipsychotics, or mood stabilizers in the year before the index diagnosis. They were less likely to be treated for anxiety, adjustment, depressive, or "other" mental disorders and were less frequently filled prescriptions for antidepressants during this period (Table 2).

Children were significantly more likely than adolescents to have been treated for a disruptive behavior disorder, including ADHD, and to have received stimulants or atomoxetine and antipsychotic medications in the year before the index diagnosis, but they were less likely to have been treated for a depressive disorder or "other" mental disorders (Table 2). They were also less likely than adolescents to be treated with antidepressants or anxiolytics or hypnotics during this period. The two age groups were approximately equally likely to be treated with mood stabilizers.

Bipolar diagnosis subgroups

Among youths with a new diagnosis of bipolar disorder, 28.8% received only the index bipolar diagnosis (incidental diagnosis group); 28.7% received the diagnosis two to four times (sporadic diagnosis group); and 42.5% received the diagnosis five or more times (consistent diagnosis group). Compared with the sporadic bipolar diagnosis group, the consistent diagnosis group included slightly, although significantly, more females (Table 3). The three diagnosis groups did not significantly differ with respect to mean±SD number of mental disorder visits for diagnoses other than bipolar disorder during the year after the index visit (mean: incidental, 10.7±12.4; sporadic, 10.1±11.2; consistent, 10.6±11.5).

During the prediagnosis year, approximately one-half (52.4%) of adolescents had been diagnosed as having a depressive disorder and a majority had filled a prescription for a psychotropic medication, most commonly an antidepressant (Table 2). During this period, most children with a new diagnosis of bipolar disorder had been diagnosed as having a disruptive behavior disorder and most had filled a prescription for psychotropic medication, usually a stimulant (Table 2).

Mental disorder diagnoses after a new diagnosis of bipolar disorder

Compared with the incidental diagnosis group, the consistent diagnosis group was more likely to have two or more claims for a diagnosis of pervasive developmental or psychotic disorder in the year after the index diagnosis. The consistent diagnosis group was also more likely than the incidental diagnosis group to have two or more claims for depressive disorder during the prediagnosis year (Table 4).

Psychotropic medication use after a new diagnosis of bipolar disorder

Using a criterion of prescriptions on two or more occasions, we found that youths in the consistent diagnosis group were significantly more likely than those in the other two groups to start psychotropic medications in the year after the new diagnosis of bipolar disorder and less likely than those in the incidental diagnosis group to stop psychotropic medications that they had previously received (Table 5). Starting antipsychotic medications and mood stabilizers after the new diagnosis of bipolar disorder was most common in the consistent diagnosis group, followed by the sporadic diagnosis group and then the incidental diagnosis group. Patients in the consistent diagnosis group were also significantly more likely than those in the incidental diagnosis group to start taking antidepressants, stimulants or atomoxetine, and anxiolytics or hypnotics during the year after the new diagnosis of bipolar disorder (Table 5).

During the course of one year, an estimated .23% of youths in this large privately insured population received a new clinical diagnosis of bipolar disorder. Although we are unable to address how the diagnosis of bipolar disorder in community practice compares with a rigorous application of DSM-IV criteria, most young people in our study who were given a new diagnosis of bipolar disorder had already been diagnosed as having other mental disorders, usually depressive or disruptive behavior disorders, and they were already being treated with psychotropic medications. In the year after the new diagnosis of bipolar disorder, more than half had three or fewer additional claims for bipolar disorder. This service pattern suggests that a new diagnosis of bipolar disorder tends to be given to young people with clinically recognized mental disorders rather than de novo and that the diagnostic label often does not persist, presumably as new clinical developments unfold. The observation that claims for other psychiatric disorders tend to precede claims for a new diagnosis of bipolar disorder may reflect the findings from prospective epidemiological research that several psychiatric disorders, including major depressive, conduct, and oppositional defiant disorders, commonly antedate mania onset (20). More generally, problems in child social, behavioral, and emotional development are often reported before full criteria are met for bipolar disorder (21,22).

Consistent with our hypotheses, the risk of receiving a new diagnosis of bipolar disorder increased with patient age. Also in keeping with our expectations, a clinical diagnosis of disruptive behavior disorder often preceded a diagnosis of bipolar disorder among children. Also, among children, boys were more likely than girls to have a diagnosis of bipolar disorder. Depressive diagnoses often preceded a diagnosis of bipolar disorder among adolescents, and among adolescents, females were more likely than males to be diagnosed as having a depressive disorder before receiving a new diagnosis of bipolar disorder. This pattern of clinical diagnosis may reflect previously described patterns in the age-related treated prevalence of disruptive behavior and mood disorders (23) through childhood and adolescence (24) along with the usual gender distribution in behavior and mood disorders.

Most young people filled prescriptions for psychotropic medications during the year before a new diagnosis of bipolar disorder. Antipsychotics, which are commonly used to treat bipolar disorder among youths (3), were often prescribed to young people during the year before the index diagnosis, especially to children and males of both age groups. Nearly one-third of the youths and proportionately more males than females who were newly diagnosed as having bipolar disorder were already receiving mood stabilizers. These prescription patterns indicate that a diagnosis of bipolar disorder is often given to young people who have been previously identified with and treated for serious mental health problems.

