In 1999 the Accreditation Council for Graduate Medical Education (ACGME) mandated that residency programs document trainees’ proficiency in patient care as one of its six core competencies (1). Documentation of trainee proficiency in many areas of psychiatric practice, such as psychopharmacology, will prove challenging. We devised this pilot study to evaluate adequacy of psychopharmacological treatment of depressive disorders by psychiatric residents.
Psychiatric research indicates that depression is typically undertreated, both in primary care and mental health settings. The National Depressive and Manic-Depressive Association Consensus Statement on the Undertreatment of Depression asserts, "it is unfortunate that the vast majority of those treated with an antidepressant medication are not prescribed an adequate dose for a long enough time" (2).
An instrument designed to assess adequacy of medication trials for major depression, the Antidepressant Treatment History Form (ATHF), objectively quantifies antidepressant medication trials using a 1—5 scale and is now available as a computer-based algorithm (3—7). This instrument can help assess whether patients have failed to respond to acceptable medication trials and thus can be termed treatment-resistant. In our study, treatment of depressive disorders in a psychiatric resident outpatient clinic was rated using the ATHF. Data analysis sought relationships between severity of illness, treatment duration, and adequacy of antidepressant therapy. To our knowledge, no study directly addressing this issue has been previously published.
This study was reviewed and approved by the Wake Forest University Baptist Medical Center (WFUBMC) Institutional Review Board. Charts of patients seen in the WFUBMC Psychiatric Outpatient Department Triage Clinic from January 1, 2000 through December 31, 2000 were reviewed. The Triage Clinic meets weekly and is observed by a faculty psychiatrist via a one-way mirror. In this clinic, patients are seen by either a psychiatric resident or by a family therapy intern. If a resident is not seeing a patient, he or she also observes interviews behind the mirror, assisting in clinical decision making. At any time, this observing team may call into the interview room to request follow-up questions or give the interviewer guidance.
After the initial visit, patients are typically followed by the same second, third, or fourth year resident who participated in the initial interview. From this point on, visits are not directly observed, and supervision for difficult cases occurs as needed only at a weekly clinic conference.
A single faculty psychiatrist (who was not involved in this study) supervises the clinic. Patients referred to the triage clinic are typically indigent or receiving government aid.
Of the 285 reviewed charts, the 176 selected for further study were those reporting that the resident had diagnosed major depression; dysthymic disorder; depressive disorder, not otherwise specified; adjustment disorder with depressed mood; or bipolar disorder, with a depressed episode during the studied period. All diagnoses were made using DSM—IV criteria during a one-hour unstructured interview. Of these, excluded charts were: 18 patients not primarily treated by residents; eight patients who were referred back to their public mental health center or primary care physician after intake; 10 patients who received treatment by psychotherapy only, refused medications, or terminated treatment before a medication trial was undertaken; 11 charts that could not be located; and 17 charts that contained insufficient documentation to establish what diagnosis was given or medical regimen used.
Data were collected from 112 remaining charts for up to 52 weeks following the initial visit and included age, race, gender, and weeks followed in the clinic. Depressive disorders were coded as bipolar or unipolar (defined as all depressive disorders except depression in bipolar disorder) and as psychotic or nonpsychotic. Otherwise, details of diagnosis were not recorded due to incompletions in the residents’ charting (e.g., charts that simply noted "depression"). Presence or absence of comorbid Axis I diagnoses (but not specific comorbid diagnoses), Axis II diagnoses, and substance abuse was recorded. The reviewing author retrospectively assigned a Clinical Global Impression Severity (CGI) score, based on the patient’s most severely ill state during the studied year. A second author reviewed 12 randomly selected charts to assess interrater reliability for this measure.
Pharmacological treatment was rated using the ATHF, which assigns a score of 1—5 to a trial based on drug choice, dose, and duration of administration. Adequate treatment is defined as at least one medication trial rating ≥3. Examples of treatment meriting a score of 3 include a 4-week trial of fluoxetine at a dose of 20—39 mg/day or 4 weeks of nortriptyline therapy with a standard blood level of 50—99 ng/ml or dose 76—100 mg/day. Trials for psychotic depression receive a rating of 3 when the antidepressant would rate 3 independently and an antipsychotic is dosed at a level equivalent to ≥400 mg/day of chlorpromazine.
