Clinical characteristics and resource utilization of patients with bipolar disorder who have frequent psychiatric interventions

Abstract

Objective:

To compare the demographics, clinical characteristics and resource utilization of patients with bipolar disorder who required frequent psychiatric interventions (FPIs) with those needing fewer interventions in the Duke Healthcare System database between 1999 and 2005.

Methods:

This retrospective analysis was conducted using electronic medical records of bipolar patients with FPIs, defined as having ≥4 clinically significant events (CSEs) in any 12-month period while in the Duke University Healthcare System. CSEs were composed of emergency room visits, inpatient hospitalizations, or a change in psychotropic medication due to psychiatric symptoms (score ≥4 on the Clinical Global Impressions–Severity scale). Data were compared between patients with and without FPIs.

Results:

Of 632 patients with bipolar disorder 52.5% were identified as having FPIs. These patients were younger and more often female and African American than those with fewer interventions (p < 0.01 for all). Patients with FPIs were generally prescribed more psychotropic and non-psychotropic medications, utilized more healthcare resources and experienced more psychiatric co-morbidities than those who did not require FPIs (p < 0.01 for all).

Limitations:

These results are from a single healthcare system and may not be generalizable to all patients with bipolar disorder. This analysis was retrospective and relied on availability of adequate information recording and coding of diagnoses by physicians.

Conclusions:

Patients with bipolar disorder who required FPIs were significantly different from those with fewer clinically defined interventions with respect to their demographic and clinical characteristics and prescribed medications.

Key words::

• Bipolar disorder
• Clinical management research information system
• Clinically significant events
• Frequent psychiatric interventions

Introduction

Bipolar disorder is a serious chronic mental illness characterized by lifelong periodic mood episodes separated by periods of relative wellness1, with a 12-month incidence in the United States estimated between 1.3% and 2.0%2–4. Patients with bipolar disorder also represent a significant economic burden, estimated at $7 billion in direct medical costs and $38 billion in indirect costs in the United States5. In addition, comparisons of resource utilization between patients with bipolar disorder and age- and gender-matched patients who do not have bipolar disorder found that those with bipolar disorder used three to four times more healthcare resources, incurred four times greater costs per patient, incurred higher drug costs and required more frequent psychiatric hospitalizations, outpatient visits and emergency room (ER) visits6–8.

In patients with bipolar disorder, the extent of healthcare resource use and overall treatment costs are expected to vary depending on the severity of illness. Indeed, those with a more insidious disease course may require frequent interventions. For purposes of this analysis, frequent psychiatric intervention (FPI) was defined as four or more interventions in the previous year to address mood symptoms and was used as a surrogate marker for severity.

Various outcome measures in bipolar patients whose illness reached this stage were analyzed using the Clinical Management Research Information System (CRIS), a large clinical database developed by the Duke University Medical Center Department of Psychiatry9. To assess differences in demographics, clinical characteristics and resources used, electronic medical records data were compared between patients whose conditions fit the FPI definition and those whose conditions did not.

Patients and methods

Study design and data source

This retrospective analysis evaluated data obtained from electronic medical records for all patients examined during visits to the Department of Psychiatry at Duke University Medical Center from 1999 to 2005. The catchment area is the Piedmont region of North Carolina, with 75% of the patients living within 100 miles of Duke University. Individual patient information (demographics, diagnoses, medications, hospitalizations, ER visits and clinical rating scales) is collected and stored in this repository in a longitudinal, anonymous manner that complies with the Health Insurance Portability and Accountability Act. The data tables from CRIS were retrospectively reviewed by the authors, enabling a large-scale, naturalistic examination of patients with and without FPIs.

Inclusion criteria

Eligible patients were in the system for 1 or more years, had a Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR), diagnosis for bipolar disorder (type I or II) and received treatment from a physician. FPI terminology was derived from the DSM-IV-TR definition of rapid-cycling bipolar disorder, characterized as four or more mood episodes a year10. Compared with patients with bipolar disorder who experienced fewer annual mood episodes, this patient population was generally considered more difficult to treat, with higher morbidity11,12.

For this analysis, patients with FPIs experienced four or more clinically significant events (CSEs) requiring intervention in any 12-month period recorded in CRIS. A CSE was defined as an ER visit related to bipolar disorder, inpatient psychiatric hospitalization or a change in psychotropic medication associated with psychiatric symptoms (i.e., score ≥4 on the Clinical Global Impressions–Severity [CGI-S] scale). All CSEs were at least 7 days apart, and hospitalizations defined as CSEs were for separate episodes rather than successive hospitalizations for a single episode.

