Article Text

Download PDFPDF

Improving depression care in a psychiatry resident psychopharmacology clinic: measurement, monitoring, feedback and education
  1. B Shiner1,2,3,
  2. R L Green1,4,
  3. K Homa3,
  4. B V Watts1,5,
  5. A Groft1,
  6. W C Torrey1,
  7. T E Oxman1,2
  1. 1Department of Psychiatry, Dartmouth Medical School, Lebanon, New Hampshire, USA
  2. 2Department of Community and Family Medicine, Dartmouth Medical School, Hanover, New Hampshire, USA
  3. 3Dartmouth Hitchcock Leadership Preventive Medicine Residency Program, The Dartmouth Institute for Health Policy and Clinical Practice, Hanover, New Hampshire, USA
  4. 4Department of Psychology, Dartmouth College, Hanover, New Hampshire, USA
  5. 5VA National Center for Patient Safety Field Office, White River Junction, Vermont, USA
  1. Correspondence to Dr Brian Shiner, VA New England Health Care Engineering Resource Center, 215 North Main Street, 11Q White River Junction VT 05009 brian.r.shiner{at}


Objective Better outcomes for major depressive disorder (MDD) are associated with proactive treatment, including timely follow-up, systematic assessment and treatment changes for inadequate improvement. The effectiveness of an intervention to facilitate proactive treatment for MDD in a resident psychopharmacology clinic was studied.

Methods A quality improvement program with administrative process changes to improve flow and a 40-week pre/post study to evaluate the effect of education and feedback was conducted. A systematic assessment and reengineered scheduling system were implemented. During the first 20 weeks, baseline data were collected; during the second 20 weeks, feedback to residents and attending psychiatrists about adherence to evidence-based treatment recommendations was added.

Results Reengineering our system to improve flow was successful. By linking outcomes collection to completion of billing sheets, outcomes at 90% of visits for MDD throughout the 40-week study was assessed. By centralising our scheduling system, the percentage of active-phase patients with MDD seen for follow-up within 6 weeks was improved from 19% to 59%. In response to feedback, residents did not make significant changes to their overall practice patterns. Patient outcomes did not improve as a result of feedback to residents. Residents did improve their practice patterns for a subset of patients including those without comorbid psychiatric disorders and those whose depressive episodes had lasted <1 year.

Conclusions Improving administrative processes for the treatment of patients with MDD resulted in rapid changes that were associated with improvements in the delivery of evidence-based care. Feedback to residents was more difficult and less successful.

  • Major depressive disorder
  • quality assurance
  • guidelines
  • feedback
  • internship and residency
  • clinical practice guidelines
  • evidence-based medicine
  • graduate medical education
  • statistical process control
  • patient outcomes

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Care of major depressive disorder (MDD) has been identified by the Institute of Medicine as a priority area for quality improvement.1 Inadequate pharmacologic treatment of MDD has been demonstrated in both general and speciality mental health settings.2–4 Delay in initiating effective treatment for MDD is harmful, as suffering is prolonged and the effects of treatments may be less robust.5 Patients with MDD who reach the inpatient psychiatry level of care for suicide attempts have frequently received inadequate antidepressant treatment before and after hospitalisation.6 7 In outpatient psychiatric settings, evidence suggests mental health specialists delay initiating antidepressant treatment and are often slow to increase to adequate doses.8 9 This is unfortunate because patients who are treated aggressively according to evidence-based guidelines fare better than those who are not.5 Furthermore, the best results are obtained when prescribing is aggressive and linked to measured clinical outcomes rather than to unsystematic clinical practice.10

Consensus-based guidelines published by the American Psychiatric Association (APA) and the Food and Drug Administration at the time our intervention was planned called for at least weekly contacts in the first month an antidepressant is prescribed, followed by at least monthly visits until the end of the acute phase of treatment, remission of the depressive episode.11–13 Recent work lends empirical support to the APA's recommendation for reassessment after 4–8 weeks because most patients who will eventually respond to an antidepressant show at least a 30% improvement after 6 weeks.14


We randomly selected 100 charts for a 6-month period to characterise baseline depression care in our resident psychopharmacology clinic. MDD was the most common diagnosis in 51 (51%) of patients. Patients were seen an average of 2.3 times over 6 months and had been treated in the clinic for their current episode for an average of 11.7 months. Analysis of Antidepressant Treatment History Form scores revealed that 78% of patients were prescribed adequate antidepressant trials, which was higher than reported in other inpatient and outpatient psychiatric settings.9 15 16 The time between appointments was almost 3 months, whereas treatment guidelines and the best evidence on monitoring, reviewed above, suggested reevaluation within 6 weeks.