The sum and average of ratings of all psychotropic trials, the highest rating received, and a rating for the first trial initiated were noted. The first trial was defined as the plan recommended when the patient was first seen by the resident and scored as though this plan were carried out. For example, if a patient were given a dose titration schedule (e.g., sertraline 25 mg/day for 1 week, then 50 mg/day), scoring would be based on the highest planned dose, assuming it would be administered for a minimum of 4 weeks. This rating was created to gauge how aggressively the resident planned to treat the patient’s condition and was accessible even if the patient did not follow instructions or return for follow up. a1 summarizes ATHF scoring variables used in this study. Information on response was frequently indiscernible from available documentation and thus could not be used in the analysis.
Univariate and multivariate regression analyses were used to assess relationships between maximal and first ATHF scores and other variables (CGI score and time in treatment). The intraclass correlation was calculated for interrater reliability.
The 112 patients had an average age of 42.7 years SD=11.6; 69.6% (N=78) were female; and 78.6% (N=88) were Caucasian. Of the remainder, 23 (20.5%) were African American, and one (0.9%) was Asian. Additionally, 8.0% (N=9) were found to have bipolar depression, and 11.6% (N=13) had psychotic features. Beyond these distinctions, specific diagnoses were not recorded. Of the patients, 43.8% (N=49) had comorbid diagnoses on Axis I, and 12.5% (N=14) were given Axis II diagnoses. Substance abuse or dependence was diagnosed in 22 patients (19.6%).
Most patients (79.5%) were assigned CGI severity scores of 4 or less, and the average CGI score was 3.8 SD=0.9. Thus, the majority of the sample fell into the range of mild to moderate depression. Intraclass correlation for the randomly selected sample was 0.57.
Patients were seen in the residents’ clinic for an average of 23.9 (SD=21.8) weeks and were prescribed an average of 3.0 (SD=2.0) medications. Overall, the average ATHF score was 1.6 (SD=0.6), while the average highest score was 2.3 (SD=1.2). Based on a highest ATHF rating of at least 3, 49.1% (N=55) of the patients underwent an adequate antidepressant trial for depression. The proportion of patients attaining each potential ATHF score (1—5) as their highest is depicted in F1. Based on the rating of the first trial, 44.6% (N=50) of charts documented adequate pharmacotherapeutic interventions.
No relationship was found between patient age and treatment adequacy (p>0.84). Although previous research has found differences in intensity of antidepressant treatment based on race (4), no relationship was found between ethnicity and highest ATHF score in this study (p>0.9). There was also no association between gender and treatment intensity (p>0.4) or any interaction between race and gender in predicting treatment intensity (p>0.4).
Effort was also made to detect whether the severity of illness would predict treatment adequacy. No association was found between CGI severity rating and the highest (p>0.05) or mean ATHF score (p>0.1).
A strong relationship was discovered between intensity of treatment and time followed in the clinic. The highest observed ATHF score was greatly dependent on weeks retained in treatment (F=128.9, df=1,110, p<0.0001). Furthermore, 53.5% of the adjusted variance in highest ATHF score was explained by weeks in treatment. However, our univariate regression model found that a mean of 40 weeks in treatment was required to achieve a highest ATHF score≥3. In a multivariate linear regression, both the number of medication trials (p<0.01) and weeks in treatment (p<0.0001) were retained in a single model predicting the highest ATHF score (model F=72, df=2,109, p<0.0001). No relationship was found between weeks in treatment and the previously defined first ATHF score, describing the intensity of the initial medication plan (p>0.42).
Given the enormous burden of disability currently attributable to depressive disorders, maximizing the proportion of patients receiving at least one adequate medication trial should be a prime goal of psychiatric education. The first step in improvement is evaluating current training and performance.
Comparison With Previous Research on Treatment Adequacy
How does the adequacy of antidepressant treatment in this resident clinic compare with antidepressant treatment intensity in other settings? Previous studies have documented a range of adequacy estimates, in various clinical settings and with varying adequacy standards. Mulsant et al. used the ATHF to evaluate treatment of 53 psychotically depressed and 134 nonpsychotic depressed patients referred for ECT. In the Mulsant et al. study, systematic review found that 52% of nonpsychotic patients, and only 4% of psychotic patients received an adequate medication trial (5). The ATHF was also used by McCall et al. who found that 70.4% of 88 inpatients with major depression had received adequate treatment within the 6 months prior to admission, and the proportion receiving adequate treatment dropped to 60.3% over 12 months following discharge (6).
