Use of Augmentation Agents for Treating Depression: Analysis of a Psychiatric Electronic Medical Record Data Set

In this retrospective, cross-sectional study, patients at least 18 years old with major depressive disorder without psychosis or psychotic features who initiated augmentation therapy between January 2001 and June 2011 were identified in MindLinc, a psychiatric electronic medical record (EMR) database representing an anonymous subset of clinical data generated from use at multiple sites. Institutional review board approval was not required because MindLinc data are anonymized and thus exempt from HIPAA requirements. The data set included diagnoses, clinical measure (Clinical Global Impressions–Severity [CGI-S] score), medications prescribed, and other medical conditions and services rendered.

The earliest date of augmentation was flagged as the index date. Augmentation was defined as prescription of a combination of antidepressants or an antidepressant and an agent not conventionally used as first-line monotherapy (second-generation antipsychotics, mood stabilizers or anticonvulsants, or stimulants). Medical records from a one-year preindex period were evaluated to identify patient demographic and clinical profiles, comorbid psychiatric and medical conditions, psychiatric drug utilization patterns, and site characteristics. Clinical severity of patients at the time of augmentation was classified as mild, CGI-S score of 2 or 3; moderate, 4; or severe, 5–7.

Descriptive statistics and frequency distributions were used to describe the sample and the type of augmentation strategy. Clinical and demographic differences between patients who received different augmentation strategies were examined by bivariate chi square analysis. Clinical and demographic predictors of the type of augmentation agent in a multivariate framework were assessed with logistic regression models. Variables that showed bivariate associations with the type of augmentation strategy at p≤.10 were used in the multivariate model. Augmentation with a second-generation antipsychotic was used as the reference category because it constitutes the only FDA-approved augmentation option. [Additional information about the study methods is available in an online data supplement to this report.]

A total of 3,209 patients initiated augmentation therapy for depression. Most patients had a diagnosis of recurrent major depressive disorder (58%), and many had moderately severe depressive symptoms (48%). A comorbid anxiety disorder diagnosis was documented for 48% of the patients. Antidepressant combination therapy was the most commonly used augmentation strategy (75%, N=2,420), followed by second-generation antipsychotics (11%, N=356), mood stabilizers or anticonvulsants (8%, N=266), and stimulants (5%, N=167).

A larger proportion of patients who were prescribed second-generation antipsychotics had recurrent major depressive disorder, compared with those who were prescribed an antidepressant combination, mood stabilizers, or stimulants (p=.034), and a larger proportion who were prescribed mood stabilizers had a diagnosis of major depressive disorder not otherwise specified (p=.039). Use of second-generation antipsychotics was associated with having a comorbid adjustment disorder (p=.016) and with having a personality disorder (p<.001), whereas augmentation with mood stabilizers was associated with having a comorbid somatic disorder (p<.001).

Among patients who were prescribed second-generation antipsychotics, a significantly larger proportion were classified as having severe symptoms at baseline (41%), compared with those who were prescribed a combination of antidepressant therapies (25%), mood stabilizers (27%), or stimulants (22%) (p<.001). Although patients who were prescribed second-generation antipsychotics had severe clinical symptoms (CGI-S scores=5–7), no significant differences in mean baseline CGI-S scores were observed by the specific agent prescribed.

A summary of predictors of augmentation with second-generation antipsychotics compared with other agents in a multivariate framework is presented in Table 1. Patients with adjustment disorder, personality disorders, somatic disorders, and prior use of benzodiazepines were more likely to receive augmentation with second-generation antipsychotics than with an antidepressant combination. Race and gender were strong predictors of augmentation with second-generation antipsychotics compared with mood stabilizers. Patients with substance use disorders and nonwhite patients were more likely to receive second-generation antipsychotics than stimulants. Disorders of infancy and childhood and adolescence were also strong predictors of stimulant use rather than use of second-generation antipsychotics.

Logistic regression analysis indicated that the baseline clinical severity of patients was the strongest and most consistent predictor of augmentation strategy. Patients with moderate clinical symptoms were 2.8 times more likely to receive a second-generation antipsychotic than combination antidepressant therapy compared with patients with mild depression (95% confidence interval [CI]=1.9–4.0). Similarly, moderate severity was associated with the selection of second-generation antipsychotics rather than mood stabilizers (odds ratio [OR]=3.4, CI=1.9–5.9) or stimulants (OR=4.1, CI=1.8–9.2). [Details of the results of all analyses are presented in the online supplement.]

This study provides real-world evidence of the treatments that specialty psychiatric providers use when managing patients with depression who do not adequately respond to initial antidepressant therapies. When a change in the treatment plan is necessary, current APA practice guidelines recommend that additional options include augmentation with either another antidepressant or with a nonantidepressant medication, such as lithium, thyroid hormone, or a second-generation antipsychotic (5). The augmentation patterns in this study are consistent with these recommendations: 75% of patients received an antidepressant combination, 11% were prescribed second-generation antipsychotics, 8% received mood stabilizers, and 5% received augmentation with stimulants.

