OUP user menu

Improving outpatient care of depression by implementing practice guidelines: a controlled clinical trial

Isaac Bermejo, Frank Schneider, Levente Kriston, Wolfgang Gaebel, Ulrich Hegerl, Mathias Berger, Martin Härter
DOI: http://dx.doi.org/10.1093/intqhc/mzn050 29-36 First published online: 6 November 2008


Objective Depressive disorders are of great medical and political significance. Although evidence-based guidelines have been published and educational initiatives have been launched to implement them, they are rarely actually used. The aim of the study was to implement clinical practice guidelines for outpatient care of depression using a practice-oriented and interdisciplinary approach.

Design Controlled clinical trial with a naturalistic design (data collection within routine practice) designed as a prospective pre-post study.

Setting Outpatient care.

Participants 29 general practitioners (intervention: 18; control: 11) and 15 psychiatrists (intervention: 11; control: 4). Overall, the treatment of 698 patients (two samples: pre: 361; post: 337) was documented.

Intervention(s) Multifaceted intervention combining benchmarking, continuous medical education and interdisciplinary quality circles for the diagnosis and treatment of depressive disorders.

Main outcome measures Mixed-effects regression models for cluster-adjusted analysis of patients' symptom reduction.

Results Although physicians in the intervention group improved their clinical effectiveness (proportion of patients with response/remission) to a greater extent than physicians in the control group (intervention: 48.6% to 66.9%; control: 54.9% to 61.5%), cluster-adjusted analysis failed to prove a statistically significant effect of the intervention on the treatment outcome.

Conclusions Although no statistically significant improvements were found regarding the outcomes, the action programme provides important work, materials and results for an integrated treatment model for depression.

  • depression
  • clinical practice guidelines
  • implementation
  • outpatient health care system
  • quality assurance


Approximately 11% of the population has suffered from a depressive disorder over the past 12 months [1, 2]. The medical and social importance of depression is set to increase in developed countries over the next decades [3]. About 50% of depressed patients in primary care are not detected, only around one-third receive adequate treatment, and treatment is complicated by system-related barriers [48]. On the other hand, there is evidence that a guideline-adherent treatment and a better collaboration between primary and specialist care lead to marked improvements [9, 10].

Evidence-based guidelines on depression have been formulated in many countries and educational initiatives have been started to transfer these into practice [1114]. Although it remains unclear how best to implement guidelines into routine care, the adherence to a multilevel care model brings about an evident improvement [10, 12]. Multifaceted implementation strategies including continuing medical education, organizational interventions and regulatory measures have to be combined [12, 15, 16]. The effectiveness of such programmes in improving treatment behaviour has been proven for different diagnoses (e.g. asthma, depression, acute coronary syndrome) [10, 17, 18].

The present paper aims to investigate the effects of a specific training programme for implementing clinical practice guidelines and quality management measures on patient outcome and physicians' treatment measures in the outpatient care of depression.



Data were recorded within the study ‘Comprehensive Quality Management in Outpatient Care’ [19]. The study was a controlled clinical trial with a naturalistic design (data collection within routine practice) designed as a prospective pre–post intervention with control group in three regions of Germany. The study protocol was approved by the local Ethics Review Committee. Informed consent was provided by all the participants.

More than 400 physicians in outpatient care (private practices) were contacted, 69 of them (46 general practitioners, 23 psychiatrists) agreed to participate. Twenty-five physicians (36%; 16 general practitioners, nine psychiatrists) were excluded because of missing data (Fig. 1). The intervention group consisted of physicians in the areas of North Rhine-Westphalia and South Baden, and the control group was made up of physicians in the area of Munich. It was not possible to randomize the practices as the aim was to link up general practitioners and psychiatrists of one region. To avoid contamination in the control group, separate regions were selected.

Figure 1

Flow of physicians and patients through the trail.

For the present contribution, pre–post data of 29 general practitioners and 15 psychiatrists were analysed. A total of 3282 patients were screened and general practitioners' diagnoses recorded. Overall, the treatment of 698 patients (pre: 361; post: 337) was precisely documented.