In a study of 4,103 prescriptions written by psychiatric and nonpsychiatric physicians for depressed patients, 35% of initial antidepressant trials were rated as inadequate in terms of duration and dosage (8). Dosage standards in that study referenced American Psychiatric Association (APA) practice guidelines for major depression, with an adequate trial defined as lasting 6 weeks. A review of pharmacotherapy following hospitalization for depression deemed 30% of trials insufficient, defining adequate continuation treatment as lasting 4 months and using dosages based on British Medical Association recommendations (9). For treatment in primary care settings, a 1992 study found that only 11% of depressed patients received an adequate dose or duration in 1 year of treatment. Time requirement for adequacy was 3 months (10). These studies may not directly compare to what should be expected of psychiatrists. In view of these previous findings, however, a rate among trainees of approximately 50% adequacy as defined by the ATHF is not surprising.
In this study, our estimate of treatment adequacy in this resident clinic could be biased based on the assessment tool. The ATHF was designed for use in major depressive episodes as part of major depressive disorder or bipolar disorder, whereas this sample contains a mix of depressive disorders. Is it possible that inadequate levels of treatment, as assessed by the ATHF, were administered to patients with conditions other than a major depressive episode? Is it a fair tool to use for these other disorders?
To assess this possibility, we must explore current treatment recommendations for miscellaneous depressive disorders. Treatment of dysthymia is less well researched than that of major or bipolar depression, but a proposed medication algorithm (11) describes target doses largely similar to those required by the ATHF. One difference in these guidelines is that the dysthymia algorithm proposes medication trials of 12 to 16 weeks, a more stringent requirement than that of the ATHF. No clear guidelines exist for treatment of adjustment disorders. However, a retrospective survey of 1,039 patients seen in consultation at university hospitals found that treatment with antidepressants for adjustment disorders did not significantly differ from other Axis I disorders (12). Therefore, there is no current research-supported evidence that would preclude using the ATHF to assess treatment of various depressive disorders although further investigation is needed.
If we assume our estimate of approximately 50% adequacy is reasonably accurate and should replication studies show these results to be representative of psychiatric residencies, how do we evaluate this result? It is certainly lower than what is called for given the pervasive and debilitating nature of depression. As a field, psychiatry must continue the campaign to improve treatment adequacy. A possible moderating factor is the population residents serve. Often, due to access and funding factors, resident clinics serve predominantly indigent patients. It is likely that this population has higher rates of noncompliance, attributable to socioeconomic and cultural factors. Dropout rates may also be higher in such clinics. The sample in our study included 26 patients (23.2%) seen only once. This group will also likely feature more chronic, treatment-resistant cases in which research-supported therapies have already failed. It may be that, with this population, attaining an adequate trial in 50% of patients approaches maximum possible success. Other possible contributors to the level of treatment adequacy by residents are decreasing available time for medication management visits (13) and suboptimal supervision or didactics.
Although no other studies, to our knowledge, have documented compliance of psychiatry residents with standards of outpatient treatment, literature on clinical performance of residents in other specialties exists. A 2001 study found a preeducational intervention rate of compliance with a hypertension practice guideline to be 32% among internal medicine residents (14). Another study found a baseline rate of compliance with standards for diabetes management of 40% among family practice residents (15). Keim et al. found that internal medicine residents documented use of standard preventive health care services in only 39% of appropriate cases (16). The recent Continuous Quality Improvement (CQI) study reported similar rates of treatment quality in 12 metropolitan areas (17). Perhaps in the academic setting, <50% compliance with treatment standards by physicians in training is typical.
Predictors of Treatment Adequacy and Implications for Psychiatric Training
In this study, we hoped that associations between patient characteristics and treatment adequacy might shed light on house officer training deficiencies. We hypothesized that more severely ill patients would be treated more aggressively and thus be more likely to receive an adequate trial. Nevertheless, there was no relationship found between CGI and adequacy. Does this imply a deficit in psychopharmacological or assessment skills? Here, the measurement tool may be suboptimal since it does not differentiate between duration and dose. It may be that antidepressant dose was increased more aggressively for severely ill patients and that the physician was more apt to abandon the trial or opt for hospitalization given their acuity.
Additionally, the CGI was assigned retrospectively by the investigators and does not necessarily correlate temporally with the maximal ATHF. Although one might expect more aggressive treatment (meriting higher ATHF scores) to correlate temporally with more severe illness (meriting higher CGI scores), our data cannot confirm this.