Study measures and statistical analysis

Patient demographics were recorded at study entry into the database, and all psychiatric and medical diagnoses were recorded for each visit. In addition, psychiatric and substance abuse co-morbidities and healthcare resource use (ER visits, hospitalization, outpatient visits and psychiatric and non-psychiatric medications) were recorded. In analyses comparing resource use and costs, only those attributable to bipolar disorder were considered. Daily inpatient physician assessments were equated with an inpatient length of stay (i.e., hospital days). All pertinent patient data were entered in the database. Resource use and costs were annualized and presented on a per-patient per-year basis.

Suicidality was determined by documenting suicidal thoughts, intent or behaviors (e.g., overdose, self-injurious behavior) based on the Mental Status Examination and information from the patient’s psychiatric history and the physician’s clinical notes. CGI-S scores were rated by the treating physician at each visit, and baseline severity measures were determined at the time of the initial psychiatric visit.

Between-group differences for continuous variables were analyzed using an analysis of variance model; categorical variables were analyzed using the chi-square test. All descriptive statistics and frequency distributions were calculated using Statistical Analysis Software (SAS Institute, Inc, Cary, NC, USA).

After examination of the full complement of study variables, a final set of the most commonly referenced variables in the database was evaluated using multivariate logistic regression models to assess their relationship with frequent interventions. The importance of each covariate for predicting this probability was measured by the magnitude of the parameter coefficient (odds ratio). The following variables were used: age (≥40 vs. <40), African American race (yes vs. no), female gender (yes vs. no), depressive disorders (yes vs. no), anxiety disorders (yes vs. no), personality disorders (yes vs. no), relationship problems (yes vs. no), substance abuse (yes vs. no), baseline CGI-S severity (≥5 vs. <5) and suicidality (yes vs. no). Statistical significance of the multivariate models was evaluated using the chi-square test. No adjustments were made for multiplicity.

A published study by Stensland et al. (2007)13 provided 2004 unit costs of resources associated with bipolar disorder. Their analysis used data from the PharMetrics Integrated Outcomes Database of adjudicated medical and pharmaceutical claims for more than 6 million patients from six United States health plans. Costs of resources were inflated to 2008 unit costs using Bureau of Labor Statistics Medical Care Services14 and Prescription Drugs Inflation Rates15. The unit costs of resources are shown in Table 1. Costs for resource utilization and medications were calculated by multiplying the values for each group by the adjusted 2008 costs. No inferential statistics were computed for the cost variables.

Table 1.  2008 unit costs of psychiatric resources.

Annual per-patient costs	2004 unit cost from Stensland et al.13 (US$)	Inflated 2008 unit cost (US$)*
Direct medical costs
ER visits	512.50	614.95
Days of hospitalization	751.27	901.45
Outpatient visits	70.87	85.04
Medications
All antipsychotics	1439.53	1604.42
Antidepressants	618.18	688.99
Anticonvulsants	1045.28	1165.01
Lithium	115.00	128.17
Benzodiazepines	92.50	103.10

*Costs inflated to 2008 on the basis of United States Bureau of Labor Statistics consumer price index14,15.

Results

Demographics and clinical characteristics

A total of 632 patients met the inclusion criteria (Table 2). The majority of the population was female (60.4%) and Caucasian (68.4%), with a mean age of 40.5 years. Of the 632 patients with bipolar disorder, 332 (52.5%) met the criteria for FPIs. The median time between CSEs for these patients was 49 (range, 7–317) days. Patients differed in their demographics and clinical characteristics based on the number of interventions required (Table 2). Those who required FPIs were younger (39.1 vs. 42.2 years; p = 0.0015) and more likely to be female (66.3% vs. 54.0%; p = 0.0016) and African American (24.5% vs. 15.4%; p = 0.0045) than patients who did not. CGI-S ratings in patients with FPIs confirmed that they had more severe illness than those without FPIs upon entry into this health system (Figure 1), with a higher percentage rated as having moderate illness (44.3% vs. 29.5%; p = 0.0002) or severe illness (31.1% vs. 23.9%; p = 0.047). Moreover, a significantly higher proportion of patients who required FPIs were classified as suicidal at some point during their recorded treatment compared with those who did not (n = 208 [62.7%] vs. n = 90 [30.0%]; p < 0.0001) (Table 2).