The intervention group received guideline training for the diagnosis and treatment of depressive disorders [20]. The training was designed as a multifaceted intervention (six sessions of 3 h) combining benchmarking, continuous medical education and interdisciplinary quality circles. Benchmarking based on baseline data comprised weak point analysis and orientation towards models of best practice. Clinical medical education entailed training regarding the clinical practice guidelines for depression, a short version of these guidelines (desk aid), and a pathway for primary care [21, 22]. The quality circles aimed to reorganize the care of depressed patients. The training package was well accepted [20]. Over 65% of the physicians evaluated the applicability and 74% the benefit for practice as good. Clinical practice guidelines were evaluated by 88% of participants as very useful for routine.

Assessment and procedures

Before and after intervention, the physicians documented the first contact of patients with a depressive episode and the monitoring contact 8 weeks later. At each measurement point, patients filled out a questionnaire.

Physicians' documentation. The tool was specifically developed to record physicians' consultation proceedings [23]. Physicians assessed sociodemographic characteristics, psychosocial stressors, and the ICD-10 criteria for depression of all included patients, and described their diagnostic and therapeutic measures. In the documentation tool, the ICD-10 criteria were enumerated and physicians were required to mark all criteria relating to symptoms presented by the patients. The adequacy of pharmacotherapy was assessed by comparing the daily dose (mg/day) indicated by physicians with the corresponding dose recommended by the German Drug Compendium (Rote Liste).

Patient questionnaire. Upon inclusion in the study (baseline) and 8 weeks after this first consultation (follow-up), patients received a questionnaire assessing depressive syndromes with the German depression module of the Patient Health Questionnaire (PHQ) [24] as well as their depression-specific medical history and complaints. The PHQ-D was designed to diagnose mental health disorders in primary care. It provides a sum score for measuring depression severity (≥11 = cut-off with a maximal Youden Index for depressive syndromes) and a categorical algorithm for major and minor depression [25]. The PHQ has shown good sensitivity and specificity in different validation studies [26, 27].

Data analysis

The primary outcome was patients' dichotomized symptom reduction (remission or response) following 8 weeks of treatment assessed with the PHQ-D. According to guidelines, a complete remission was defined as the absence of clinically relevant depressive symptoms (PHQ score ≤9), and a response was defined as an improvement ≥50% in relation to the admission score. Secondary outcomes were the PHQ-D scale values (interval-scaled) and use of ICD-10 criteria, prescription of adequate antidepressant dose, and referrals to specialists (psychiatrists, psychotherapists) or inpatient care (all dichotomous).

Statistical analyses encompassed the fitting of mixed-effects regression models for cluster-adjusted analysis. For dichotomous outcomes, logistic models were applied and for the interval-scaled outcome, linear models were used. A random intercept was included in all models because data were grouped (clustered) by physicians. The models included main effects of intervention and time as well as their interaction term. Physician- and patient-level covariates were considered (adjustment). Effect sizes for dichotomous outcomes were displayed as adjusted odds ratios (adjOR) (i.e. the ratio of the odds of an outcome between two groups) with 95% confidence intervals. An odds ratio >1 indicates higher odds of the outcome in the group of interest compared with a reference group. In the present analysis, odds ratios >1 indicate superiority of the intervention group (over the control group), of the follow-up measurement point (compared with baseline), and a larger increase over time in the intervention group than in the control group. To describe the amount of heterogeneity between individual physicians, intraclass correlation coefficients were calculated for each outcome. Findings with a significance level of P < 0.05 were considered to be statistically significant (two-tailed tests). Analyses were performed with Stata 10 (StataCorp, College Station, TX, USA).


Apart from the number of depressive patients visiting the practice, which was significantly higher in psychiatrists' practices, no statistically significant differences were found between groups, either between physicians in the intervention and the control group, or between included and excluded physicians. Moreover, the sample was similar regarding outpatient care in Germany (Table 1) [28].