We had anticipated that during the year before the index bipolar diagnosis, patients who had more frequent (consistent) diagnoses of bipolar disorder would have experienced more intensive mental health service use, higher rates of other psychiatric disorder diagnoses, and more complex medication patterns, compared with patients with less frequent insurance claims for bipolar disorder. The groups in fact had generally similar service patterns during the prediagnosis year. Without more detailed information, including symptom patterns, psychiatric history, family history, and other clinical considerations, it was not possible to predict frequency of visits for the treatment of bipolar disorder among youths with a new diagnosis of bipolar disorder. In one prospective study of well-characterized youths with bipolar disorder, greater initial bipolar symptom severity and rapid cycling as well as older patient age and female gender predicted higher levels of care (25).

Although it is difficult to understand why a young person has one claim as opposed to five or more claims in community practice, it is possible to address the treatment consequences. First, approximately one-third of patients with an incidental diagnosis of bipolar disorder (only the claim for the index diagnosis) did not receive two or more prescriptions of any one class of psychotropic medications in the year after the diagnosis, so they were generally not medicated. This suggests that whatever was occurring to prompt the diagnosis was likely transient for some children or that perhaps appropriate medication prescriptions were not filled. By contrast, nearly 90% of youths who received consistent bipolar disorder diagnoses (four or more claims after the index diagnosis) received some psychotropic medication treatment in the year after the new clinical bipolar diagnosis. Moreover, prescription patterns were moderately responsive to new bipolar diagnoses. Use of mood stabilizers and antipsychotics, although used with some frequency before the index diagnosis, increased significantly after the index diagnosis, especially in the group with a consistent diagnosis.

We had hypothesized that a new diagnosis of bipolar disorder would be used by clinicians to explain patient symptoms and would lead to a decrease in other psychiatric diagnoses and cessation of treatments used to address the previously diagnosed disorders. Specifically, we anticipated that many young people diagnosed as having disruptive behavior disorder or depressive disorder before receiving a new diagnosis of bipolar disorder would not continue to receive these diagnoses after the new diagnosis. However, a substantial proportion of youths continued to receive diagnoses of depressive and disruptive behavior disorder after the new diagnosis of bipolar disorder. Similarly, most youths treated with antidepressants or stimulants or atomoxetine continued to receive these medications after the new diagnosis of bipolar disorder. This suggests that for many patients, a diagnosis of bipolar disorder is added to the diagnostic list rather than being used as an alternative explanation for the youth's behaviors, and similarly, treatments are added rather than substituted.

We found that around 2% of youths treated for ADHD during the prediagnosis year went on to receive a new diagnosis of bipolar disorder. In one clinical study, which was limited to children with ADHD and moderate to severe impairment, poorer baseline function predicted a change in diagnosis from ADHD to bipolar disorder (26). The disproportionate number of boys treated for ADHD in the year before a new diagnosis of bipolar disorder that we found in our sample raises the possibility that some community clinicians may use a diagnosis of bipolar disorder among boys with particularly severe ADHD.

This study has several limitations. As noted, there is no way to validate the accuracy of diagnosis in this claims database. No information is available concerning onset of psychiatric symptoms or psychiatric disorders. The administrative data also do not measure functional outcomes. For these reasons, it is not possible to draw causal inferences from associations of clinical diagnoses and medications with new onset of bipolar disorder. Associations between prior medication use and a subsequent new diagnosis of bipolar disorder should not be interpreted as evidence that these drugs sensitize patients to developing bipolar disorder. Finally, whether similar service use patterns would be observed among publicly insured or uninsured patients also remains unknown.

Most youths from a large privately insured population with new diagnosis of bipolar disorder received treatment for other mental disorders in the year before the new diagnosis. Among children, symptoms such as impulsivity, hyperactivity, irritability, aggression, and school failure are common to both juvenile bipolar disorder and disruptive behavior disorders. Among adolescents, depression also occurs with behaviors that raise questions about whether the condition is a manic or mixed-episode depression or depression with a comorbid condition. Perhaps as a result of diagnostic complexities, clinical diagnoses of bipolar disorder are often given in a tentative manner and do not persist as new symptom patterns emerge or resolve. Administrative requirements to enter diagnoses after each visit, even when the diagnosis is unclear, may further influence clinical diagnosis patterns. Some physicians may be addressing their patients' symptoms with different medications and justifying their use with a diagnosis, whereas others may make a diagnosis based on treatment response. On the basis of medical claims, however, it was not possible to predict which youths would go on to receive a consistent course of bipolar treatment from those whose diagnosis proved transient.

This work was supported by award U18-HS016097 from the Agency for Healthcare Research and Quality (Center for Education and Research on Mental Health Therapeutics).

Dr. Olfson has received research funding from Eli Lilly and Company and AstraZeneca Pharmaceuticals. Dr. Olfson has also served as a consultant for Pfizer, Janssen, and AstraZeneca Pharmaceuticals. Dr. Carlson has received research funding from and served as a consultant for Otsuka America Pharmaceutical and Bristol-Myers Squibb and has received research funding from GlaxoSmithKline. The other authors report no competing interests.