Purpose of change

Our overall goal was to improve our treatment for MDD. Based on our assessment, we designed process improvements to reduce the number of weeks between visits and to routinely measure clinical outcomes. We combined process improvements with an educational intervention designed to improve residents' skill in proactively managing MDD. Because we followed methods largely developed in primary care, our first study question was whether the methods that we chose were adaptable to a psychiatric clinic.17 18

Our second study question was whether an improved process would improve patients' outcome. We hypothesised that prolonged time between medication appointments was a major factor leading to lengthy active-phase treatment. Therefore, we sought to determine whether more timely follow-up visits and more proactive treatment would improve patient outcomes.



Our psychopharmacology clinic is located at a rural academic medical centre. Twelve residents were supervised in pairs by four attending psychiatrists. Each resident saw patients in the clinic one afternoon per week. The clinical process followed a standard teaching model: a resident interviews the patient and then discusses findings with an attending psychiatrist in another room. Both physicians then see the patient together to confirm findings and convey the treatment plan.


All patients diagnosed with MDD were included in the data collection process. We relied on a weekly chart review, rather than a formal diagnostic assessment, for diagnosis of MDD in this naturalistic study. The diagnosis of MDD was assigned if it was reported in the problem list or clinical assessment. Psychiatric comorbidities were assessed in the same way. Patient and treatment variables were collected for each patient visit and are summarised in table 1 below.

Table 1

Primary variables collected with chart review

In establishing our measurement procedure, residents felt that the intervention would be most helpful for patients who were not “chronic” (depressed for <1 year) and not “complicated” (no psychiatric comorbidity). Therefore, the data collection procedure included classification of patients based on these categories. The third patient variable, clinical status, was created to identify patients in the active phase of depression treatment. Clinical status was determined by review of the resident and attending psychiatrist's written assessment and diagnostic code, as MDD diagnostic codes include an indicator for clinical status.

There were four treatment variables. We chose 6 weeks as our timely follow-up standard. Although we recognised that Food and Drug Administration and APA guidelines called for a more rapid follow-up, we chose a goal of 6 weeks because this would represent almost a 50% improvement and our literature review suggested that 4 to 6 weeks is an ideal time to make an antidepressant change due to non-improvement. Adequacy of antidepressant trials was assessed with the Antidepressant Treatment History Form. The last two treatment variables were linked to assessing depression outcomes using the Patient Health Questionnaire-9 (PHQ-9).19 The PHQ-9 is sensitive to change, and the minimal clinically important difference (MCID) in score has been determined to be five points.20 21 This allows physicians to know whether the antidepressants they prescribe have been effective. In the active management of MDD, clinically important improvement should be seen within 4 to 6 weeks of a change in antidepressant medication or dose.22 We expressed clinical improvement rate as PHQ-9 change per month. An adequate rate of improvement was considered five points per month. Prescribing behaviour was assessed by determining whether residents made changes to the antidepressant regimens when patients failed to improve at an adequate rate.

Planning the intervention

The first system change involved the routine administration of the PHQ-9. The questionnaire was printed with the billing sheet when residents' patients signed in for their appointments. The printing was a cue for the receptionist to ask the patient to complete the form, which the patient handed to their resident at the start of an appointment. Over the 3 months before the study, residents were trained to interpret the PHQ-9 by a three-part educational process, based on a predisposition–enablement–feedback model.18 Predisposition began with introducing residents to the tool, which generated interest and some experience in a “measurement-based” model of care for depression.10 Enablement involved minimising residents' expenditure of extra time or effort by having the patient fill out the questionnaire in the waiting room and by posting rules for interpretation in exam rooms. Feedback is described below, as it comprised the intervention phase.

The other process change aimed to improve the time between appointments. Previously, residents kept their own calendars and generally scheduled patients for follow-up at their next available appointment. This created a backlog of patients and contributed to long follow-up times. Beginning on week 1 of the study, residents wrote the desired follow-up interval on a card which the patient then gave to the receptionist. Whenever possible, the appointment was scheduled with the patient's own resident; when this was not possible, the appointment was scheduled with the other resident supervised by the same attending psychiatrist. If no appropriate appointments were available, patients were scheduled in the next available slot, but the clinic secretary moved the appointment forward when there were cancellations.

Data collection followed the implementation of process improvements and was planned for 40 weeks (3 July 2006 to 30 March 2007). Data collection occurred in two blocks of 20 weeks each. The first 20 weeks was considered baseline so improvements could be assessed relative to the implementation of feedback which occurred during the second 20 weeks. The Dartmouth College Committee for the Protection of Human Subjects determined that this project did not meet the definition of human subjects' research. This project was classified as quality improvement.