Figure 1.  Baseline CGI-S ratings at first visit for patients who required FPIs vs. those who did not require FPIs. CGI-S, Clinical Global Impressions–Severity; FPIs, frequent psychiatric interventions. N = 610 (FPIs, n = 325; no FPIs, n = 285).

Table 2.  Comparison of baseline demographics, medication use and suicidality in patients with or without FPIs.

	Total patient population (n = 632)	Patients with FPIs
		Yes (n = 332)	No (n = 300)	p-value*
Age, mean (SD), years	40.5 (12.4)	39.1 (11.4)	42.2 (13.2)	0.0015
Gender, n (%)
Male	250 (39.6)	112 (33.7)	138 (46.0)	0.0016
Female	382 (60.4)	220 (66.3)	162 (54.0)
Race, n (%)†
Caucasian	431 (68.4)	218 (65.9)	213 (71.2)	0.0045‡
African American	127 (20.2)	81 (24.5)	46 (15.4)
Other	72 (11.4)	32 (9.7)	40 (13.4)
Suicidality, n (%)	298 (47.2)	208 (62.7)	90 (30.0)	<0.0001
Resource utilization, visits per patient
ER	1.4	2.2	0.5	0.0001
Inpatient days§	3.7	5.5	1.7	0.0001
Outpatient visits	18.1	24.4	11.0	0.0001
Psychotropic medications per visit, mean (SD)	2.7 (1.2)	2.9 (1.1)	2.3 (1.2)	<0.0001
Non-psychotropic medications per visit, mean (SD)	2.2 (2.3)	2.5 (2.5)	1.9 (2.0)	0.0025

ER, emergency room; FPIs, frequent psychiatric interventions; SD, standard deviation.

*p-value corresponds to patients who have FPIs versus those who do not have FPIs.

†N = 630 (FPIs, n = 331; no FPIs, n = 299).

‡p-value corresponds to comparison of African American patients and Caucasian patients, plus others.

§Daily inpatient assessments of a patient by a physician were equated with days of inpatient hospitalization.

Patients with FPIs also used more psychiatric resources than those without FPIs, including mean ER visits per patient (2.2 vs. 0.5 visits; p = 0.0001), mean hospitalization days (5.5 vs. 1.7 days; p = 0.0001) and mean outpatient visits per patient (24.4 vs. 11.0 visits; p = 0.0001) (Table 2). As expected, patients with FPIs were prescribed a significantly higher mean number of psychotropic (2.9 vs. 2.3; p < 0.0001) and non-psychotropic (2.5 vs. 1.9; p = 0.0025) medications per psychiatric visit. Specific medication classes prescribed were statistically different between patients with or without FPIs and are summarized in Table 3.

Table 3.  Concomitant medications with statistically significant use in patients with or without FPIs*.

Medication, n (%)	Patients with FPIs
	Yes (n = 332)	No (n = 300)	p-value
Antipsychotics
Typical	33 (9.9)	12 (4.0)	0.0037
Atypical	238 (71.7)	112 (37.3)	<0.0001
Antimania
Lithium	152 (45.8)	91 (30.3)	<0.0001
AEDs	188 (56.3)	136 (45.3)	0.0046
AEDs, second generation	169 (50.9)	84 (28.0)	<0.0001
Antidepressants
Atypicals	195 (58.7)	119 (39.7)	<0.0001
TCAs	33 (9.9)	12 (4.0)	0.0037
SNRIs	93 (28.0)	36 (12.0)	<0.0001
SSRIs	224 (67.5)	133 (44.3)	<0.0001
Anxiolytics
Benzodiazepines	220 (66.3)	129 (43.0)	<0.0001
Non-benzodiazepines	42 (12.7)	19 (6.3)	0.0072
Anti-sedation	15 (4.7)	5 (1.7)	0.0409
Extrapyramidal symptoms, Parkinson’s disease	28 (8.4)	10 (3.3)	0.0071
Amphetamines	55 (16.6)	27 (9.0)	0.0047
Selective GABA-receptor agonists	28 (8.4)	8 (2.7)	0.0018
Alcohol abuse	9 (2.7)	2 (0.7)	0.0497

AED, antiepileptic drug used as a mood stabilizer; FPIs, frequent psychiatric interventions; GABA, gamma-aminobutyric acid; SNRI, selective norepinephrine reuptake inhibitor; SSRI, selective serotonin reuptake inhibitor; TCA, tricyclic antidepressant.

*Concomitant medications that were commonly used in patients with and without FPIs but not statistically significant between the two populations were medications used for opioid addiction (5.4% vs. 3.0%; p = 0.132), tremor/akathisia (4.5% vs. 2.7%; p = 0.214) and medications for insomnia (benzodiazepines: 4.2% vs. 3.7%; p = 0.723; non-benzodiazepines: 4.8% vs. 3.0%; p = 0.241).