View this table:
Table 1

Physician characteristics

General practitionersPsychiatristsGeneral practitionersPsychiatrists
IG (n = 18)CG (n = 11)IG (n = 11)CG (n = 4)IG (n = 10)CG (n = 7)IG (n = 6)CG (n = 2)
Sex; male n (%)12 (66.7)8 (72.2)8 (72.7)2 (50.0)5 (50.0)6 (85.7)3 (50.0)1 (50.0)0.76
Age in years; mean (SD)48.9 (7.0)44.9 (5.9)45.7 (8.2)55.8 (9.1)48.0 (8.2)45.7 (7.1)49.8 (7.1)54.0 (14.1)0.23
Duration of practice in years; mean (SD)12.9 (8.1)9.4 (5.3)10.8 (8.0)18.8 (10.9)11.3 (6.2)11.4 (9.0)10.3 (8.5)13.5 (14.8)0.66
Type of practice; single n (%)15 (83.3)6 (54.5)7 (63.6)2 (50.0)7 (70.0)4 (57.1)5 (83.3)2 (100.0)0.61
Number of depressive patients per week; mean (SD)13.9 (8.5)12.5 (5.1)42.8 (25.1)53.8 (25.6)13.6 (13.0)21.4 (12.5)21.5 (12.6)35.0 (28.3)<0.001
Number of new depressive patients per week; mean (SD)2.31 (1.5)1.68 (1.5)5.6 (2.3)9.3 (3.0)1.3 (0.8)2.8 (1.9)5.1 (5.1)4.8 (4.6)<0.001
  • n = Sample size (total sample size ranges from 63 to 69 due to missing data across analyses); IG, intervention group; CG, control group; SD, standard deviation; P= statistical significance of test results for between-group differences.

Most of the general practitioner and psychiatrist patients were female and ∼50 years old, with general practitioners' patients in the intervention group being somewhat older (Table 2). Differences were found regarding family status, with the patients in the intervention group being more likely to be married. There were also differences regarding the relationship between minimal/mild and moderate/severe depression disorders. At baseline, more general practitioner patients of the control group had a moderate to severe depressive episode, and at both measurement points, more psychiatrist patients had a moderate to severe depression.

View this table:
Table 2

Patient characteristics

At general practitionersAt psychiatristsP-value
Intervention groupControl groupIntervention groupControl group
Baseline (n = 145)Follow-up (n = 147)Baseline (n = 70)Follow-up (n = 57)Baseline (n = 114)Follow-up (n = 100)Baseline (n = 32)Follow-up (n = 33)
Sex; male n (%)41 (28.7)31 (23.1)12 (27.1)17 (33.3)22 (19.6)29 (29.3)8 (25.0)10 (33.3)0.51
Age in years; mean (SD)52.5 (15.2)50.7 (14.8)48.1 (18.0)47.2 (16.6)46.1 (13.4)46.8 (12.9)47.5 (13.6)43.1 (13.2)0.006
Family status; n (%)
 Single10 (12.2)18 (15.0)13 (25.5)14 (28.0)14 (15.9)11 (13.9)7 (29.2)10 (40.0)0.001
 Married52 (63.4)79 (65.8)24 (47.1)23 (46.0)54 (61.4)51 (64.6)10 (41.713 (52.0)
 Divorced7 (8.5)16 (13.3)9 (17.6)3 (6.0)13 (14.8)10 (12.7)7 (29.2)2 (8.0)
 Widowed13 (15.9)7 (5.8)5 (9.8)10 (20.0)7 (8.0)7 (8.9)0 (0.0)0 (0.0)
Depression severity by the PHQ-D at beginning of treatment; mean (SD)11.2 (5.9)12.6 (6.3)13.3 (5.6)12.4 (5.9)15.1 (6.0)16.2 (6.1)13.6 (7.0)12.5 (5.8)<0.001
Depressive disorder by the PHQ-D at beginning of treatment; moderate/severe n (%)48 (52.2)73 (66.4)39 (78.0)31 (62.0)70 (78.7)66 (84.6)17 (73.9)18 (75.0)<0.001
  • n = Sample size (total sample size ranges from 516 to 671 due to missing data across analyses); SD, standard deviation; P= statistical significance of test results for between-group differences; PHQ-D, Patient Health Questionnaire (German version).