The utilization rates of augmentation agents found in this study differ from those in the only previously published study on this topic (10), which found that among patients with major depressive disorder receiving antidepressant therapy augmentation, approximately half received a combination of antidepressants, one-third received second-generation antipsychotic augmentation, and the rest received anticonvulsants. Differences between our findings and the previously published evidence may be attributed to differences in setting, time frame, and data source. Our study used EMRs with a DSM-IV diagnosis, whereas the previous study identified VA patients by codes in administrative claims data, which are prone to missing and inaccurate diagnoses (11). It is also possible that the VA study included a disproportionately high number of patients with undocumented psychosis and other conditions who should have been excluded. Also, the previous study defined augmentation on the basis of a conservative estimate of a 60-day overlapping supply by using pharmacy claims (10), whereas our analysis assessed data on medication use with prescription records from providers. Use of a more stringent criterion through a different data set on medication use could have identified a lower proportion of patients treated with combination antidepressants compared with our estimates (10).

Our analysis showed that clinicians mainly prescribed a combination of antidepressants and used second-generation antipsychotics primarily for patients with severe depression; concomitant psychiatric conditions, such as adjustment disorder and personality disorders; and prior use of benzodiazepines. This finding is consistent with prior evidence showing that patients were more likely to receive augmentation with an antipsychotic agent if they had previous hospitalizations or comorbid posttraumatic stress disorder or were nonwhite (10).

Consistent with previous evidence, race and gender were strong predictors of augmentation with second-generation antipsychotics compared with mood stabilizers; second-generation antipsychotics were prescribed mainly to nonwhite males. Although African Americans and Hispanics were more likely to receive augmentation with second-generation antipsychotics, the overall rate of augmentation was lower for this population than for white patients (10). It is possible that these racial differences are driven by insurance status or treatment settings. In addition, the different rates of augmentation may result from differences in provider-patient communication or patient treatment preferences (10). In the current analysis, consistent with previous findings, patients with substance use disorders were more likely to receive augmentation with second-generation antipsychotics than with stimulants, possibly because these antipsychotic agents have been used off-label in the management of substance use disorders (12).

Although this study provides information on the type of augmentation agents used in clinical practice for major depressive disorder, it does not provide insight into the overall rate of use of augmentation therapy. Prior evidence has shown that only 4.2% of patients receive augmentation with an antipsychotic, anticonvulsant, or lithium, despite evidence showing significantly longer overall treatment duration compared with patients who continued on their initial antidepressant monotherapy or were switched to another antidepressant (13). In primary care settings alone, augmentation use was found to be only about 2% over a six-month follow-up period from the start of antidepressant monotherapy (14).

The main limitation of this study was the possible lack of comprehensive health care information in an EMR data set. Although medical records should have a detailed medical history, the data set is not based on a closed system and may lead to underestimation of medical conditions or prescription use. Therefore, it is possible that augmentation therapies were recorded at a date later than their initiation. However, exclusion criteria were established to ensure that the initiation of augmentation therapy was captured as accurately as possible. In addition, these results may underestimate comorbidities because some patients might not have been systematically evaluated for all psychiatric diagnoses, and diagnostic patterns are affected by variations in clinical detection over time. Insurance status was not available and can be an important factor affecting choice of treatment in real-world settings. Also, the longitudinal data repository is built through the use of EMRs, and not all data fields are mandatory or completed in full by physicians. Finally, the database is specific to specialty psychiatric providers, and the results may not be generalizable to the primary care setting.

Nonetheless, this study has unique strengths not available from claims data sets or survey research. MindLinc serves large, diverse populations, care settings, and a variety of treatment centers. This diversity and inclusion of all available psychiatric diagnoses provides an opportunity to examine more wide-ranging patterns of comorbid diagnoses in real-world settings and to document clinically important diagnoses omitted or overlooked in major surveys or administrative claims data. Another strength was the assessment of utilization patterns in relation to clinical severity, a data point generally not available in traditional sources of treatment pattern data, such as insurance claims.

In specialty settings, patients with major depressive disorder were often treated with a combination of antidepressants, whereas second-generation antipsychotics were more often used for patients with severe depression than for those with mild symptoms.

This study was supported by Bristol-Myers Squibb and Otsuka Pharmaceutical Co., Ltd. Editorial support for manuscript preparation was provided by Ogilvy Healthworld Medical Education and funded by Bristol-Myers Squibb, Scientific Connexions, Inc., and Otsuka America Pharmaceutical, Inc.

Dr. Bates is an employee of Bristol-Myers Squibb, where Dr. Sheehan, Dr. Zhu, and Dr. Kalsekar were formerly affiliated. Dr. Sheehan and Dr. Kalsekar are employees of AstraZeneca. Dr. Baker is an employee of Otsuka Pharmaceutical Development and Commercialization, Inc. The other authors have no competing interests.