Prescribing psychoactive medications is the core of most psychiatrists’ clinical work. Many factors shape how, when, and for whom psychiatrists prescribe, including the product labeling authorized by the U.S. Food and Drug Administration (
1), treatment guidelines (
2), published clinical trials (
3), continuing medical education programs (
4), interactions with representatives of pharmaceutical companies (
5), training experiences (
6,
7), and formulary restrictions. The guidance from these sources is often conflicting, leading to uncertainty and inconsistency in psychopharmacological practice across prescribers. However, the degree to which psychiatrists agree or disagree on pharmacological approaches to clinical presentations has received little attention (
8).
Here we report the results of a survey distributed to psychiatrists attending one of two professional meetings in 2017. The survey aimed to assess the level of agreement among psychiatrists in the psychopharmacological management of patients presenting with common clinical scenarios.
Method
A survey presenting 15 clinical statements was administered at two separate psychiatric meetings in 2017: the Georgia Psychiatry Association (GPPA) annual meeting in Atlanta, Georgia, and the American Society of Clinical Psychopharmacology (ASCP) in Miami Beach, Florida. At the GPPA meeting, a paper version of the survey was distributed; at the ASCP meeting the survey was conducted via a Survey Monkey Web link. The wording of the survey statements was identical for both meetings, and no clarifications of the statements was provided by the investigators. The only difference between the surveys was a question for the respondent’s age: at the GPPA, respondents wrote in their current age; for the on-line survey, the ASCP respondents selected from age categories. Data collection was anonymous, and the Emory University institutional review board approved the study.
The topics addressed in the survey statements were selected by the authors on the basis of questions that repeatedly emerged during supervision of a resident’s psychopharmacology clinic. The survey collected information about the respondent’s age, gender, practice setting, practice type, primary patient population, and training status. The statements on the survey asked respondents to answer “yes,” “no,” or “not applicable” (N/A) (if a respondent deemed a statement to be outside his or her area of practice) in response to 15 statements. Fourteen of the statements were about clinical scenarios across a variety of psychiatric illnesses. The statements were phrased so that a “yes” answer indicated that the psychiatrist frequently, often, or usually used psychiatric medications as described in the statement. For example, “I frequently use an SSRI [selective serotonin reuptake inhibitor] as monotherapy (without a mood stabilizer) to treat depression in patients with Bipolar Type II disorder.” Inclusion of the frequency of the practice in the statement was made to avoid high rates of “yes” answers that could occur if the respondent could think of isolated cases in which she or he had prescribed medication as described in the statement.
The 15th statement read, “The psychopharmacology training I received during my residency adequately prepared me to manage psychiatric patients with confidence,” which was included as a potential moderator of the responses to the other statements. Pretesting of the survey included expert review of the items by psychiatry faculty who work in an academic psychopharmacology outpatient clinic and the solicitation of feedback from psychiatry residents. Their feedback was used to modify the phrasing of the survey items and its layout.
Survey responses were entered into SPSS (version 21) for data analysis. The simple percentages of “agree” or “disagree” responses were the primary outcomes to be reported. Comparisons between the GPPA and ASCP attendees were made using t tests for continuous variables and chi-square tests for categorical variables. For analyses of age on responses, the sample was divided into two groups, <55 years and ≥55 years, on the basis of a median split. We used a threshold of ≥80% agreement in responses to reflect “high consistency”; statements below this threshold were considered to have “low consistency.” Statistical significance testing was conducted without correction for multiple comparisons due to the exploratory nature of the survey.
Results
Forty-seven psychiatrists completed the survey at the GPPA meeting; 64 completed the on-line survey while at the ASCP meeting. The GPPA respondents were more likely to be female than were the ASCP respondents (
Table 1). Answers were missing for 0.005% of the survey statements. The number of N/A responses ranged from 0 to 5 for all statements, except for the statement related to requiring neuropsychological testing prior to prescribing stimulants (11 [10%] answered N/A).
Only one (1%) respondent reported his or her practice to be “mostly psychotherapy.” The overwhelming majority reported “mostly pharmacotherapy” (66%) or an “equal mix of pharmacotherapy and psychotherapy” (30%), with 3% indicating “other.” Nine (8%) of the respondents identified as psychiatry residents or fellows; all others had completed training. Thus, the sample can be considered to be highly engaged and experienced in the use of psychoactive medications.
Table 1 presents the characteristics of the respondents.
The proportions of psychiatrists endorsing “agree” or “disagree” to each of the 14 clinical scenario statements are presented in
Table 2. Of the 14 statements, six were considered highly consistent, and seven were considered to have low consistency.
Consistency of practice was greatest for the following: often prescribing bupropion monotherapy in patients with anxious depression (94.5% disagree); nearly always recommending avoiding marijuana to psychiatric patients (89.9% agree); and frequently using an SSRI as monotherapy for patients with bipolar depression type 2 (85.2% disagree).