Regarding symptom reduction following 8 weeks of treatment (primary outcome), physicians of both the intervention and the control group improved their clinical effectiveness after guideline implementation. The patients' outcome (remission or response) in the intervention group increased by 18.3% (from 48.6 up to 66.9%) in the post-measurement point, whereas the control group only increased by 6.6% (54.9 vs. 61.5%) (Table 3). With regard to the secondary outcomes, no changes were found in the use of the ICD-10 criteria to confirm the clinical diagnosis with respect to the prescription of adequate antidepressant dose or in terms of the referral behaviour of the physician.

View this table:
Table 3

Results of mixed-effects regression analyses

OutcomenICCIntervention groupaControl groupaIntervention main effectbTime main effectcIntervention × time interaction
Remission or responsed3360.18353 (48.6)83 (66.9)28 (54.9)32 (61.5)1.080.881.680.311.770.35
Fulfilling ICD criteria4980.098120 (73.6)137 (72.9)50 (68.5)45 (60.8)1.130.780.770.501.340.55
Adequate AD dose3280.24493 (82.3)119 (86.9)26 (78.8)36 (80.0)1.500.551.460.570.970.97
Referral to psychiatrist4980.32216 (9.8)22 (11.7)11 (15.1)11 (14.9)0.900.881.050.921.010.99
Referral to psychotherapist4980.27124 (14.7)33 (17.6)17 (23.3)16 (21.6)0.590.350.870.751.480.48
Referral to clinic4980.1138 (4.9)11 (5.9)2 (2.7)5 (6.8)2.620.
Depression severity by the PHQ-D at end of treatment; mean (SD)3360.11210.4 (6.1)8.4 (5.8)10.1 (6.4)9.5 (5.4)−0.570.580.050.96−1.950.10
  • n = Number of subjects included in the analysis; ICC, intraclass correlation coefficient; adjOR, adjusted odds ratio (covariates: physician profession; family status of patients; depression severity at beginning of treatment); P = level of significance; apresented are absolute number of subjects (proportions), if not specified otherwise; bintervention group (coded as 1) vs. control group (coded as 0); cfollow-up (coded as 1) vs. baseline (coded as 0); dprimary outcome; B = regression coefficient (non-standardized) PHQ-D, Patient Health Questionnaire (German version); SD, standard deviation.

The cluster-adjusted analysis did not show any statistically significant effect of the intervention either for primary or secondary outcomes (Table 4). However, in the case of the primary outcome and for most of the secondary outcomes, adjOR of the mixed-effects regression analyses pointed in the expected direction (treatment improvement through intervention). In all analyses, a statistically significant effect was found for depression severity at the beginning of treatment, showing that with decreasing depression severity at the beginning, the probability of response/remission following 8 weeks of treatment increased.

View this table:
Table 4

Detailed results of mixed-effects logistic regression analysis on the primary outcome

Independent termAdjusted odds ratio95% Confidence intervalP-valuea
Intervention main effectb1.080.38 to 3.070.88
Time main effectc1.680.61 to 4.590.31
Intervention × time interaction1.770.53 to 5.920.35
Physician professiond0.700.31 to 1.570.39
Family statuse
 Single1.270.62 to 2.600.52
 Divorced1.120.49 to 2.550.80
 Widowed0.470.18 to 1.250.13
Depression severity at beginning of treatmentf0.830.78 to 0.87<0.001
  • aLevel of significance; bintervention group (coded as 1) vs. control group (coded as 0); cfollow-up (coded as 1) vs. baseline (coded as 0); dpsychiatrist (coded as 1) vs. general practitioner (coded as 0); edummy coded, reference category is ‘married’; fas in points as measured by the Patient Health Questionnaire.