Co-morbidities

Patients with FPIs had a significantly higher percentage of psychiatric co-morbidities than those with fewer interventions (Table 4). The most common psychiatric co-morbidities that were more prevalent in patients with FPIs were anxiety disorders (40.4% vs. 25.7%; p < 0.0001), depressive disorders (38.6% vs. 23.7%; p < 0.0001) and personality disorders (31.9% vs. 10.0%; p < 0.0001). In addition, a significantly higher proportion of patients in this subgroup were diagnosed with a substance abuse or dependence disorder (39.8% vs. 28.0%; p = 0.0019). In patients needing FPIs compared with those needing fewer interventions, the most common substance abuse disorders were related to alcohol (24.4% vs. 17.7%; p = 0.0387) and cocaine (18.1% vs. 10.3%; p = 0.0057) (Table 4).

Table 4.  Psychiatric and substance abuse co-morbidities that were statistically significant in patients with and without FPIs*.

Co-morbidity, n (%)	Patients with FPIs
	Yes (n = 332)	No (n = 300)	p-value
Anxiety disorders	134 (40.4)	77 (25.7)	<0.0001
Depressive disorders	128 (38.6)	71 (23.7)	<0.0001
Mood, other†	38 (11.5)	11 (3.7)	0.0003
Schizophrenia	9 (2.7)	2 (0.7)	0.0497
Psychotic disorders, other‡	38 (11.5)	12 (4.0)	0.0005
Substance abuse diagnosis	132 (39.8)	84 (28.0)	0.0019
Alcohol use disorders	81 (24.4)	53 (17.7)	0.0387
Cocaine use disorders	60 (18.1)	31 (10.3)	0.0057
Nicotine use disorders	28 (8.4)	8 (2.7)	0.0018
Poly-substance use	13 (3.9)	1 (0.3)	0.0022
Personality disorders	106 (31.9)	30 (10.0)	<0.0001
Eating disorders	9 (2.7)	1 (0.3)	0.0168
Dissociative disorders	6 (1.8)	0 (0)	0.0193
Relational problems	52 (15.7)	12 (4.0)	<0.0001
Additional codes§	169 (50.9)	97 (32.3)	<0.0001
Additional conditions that may be a focus of clinical attention¶	27 (8.1)	13 (4.3)	0.0501

FPIs, frequent psychiatric interventions.

*Psychiatric or substance abuse co-morbidities commonly found in patients with and without FPIs, with no significant difference between the two populations, were attention-deficit and disruptive behavior disorders (10.8% vs. 8.7%; p = 0.358), adjustment disorders (4.5% vs. 3.0%; p = 0.319), cannabis abuse (12.4% vs. 8.7%; p = 0.133) and opioid abuse (5.7% vs. 5.3%; p = 0.831).

†Mood Disorder Due to General Medical Condition and Mood Disorder NOS.

‡Brief Psychotic Disorder, Delusional Disorder, Psychotic Disorder Due to General Medications, Psychotic Disorder NOS, Schizoaffective Disorder, Schizophreniform Disorder, Shared Psychotic Disorder.

§Diagnosis or Condition Deferred on Axis, No Diagnosis on Axis II, No Diagnosis or Condition on Axis I, Unspecified Mental Disorder (Nonpsychotic).

¶Academic Problem, Acculturation Problem, Adult Antisocial Behavior, Age-Related Cognitive Decline, Bereavement, Child or Adolescent Antisocial Behavior, Identity Problem, Malingering, Non-Compliance With Treatment, Occupational Problem, Phase of Life Problem.

Cost analysis

Mean direct 12-month psychiatric per-patient costs for patients with FPIs were more than double those for patients with fewer interventions ($11,414 vs. $4790, respectively) (Table 5). The largest percentage of total direct psychiatric costs was attributable to inpatient hospitalizations in patients with FPIs (43.7% of total direct costs). Psychiatric medication costs were the second-largest cost component (26.1%), followed by psychiatric outpatient visits (18.2%) and psychiatric ER visit costs (12.0%). In patients with fewer interventions, psychiatric medication costs were the largest cost component (42.8%), followed by inpatient psychiatric hospitalization (31.4%), psychiatric outpatient visits (19.6%) and psychiatric ER visits (6.2%). For psychiatric medication costs, antipsychotics were the highest cost driver in patients with FPIs compared with patients without FPIs ($1165 vs. $626), followed by anticonvulsants ($1081 vs. $850).