In contrast, consistency in practice patterns was lowest for the following: whether the potential for weight gain or metabolic disturbances was the primary factor in choosing a first-line antipsychotic for a non-obese patient diagnosed as having schizophrenia (50.5% disagree); often prescribing a second SSRI for a patient with major depression who failed to respond to an initial SSRI trial (55% agree); and often co-prescribing a benzodiazepine with an SSRI when starting treatment for a patient with panic disorder (60.4% agree).
Although the results were largely consistent across the two groups of meeting attendees and practice variables, a few significant differences emerged. GPPA meeting attendees were more likely than ASCP attendees to endorse often using a second SSRI after one SSRI failure for depression (73.9% vs 42.2%; p=.001), more likely to require neuropsychological testing before prescribing stimulants (28.9% vs 12.9%, respectively; p=.047), and less likely to endorse frequently using quetiapine as a sedative (11.1% vs 26.7%, respectively; p=.049).
At a trend level, women were less likely than men to endorse prescribing benzodiazepines on the three statements that involved this class of medications (avoiding benzodiazepines for patients with a history of alcohol dependence: 78.1% vs 60.5%, p=.078; prescribing benzodiazepines for social phobia patients who had failed an SSRI and an serotonin–norepinephrine reuptake inhibitors (SNRI): 25.0% vs 42.5%, p=.088; and co-prescribing benzodiazepines with an SSRI when initiating treatment for panic disorder: 45.5% vs 66.2%, p=.041; see
Table 2). Women were significantly more likely to be <55 years of age, but age was not related to the responses to the benzodiazepine statements. Women were also significantly more likely than men to nearly always require neuropsychological testing before prescribing a stimulant for an adult with attention deficit/hyperactivity disorder (ADHD) (35.7% vs 12.7%; p=.009).
Residents and fellows, in comparison with those who had completed training, were less likely to feel comfortable using three psychotropics for bipolar depression (11.1% vs 67.0%; p=.001), more likely to often use a second SSRI after one SSRI failure among patients with depression (88.9% vs 52.6%; p=.036), and more likely to require neuropsychological testing for ADHD before prescribing stimulants (50% vs 16.7%; p=.043).
Notably, academic psychiatrists’ responses did not differ from all other respondents on any of the statements (all ps>.1). Psychiatrists practicing in public settings were significantly less likely than psychiatrists in all other settings to co-prescribe a benzodiazepine with an SSRI for panic disorder (33.3% vs 67.8%; p=.002).
Only 11 child psychiatrists completed the survey. In comparison with all other respondents, child psychiatrists more often endorsed the use of a second SSRI after a first SSRI had failed in a patient with depression (100% vs 51.1%; p=.002), and significantly greater emphasis on potential metabolic effects of antipsychotic selection for patients with schizophrenia (81.8% vs 47.2%; p=.03),
Respondents who reported nearly always requiring neuropsychological testing prior to prescribing a stimulant for ADHD were also significantly more likely to disagree with co-prescribing a benzodiazepine with an SSRI for panic disorder (p=.02) and with using a long-acting benzodiazepine for patients with a social anxiety disorder who failed to benefit from an SSRI and an SNRI (p=.001).
Eighty (72.1%) of respondents stated that their psychopharmacology training during residency adequately prepared them for practice, which was significantly related to age (<55 years: 86%; ≥55 years: 66%; p=.018). However, responses to this inquiry about training were not significantly related to responses for any of the 14 pharmacology statements.
Discussion
This survey identified several practice approaches that were highly consistent across psychiatrists for some clinical issues but found low consistency for other common and important clinical questions. The lower-consistency statements primarily involved medications with significant risks for harm, specifically benzodiazepines and antipsychotics.
The statement with the highest level of consistency was found for the statement, “In patients with major depression with anxious features, I
often prescribe bupropion as a monotherapy,” with ≥92% of both groups disagreeing with the statement. Bupropion is not approved for any anxiety disorder. Although early trials suggested that bupropion was as effective as SSRIs in patients with anxious depression (
9), a subsequent meta-analysis of 10 GlaxoSmithKline-funded trials comparing bupropion with an SSRI found a small but significant advantage for SSRIs (
10). In a previous survey of academic psychiatrists, the lack of efficacy for anxious symptoms was the primary reason identified for not using bupropion as the primary first-line antidepressant (
11). Our survey results are strikingly similar to those of a survey of ASCP members in 2013, which found that only 4% of respondents identified bupropion as their preferred first-line treatment for anxious depression (
8).