As indicated by the intraclass correlation coefficients (Table 3), the proportion of explained variance due to variation between physicians ranged from roughly 10–30% across the investigated outcomes.


In the framework of a controlled trial, an evidence-based treatment strategy was transferred to the outpatient care of depressive patients. The main objective was to test whether a practice-oriented and interdisciplinary approach to transferring guidelines into practice would enhance treatment outcome and improve physicians' procedure in the primary health care of depression. In 44 general practitioners' and psychiatrists' practices in South Germany, the treatment of 698 depressed patients was analysed.

Due to possible heterogeneity between the practices, data were analysed regarding the primary outcome (response or remission after 8 weeks) as well as the secondary outcomes by means of a mixed-effects logistic regression. Although a hint of improvement of physicians' clinical effectiveness in the intervention group was found, the cluster-adjusted analysis failed to indicate a statistically significant effect of the intervention on the treatment outcome. It is, however, possible that analyses were underpowered due to the missing consideration of power reduction as a consequence of heterogeneity between single physicians at the stage of sample size calculation. In the intervention group, ∼50% more patients reported a significant improvement in their symptoms after intervention compared with that before intervention. Moreover, the adjOR pointed in the expected direction (treatment improvement through intervention). The results are in line with studies and reviews showing that a substantial improvement in care can be achieved when using specific training sessions and target-oriented feedback [10, 12, 14, 16]. Furthermore, the results challenge other studies which were unable to find changes in clinical outcome [29].

No statistically significant improvements were found regarding the secondary outcomes (use of ICD-10 criteria, prescription of adequate antidepressant dose and referrals). However, as reported elsewhere [30], general practitioners in the intervention group showed a significantly increased rate of accurately diagnosed patients with depressive symptoms (sensitivity) after intervention. At the same time, variances in the non-depression accuracy (specificity) highlight that it is possible to improve the sensitivity of general practitioners for depression without disproportionately reducing the specificity. Considering physicians' evaluation of the training regarding its usefulness for practice and applicability to routine [20], it is likely that general practitioners stated a higher self-confidence in their own treatment behaviour compared with that before training, but may need more specific training to acquire new behavioural skills [16, 31].


Limitations result first of all from the non-randomized control design. However, the main aim of the study was to ensure high proximity to routine care regarding the feasibility of guideline implementation. Secondly, the participating physicians were volunteers. The comparison of sociodemographic data and professional skills of the participating physicians with data from outpatient care in Germany as a whole indicates that the participating physicians do not differ from the reference groups [28]. Thirdly, patients were not independently recruited, but selected by physicians. Consequently, there may be other factors that influenced the results. Finally, no interview-based symptom assessment was used. The aim of the study was to analyse guideline implementation and adherence in everyday routine practice. Clinical trials have also shown a good sensitivity for detection of change in the PHQ [26].

The implications of these results underline the importance of quality management in outpatient depression care [32]. Although there might be other useful indicators for evaluating the effect of training programmes, the most important in clinical terms is the improvement with regard to the patients' symptoms. A wider implementation seems to be possible against the background of specific and evaluated documentation systems, evidence-based clinical practice guidelines and quality assurance measures. However, to ensure a broad acceptability and to increase the effectiveness, quality assurance measures must be behaviour-specific and embedded in an integrated care programme. Multilevel strategies to enhance evidence-based treatment in the outpatient care of depressive patients require accompanying training and support to ensure the acquisition of new behavioural skills. Further work is needed to analyse the long-term changes in physicians' treatment behaviour in the outpatient care of depressive disorders.


This work was made possible by the research funding programme ‘Competence Network on Depression and Suicidality’ (CNDS), Project 3 ‘Quality management in the diagnosis and therapy of depression’ (BMBF; grant # 01 GI 9922/0222) and the project ‘Transfer of quality improvement in the care of depressive disorders – transfer of evaluated quality management practice tools into routine’ (grant # 01 GI 0452).


We thank the members of the external expert group of the CNDS project 3 and the members of the multidisciplinary focus group of the Critical Appraisal Project on Depression Guidelines.


View Abstract