Feedback was provided in a variety of forums throughout the second phase of the project. Monthly, residents received statistical process control and analysis of means charts describing depression treatment in their own panel of patients and in comparison with the rest of the clinic.23 Residents were instructed in interpretation of statistical process control and analysis of means charts.24 In addition, residents received a data sheet that included raw data on patient and treatment variables including specific psychiatric comorbidities, psychotherapy participation and missed appointments for each of their patients with MDD. Attending physicians received their individual residents' data as well as aggregate data on all residents whom they supervised. Finally, two of the authors (BS and RG) met with each of the attending psychiatrists and one of their residents to coach them on use of feedback as part of routine supervision.

Analytical methods

To determine whether the scheduling system changes were effective, a statistical process control chart (P chart) was used to plot the proportion of patient visits seen within 6 weeks for the first phase of the study. To determine whether the resident feedback intervention was effective, data were analysed pre (first 20 weeks, no feedback) versus post (second 20 weeks, feedback) using χ2 analysis (Fisher exact test for when there were less than five patient visits in a cell). A p value of <0.05 was considered significant.

Generalised estimating equation (GEE) logistic regression was used to analyse the association between adequate improvement rate and timely follow-up since logistic regression methods are not designed to account for the correlated nature of the data correct. This method was also used to determine whether the rate of adequate improvement on subsequent visits was enhanced if medication was changed for the subgroup of patients who failed to improve despite timely follow-up. Models were adjusted for comorbidity, chronicity and phase of the study (pre vs post feedback). Statistical analyses were performed using Intercooled STATA V.9 (STATA Corporation, College Station, Texas, USA).


Situation analysis

Over the study period, there were 549 visits from 187 patients with MDD with a median of three visits per patient with MDD. The PHQ-9 was completed for 90% (492/549) of the MDD patient visits; 492 of 549 visits. Sixty-five patients either completed only one PHQ-9 during the study period or did not complete the PHQ-9 at all, so their outcomes could not be assessed. The PHQ-9 was administered twice or more on 122 patients (433 visits).

Patient characteristics

As seen in table 2, the mean age of patients with MDD as of the first study visit was 47.8. 68% were women and 60% were married. Clinic patients were 55% employed and 65% had private insurance. Only 3% of patients had no form of insurance. Data on race were not available. With regards to chronicity, 42% of patients had a new depressive episode and 29% of patients had a depressive episode of >1 year at their first study visit. Psychiatric comorbidity was common, with 58% of patients carrying a mental health diagnosis in addition to MDD. The most common comorbidities were post-traumatic stress disorder (13%) and generalised anxiety disorder (11%). With regards to severity as assessed with the PHQ-9, 15% of patients were followed for ongoing remission (score <5) and 44% were moderately to severely depressed (score of ≥10) as of their first study visit.

Table 2

Patient demographic characteristics as of first visit during the data collection period


Figure 1, a statistical process control chart (P chart) shows the percentage of patients per week who were seen within 6 weeks of their last appointment; scheduling system changes improved the rate from 19% to 59%.

Figure 1

P chart of percentage of follow-up visits seen within 6 weeks (dash lines are the σ control limits).

To determine intervention efficacy of the feedback intervention, the proportions of four treatment characteristics (timely follow-up, adequate trial, adequate improvement and made indicated change in medication) were compared using χ2 tests (first phase vs second phase). Table 3 shows the χ2 results for the four treatment variables and also the results relative to the two specific patient populations. The unit of analysis is patient visits rather than patients. For all patient visits, there were no differences in the variables between the first and second half of the study period, whereas changes were seen in the subgroups.

Table 3

Comparison of outcomes measures before and after the feedback intervention

Over the 40-week study period, 51% of the patient follow-up visits occurred within 6 weeks and 82% of patient visits included adequate antidepressant trials. Patients on PHQ-9 improved at an adequate rate in 15% of follow-up visits, and patients had a change in their medication when one was indicated at 50% of follow-up visits. When the stratified analysis was performed, there were two significant changes. A larger proportion of follow-up visits for patients without comorbid psychiatric disorders occurred within 6 weeks of the second phase of the study compared with the first (30% vs 48%; p=0.042). Residents made indicated medication changes during a larger proportion of follow-up visits for patients who had a depressive episode of <1 year in the second phase of the study compared with the first (40% vs 58%; p=0.046).

To determine the relation between follow-up interval and outcome, a GEE logistic regression model was adjusted for comorbidity, chronicity and phases of the study. There were 283 patient visits (118 patients) in this analysis. The odds of clinical improvement were 17.3 times greater in timely follow-up visits (95% CI 5.1 to 58.6; p<0.001) compared with follow-up visits that occurred in >6 weeks. Thus, at 26% of timely follow-up visits (39 out of 152), patients improved at a significant rate compared with 2% at follow-up visits (3 out of 131) that occurred in >6 weeks.