Data analysis found only one statistically significant relationship: weeks of follow up predicted treatment adequacy. What this means is not entirely clear. One problem with interpretation is that data were not coded to determine when in the follow-up period adequacy was attained. It is possible that patients retained for long periods acquired adequacy relatively early, began to recover, and stayed in treatment due to better satisfaction with their care. Those not managed aggressively early in treatment may have dropped out due to lesser satisfaction with the clinic. The lack of correlation between the first ATHF score and weeks in treatment argues against this hypothesis. However, it would be expected that adequate treatment early in the clinical course would be reflected in a more aggressive plan in place following initial presentation.
Alternatively, the correlation may mean that longer time in the clinic was required to attain treatment adequacy. This could be the case, for example, if dosages required slow titration or if multiple antidepressants were tried due to intolerability, both common clinical scenarios.
If this last possibility should prove the case, the implication for psychiatric residency training is that the therapeutic alliance is vital for adequate treatment of depression. Retaining patients in treatment, a difficult concept to teach, stands as the main predictor of an acceptable antidepressant trial in this study. Treatment alliance remains one of the most important components of psychiatric education, a skill residencies should ensure they are imparting to trainees.
There are several limitations and possible sources of bias. At 112, the sample size was relatively small. Furthermore, treatment by only 12 house officers was surveyed. Clinical Global Impression was assigned retrospectively by the authors and was not necessarily temporally correlated to the highest ATHF score.
Additionally, this study does not make provision for cases in which remission occurred before dosage adequacy requirements were met (e.g., a patient whose symptoms remitted on 10 mg of fluoxetine). This cannot be assessed due to the lack of response data. However, one would tend to assume such cases would be milder depressions. If milder depressions frequently required doses lower than considered adequate by the ATHF, one would also expect low CGI scores to correlate with low ATHF scores. No such correlation was found in this study. In any case, the results must be interpreted in light of the lack of outcome data, in part due to poor standardization of clinical records. Should these results prove typical of psychiatric residencies, it will become clear that record keeping should be better emphasized in residency training.
Moreover, the clinical use of polypharmacy was not addressed. Clinicians could use doses of more than one antidepressant, both lower than the ATHF definition of adequacy, and attain a good result. However, the patient would be characterized as having been given inadequate treatment. We cannot, nonetheless, directly assess the strategic use of polypharmacy since we did not code whether medications were used concurrently.
This study is also limited in its lack of assessment of nonpharmacological treatment (i.e., psychotherapy). Residents must meet specific psychotherapy practice requirements. We might assume residents use psychotherapy as the primary treatment for depression in a proportion of cases. It is also true that within medication management visits, psychiatrists and residents employ techniques characteristic of psychotherapy. How this may affect medication treatment is not well-characterized.
Another possible concern is that the residents did not make treatment choices independently and thus the results do not represent their actual skills. Whether the treating resident continued medication initiated elsewhere was not recorded in chart review. Quite possibly, the medication prescribed by the referring provider was continued very often. However, the results should, to some degree, reflect whether the treating resident felt the outside provider’s medication regimen was appropriate.
A final concern is that treatment primarily reflects preferences of the residents’ faculty mentors rather than their own. A single faculty psychiatrist who supervised the triage clinic for the entire study period observed the first interview of all patients. It is probable that treatment preferences of this attending psychiatrist exerted significant influence on which medications residents prescribed. Their choices were also most likely influenced by usual practice of their other faculty supervisors. This should not necessarily be considered a weakness, as the objective of this study was to describe the prescription practices trainees learn in residency programs where working with supervisors is a dominant educational tool.
Providing an adequate trial of medication to only one-half of patients presenting with depression is not an acceptable endpoint. Educational effort must continue pushing to improve this figure, and this data seems to indicate that further training in treatment alliance is a place to start.
The ACGME has moved toward outcome measurement as a determinant of accreditation. Psychiatric residencies will need ways to demonstrate that these skills, such as treatment alliance and pharmacological management, are being learned. This leads back to the issue of quantifying pharmacotherapeutic skill. Further research will be needed to elucidate the best way to document pharmacotherapeutic skill. This study used the ATHF to provide this documentation. Further research is also necessary to determine whether this could serve as a standard tool for assessing resident skill in antidepressant management. Future studies must require larger samples surveying greater numbers of residents and providing improved standardization of records (particularly the standardized rating of clinical outcome). Possibly, the ATHF could serve as a prototype for assessment scales addressing other psychopharmacologic issues. Its clarity, organization, ease of administration, and objectivity weigh in its favor as a useful tool for resident evaluation.