Table 5.  Annual per-patient direct psychiatric medical and pharmacy costs in patients with and without FPIs.

Annual psychiatric costs	Patients with FPIs
	Yes (n = 332) (US$)	No (n = 300) (US$)
Direct psychiatric medical costs
ER visits	1371	299
Inpatient hospitalization	4982	1505
Outpatient visits	2078	937
Total direct psychiatric medical costs	8431	2742
Psychiatric pharmacy costs
All antipsychotics	1165	626
Antidepressants	610	487
Anticonvulsants	1081	850
Lithium	59	39
Benzodiazepines	68	46
Total psychiatric pharmacy  medication costs	2983	2048
Overall total direct psychiatric cost per year (direct medical costs plus pharmacy costs)	11,414	4790

FPIs, frequent psychiatric interventions.

Predictors of frequent psychiatric interventions

Multivariate logistic analysis demonstrated that several patient characteristics were associated with FPIs (Table 6). When only demographics were included in the model, females were 1.595 times more likely to require interventions than males (p = 0.005) and African Americans were 1.564 times more likely to require FPIs than non-African Americans (p = 0.033). However, when all factors were included in the model, gender and race were not statistically significant. Patients with a personality disorder and those with relationship problems were 2.675 (p < 0.0001) and 2.818 (p = 0.0034) times more likely to require FPIs, respectively. Patients exhibiting a high degree of suicidality were 2.561 (p < 0.0001) times more likely to require FPIs than those without demonstrated suicidal tendencies.

Table 6.  Variables associated with FPIs (multivariate analysis).

Variable	OR	95% CI	p-value*
Only demographic factors included in the model
Age, years (≥40 vs. <40)	0.750	0.568–1.077	0.132
Race, African American (yes vs. no)	1.564	1.036–2.362	0.033
Gender, female (yes vs. no)	1.595	1.149–2.214	0.005
All factors included in the model
Age, years (≥40 vs. <40)	0.950	0.666–1.355	0.776
Race, African American (yes vs. no)	1.145	0.717–1.829	0.570
Gender, female (yes vs. no)	1.288	0.897–1.849	0.171
Depressive disorders (yes vs. no)	1.373	0.930–2.026	0.111
Anxiety disorders (yes vs. no)	1.317	0.892–1.943	0.166
Personality disorder (yes vs. no)	2.675	1.666–4.296	<0.0001
Relationship problems (yes vs. no)	2.818	1.408–5.642	0.0034
Substance abuse (yes vs. no)	1.077	0.732–1.584	0.708
Baseline CGI-S score (≥5 vs. <5)	1.148	0.991–1.331	0.066
Suicidality (yes vs. no)	2.561	1.764–3.720	<0.0001

CGI-S, Clinical Global Impressions–Severity; CI, confidence interval; FPIs, frequent psychiatric interventions; OR, odds ratio.

*Calculated using chi-square test.

Discussion

This study evaluated a patient population with bipolar disorder and clinically defined FPIs. As described previously, FPIs were used as a surrogate for disease severity. Given the nosological limitations of ‘rapid-cycling’ as a course descriptor for bipolar disorder16–18, a bipolar patient population with functionally defined FPIs was studied. Compared with rapid-cycling patients, patients in the current study did not necessarily meet all the specified criteria for ‘rapid-cycling’. Approximately 50% of the total bipolar population evaluated in this retrospective review were classified as requiring FPIs.

The results of this study suggest bipolar patients requiring FPIs can be distinguished by several important characteristics. In this database, patients with FPIs were more likely to be younger, to be African American and to be female. Their disease was more severe as measured by baseline CGI-S ratings, use of psychotropic and non-psychotropic medications, psychiatric co-morbidities and concurrent substance abuse. In addition, patients with FPIs exhibited more suicidal thinking and behavior than patients without FPIs. Additionally, the multivariate analysis suggests that suicidality, relationship problems and personality disorders are also significantly associated with FPIs. Although race and gender were initially significant predictors of FPIs, they were not significant predictors when co-morbidities, baseline CGI-S and suicidality were included in the multivariate model.

Based on substantial evidence that bipolar disorder is associated with significant impairment to functioning and well-being19, healthcare resource use and cost were examined. A better understanding of these variables in patients with bipolar disorder may be helpful for healthcare payers and others interested in disease management and cost containment. Cost-based estimates of the disease burden are useful because the unit of analysis is comparable across different illnesses and the cost estimates are comprehensive.