In the group that failed to improve at a given visit despite timely follow-up, but had subsequent timely visits (n=135, 68 patients), a GEE logistic regression model was adjusted for comorbidity, chronicity and phase of the study. The odds of clinical improvement at subsequent visits were two times greater for visits which followed a visit where an indicated medication change was made compared to visits which followed a visit where an indicated medication change was not made. However, this result was not statistically significant (95% CI 0.8 to 5.5; p=0.133) but suggests a possible trend.



Administrative process changes were effective both in ensuring that the PHQ-9 was reliably administered and that more patients were seen for follow-up within 6 weeks. Linking the PHQ-9 questionnaire to the billing sheet, which follows the patient throughout the visit, was a reliable way for the questionnaire to be completed by the patient before provider interaction. The creation of centralised scheduling created a significant improvement in timely follow-up.

The hypothesis that residents would change their overall practice of care delivery when given feedback was not substantiated. Residents did, however, make changes in their practice patterns in subgroups that they had previously identified as most likely to benefit from the intervention. Although they initially lagged in scheduling timely follow-up for “uncomplicated” patients and at making indicated medication changes in “acute” patients, they met these treatment standards at significantly higher rates during the feedback period. This suggests residents' adaptation of a standard methodology for treating depression as a result of the intervention.

Consensus-driven guidelines published by the APA recommend reassessment of the adequacy of a response to antidepressant medication after 4 to 8 weeks.11 Our work seems to provide support for this guideline. We determined that patients seen for follow-up within 6 weeks had better odds of improvement at significant rates on the PHQ-9 compared to patients seen at >6 week intervals.


This project has several important limitations. First, the lead author (BS) was centrally involved in the hypothesis generation, resident feedback intervention and data collection. As such, the results may have been unknowingly biased.

The use of the resident's diagnosis rather than a structured interview may limit this study's validity. However, the use of a structured research interview may have decreased the ability to generalise the results of the study by limiting the number of patients participating. A significant amount of teaching time was spent in ensuring that residents knew the value of the PHQ-9 in relation to the Diagnostic and Statistical Manual, Fourth Edition, Text Revision, criteria for major depression. The PHQ-9 provided residents with a criteria-based assessment of all patients who completed the instrument. Even if significant misdiagnosis did occur it would not alter the main results of this study because the treatments were selected based on the residents' belief that the patients had MDD.

Confounding by severity

The PHQ-9′s MCID five points, used here to identify patients not improving at an adequate rate, may not be appropriate to all score ranges. Using the Hamilton Rating Scale for Depression, Sackeim et al14 found that failure to achieve a 30% improvement at given time points was predictive of ultimate non-remission and may be an indication for medication adjustment. If this was applied to the PHQ-9, the true MCID would be eight points for a patient starting out with a score of 27 and three points for a patient starting out with a score of 10. Therefore, the MCID may overestimate adequate change for patients with high PHQ-9 scores and underestimate adequate change for patients with low PHQ-9 scores. In addition, we used the more aggressive window of 4 weeks to look for change in clinical outcomes. Given the finding of Sackeim et al that the best sensitivity and specificity in identifying ultimate non-remitters was at 6 weeks, this may be a more appropriate window.


We have demonstrated that depression management strategies developed in primary care can be successfully adapted to a speciality psychopharmacology clinic. It is possible to use routine outcomes assessment in a psychiatry resident psychopharmacology clinic. We replicated the previous findings suggesting that follow-up within 6 weeks is an important determinant of treatment success in the acute management of MDD. As our intervention did not change prescribing practices, further efforts should be made to ensure that changes to treatment regimens are made when patients fail to improve. It may be more helpful to provide information in real time rather than delayed pending chart review.


We would like to thank Leslie Papa, Marie Blair, Diana Shaw, Martha Catalona, Aroura Mazkin, Emily Woltman, Fred McClurg, Melissa Kostic, Justin Knapp, Jude Stenovitch, Bradley Erickson, Jennifer McLaren and Pamela Bokat for their help in making observations and leading changes to the clinic that made this work possible.



  • Funding Financial support for this work was provided by a grant from the Dartmouth-Hitchcock Quality Research Grant Program.

  • Competing interests TEO is a partner in 3CM, LLC, a company that provides consultation and training to healthcare organisations on management of mental health disorders in primary care settings. BW is a consultant for Pfizer, Janssen and Forrest Pharmaceuticals.

    BS had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of its data analysis.

  • Ethics approval The Dartmouth College Committee for the Protection of Human Subjects determined that this project does not meet the definition of human subjects research. This project has been classified as quality improvement (CPHS #20142).

  • Provenance and peer review Not commissioned; externally peer-reviewed.

Linked Articles