Another statement with high levels of agreement was, “I
always, or nearly always, recommend my psychiatric patients avoid using marijuana” (90% agreed). With several states now having approved medical uses of marijuana, some for psychiatric indications (posttraumatic stress disorder [PTSD]), the question of whether marijuana has any role in psychiatric treatment is coming to the forefront. Psychiatrists’ perceptions of patients’ use of cannabis have been little studied, despite the high use of cannabis use by patients with mental health disorders. An online 2013 poll of psychiatrists found that 85% expressed at least a moderate level of concern that cannabis is associated with adverse psychiatric effects (
12), consistent with the findings of the current survey.
Other scenarios with high consistency were as follows: not to frequently use an SSRI as a monotherapy for patients with bipolar type II disorder, consistent with the Canadian Network for Mood and Anxiety Treatments guideline for bipolar disorder, which lists SSRI monotherapy as a third-line option for bipolar II depression (
13); not to combine two antipsychotics after partial response to an initial antipsychotic, consistent with the American Psychiatric Association guideline for schizophrenia recommendation against routinely combining two antipsychotics (
14); not frequently using quetiapine as a sedative; and not to require neuropsychological testing prior to starting a stimulant in an adult with previously undiagnosed ADHD. No guidelines specifically recommend conducting neuropsychological testing for diagnosing ADHD in adults, though testing may have value for identifying comorbid learning disabilities, establishing a patient’s baseline cognitive performance, and identifying malingering (
15).
There was a nearly even split on whether the risk for metabolic side effects was the primary basis for choosing an antipsychotic for a patient with schizophrenia. This inconsistency may reflect the guidelines’ nonspecific recommendations to encourage prescribers to consider the various side effect profiles of antipsychotics when choosing an initial medication, without prioritizing any particular tolerability issue (
14,
16). Similarly, the high variability of using a second SSRI after initial SSRI failure among patients with depression reflects the nonspecific guideline recommendations for second-step medication treatment, which include considering a switch to a second SSRI after initial SSRI failure (
17,
18). The variability in clinical practices for these questions revealed in our survey may reflect variability in psychopharmacological training received and the tendency for residents to adopt the practices of their faculty supervisors rather than conform to treatment guidelines (
19).
The low level of agreement about the use of benzodiazepines for nonresponse to SSRI and SNRI among patients with social anxiety disorder is particularly noteworthy. In a meta-analysis of multiple placebo-controlled clinical trials for social anxiety disorder/social phobia, benzodiazepines were found to have an effect size greater than that of SSRIs (
20). The Canadian Network for Mood and Anxiety Treatments guidelines list long-acting benzodiazepines as a second-line intervention after SSRI or SNRI failure for social anxiety disorder (
21); in contrast, the National Institute for Health and Care Excellence guideline recommends using a monoamine oxidase inhibitor for these patients and against using benzodiazepines (
22). The lower consistency for all three statements related to benzodiazepine prescriptions is reflective of the ongoing debate about their appropriate clinical use (
23,
24).
That women were significantly less likely to co-prescribe a benzodiazepine with an SSRI for initial treatment of panic disorder, along with the trend-level findings of lower endorsement of using benzodiazepines for social anxiety disorder or previously alcohol-dependent patients with generalized anxiety disorder, is consistent with other studies that found that female psychiatrists were less likely to prescribe benzodiazepines than were male psychiatrists (
25,
26). Combined with the finding that female psychiatrists more frequently endorsed requiring neuropsychological testing prior to prescribing a stimulant for ADHD, these data suggest that, in general, female psychiatrists may be more cautious than males around prescribing medications with addictive potential.
Several limitations apply to the results of this survey. We could not determine the exact number of psychiatrists who received the survey, so the proportion of respondents out of all meetings’ attendees is unknown. In addition, the survey focused exclusively on psychopharmacologic practices. These two limitations may have introduced a participation bias in that the psychiatrists who chose to complete the survey may have been particularly interested in psychopharmacology and thus not representative of all practicing psychiatrists. For purposes of brevity, the survey did not ask the reasons why the respondents selected their answers, so it remains uncertain how much prescribing patterns are determined by product labeling, treatment guidelines, formularies, personal experience, or other factors. Several other areas of potential practice inconsistency, such as the pharmacological approach to treatment of PTSD or behavioral disturbances in dementia, were not addressed. Finally, our thresholds used to define high versus low consistency of responses were arbitrary, and our statistical tests did not correct for multiple comparisons.
Overall, these results suggest that most psychiatrists prescribe in accordance with many guideline recommendations but that they also deviate frequently. Whether these deviations from treatment guidelines are associated with better outcomes is unknown. The evidence base appears to inform, but not dictate, the practice patterns, which was the original intent behind developing evidence-based medicine (
27). Conducting practical clinical trials designed to answer treatment questions in which ambiguity currently exists would be likely to improve consistency between prescribers. For less ambiguous treatment issues, application of educational materials that increase prescriber awareness of the evidence base may also enhance consistency.