The resource use observed in this analysis indicates the number of days of acute bipolar illness and the extent of the associated disability, particularly in patients requiring FPIs. On average, within a 12-month time frame, this subgroup experienced approximately 2.2 ER visits, 5.5 days of hospitalization and more than 24 outpatient visits. The direct per-patient 12-month healthcare cost for bipolar patients with FPIs was in excess of $11,000 – more than twice the cost for patients without FPIs. The average per-patient cost observed in this analysis is similar to that found in previous cost studies of bipolar disorder8,13,20,21. The majority of total direct costs for patients with bipolar disorder requiring FPIs resulted from inpatient hospitalization.

Several limitations of this work should be noted. First, these results are from a single healthcare system and therefore may not be generalizable to all patients with bipolar disorder. Differences in healthcare plans and patient characteristics must be considered before attempting to extrapolate these findings to other settings and patient populations. Furthermore, this analysis was retrospective and relied on adequate information recording and coding of diagnoses subject to physician judgment. As a consequence, results reported here represent an approximation of both disease classification and disease burden. Finally, any resource utilization outside the Duke system was not captured in this analysis. However, most patients received the majority of their healthcare services within the Duke system.

Cost data from the PharMetrics database may not reflect the cost of patients who have no private health insurance, and socio-economic status was not considered for this analysis. Additionally, costs for psychiatric co-morbidities were assessed at the final visit. Costs for co-morbid medical conditions were not readily available, so this comparison and comparisons of total healthcare costs (including all co-morbidities) could not be made.

This was the authors’ most homogenous and recent source of cost data. Identification of cost data for bipolar disorder is difficult because psychiatric care is not included in many available databases. However, this database contains information across more than 30,000 de-identified patients who have experienced more than 200,000 qualified clinical visits. As such, it provides a naturalistic, ‘real world’ insight into characteristics of persons with bipolar disorder who require FPIs. Additionally, patient visits are tracked over time, enabling large-scale, naturalistic examination of patient progress over long periods. A future goal is to continue to build a model using a longitudinal psychiatric database to identify a bipolar patient profile. Clinicians may use this profile for their patients with bipolar disorder to avoid FPIs.

Conclusion

Patients with bipolar disorder who were identified in this analysis as requiring FPIs were significantly different from those requiring fewer interventions on assessments of demographic and clinical characteristics, psychiatric co-morbidities, resource utilization and medication type. These data suggest patients with bipolar disorder with FPIs have distinguishing characteristics and require significant psychiatric intervention. The work described here may facilitate the prospective identification of this particular subpopulation, and cost–benefit analyses can be conducted to improve patient outcomes and reduce overall healthcare costs.

Transparency

Declaration of funding:

This analysis was supported by funding from Ortho-McNeil Janssen Scientific Affairs, LLC, Titusville, NJ, USA. Duke University Behavioral Health Informatics was contracted by Johnson & Johnson Pharmaceutical Services, LLC, to access and analyze the data provided in this manuscript.

Declaration of financial/other relationships:

J.T.H. is a full-time employee of Johnson & Johnson Pharmaceutical Research and Development, LLC, and a Johnson & Johnson stockholder. N.T., C.C., R.D. and L.A. have disclosed that they are full-time employees of Ortho-McNeil Janssen Scientific Affairs, LLC, and Johnson & Johnson stockholders. W.M. has disclosed that at the time of this analysis he was a full-time employee of Ortho-McNeil Janssen Scientific Affairs, LLC. B.B. and K.G. have disclosed that they have no conflicts of interest to report.

Acknowledgments:

The authors wish to acknowledge the writing and editing assistance provided by Matthew Grzywacz, PhD, and ApotheCom (funding supported by Ortho-McNeil Janssen Scientific Affairs, LLC) in the development and submission of this manuscript. The authors also wish to acknowledge Earle Bain, MD, for his significant contributions to the early stages of this work. Dr Bain is a former employee of Ortho-McNeil Janssen Scientific Affairs, LLC, and is currently an employee of Abbott Laboratories, Inc.

The study data have been presented at the following congresses:

Society of Biological Psychiatry Annual Meeting, May 17–19, 2007, San Diego, CA; International Conference on Bipolar Disorder, June 7–9, 2007, Pittsburgh, PA; US Psychiatry and Mental Health Congress Annual Meeting, October 11–14, 2007, Orlando, FL; Institute on Psychiatric Services Annual Meeting, October 2--5, 2008, Chicago, IL, USA.