Research report A target-driven collaborative care model for Major Depressive Disorder is effective in primary care in the Netherlands. A randomized clinical trial from the depression initiative Klaas M.L. Huijbregts a,b,c,n , Fransina J. de Jong a,c , Harm W.J. van Marwijk c , Aartjan T.F. Beekman d , Herman J. Ad er e , Leona Hakkaart-van Roijen f ,J¨ urgen Un ¨ utzer g , Christina M. van der Feltz-Cornelis a,h,i a Department of Diagnosis and Treatment, Netherlands Institute of Mental Health and Addiction (Trimbos Institute), Utrecht, The Netherlands b GGZinGeest, partner VU University Medical Center, Amsterdam, The Netherlands c Department of General Practice and the EMGOþ Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlands d Department of Psychiatry and the EMGOþ Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlands e Johannes van Kessel Advising, Huizen, The Netherlands f Institute for Medical Technology Assessment, Erasmus Medical Centre, Rotterdam, The Netherlands g University of Washington, Psychiatry and Behavioral Sciences, Seattle, Washington, USA h Department of Tranzo, University of Tilburg, Tilburg, The Netherlands i Clinical Center for Body, Mind and Health, Academic Psychiatry Department, GGZBreburg, Tilburg, The Netherlands article info Article history: Received 25 May 2012 Received in revised form 12 September 2012 Accepted 15 September 2012 Available online 13 October 2012 Keywords: Primary health care Depressive disorder Collaborative care Randomized clinical trial Decision aid abstract Background: Practice variation in the primary care treatment of depression may be considerable in the Netherlands, due to relatively small and unregulated practices. We adapted the collaborative care model for the treatment of Major Depressive Disorder (MDD) to accommodate existing practice variation and tested whether this had added value over Care as Usual (CAU). Methods: A cluster randomized controlled trial was conducted to compare an adapted target driven collaborative care model with Care as Usual (CAU). Randomization was at the level of 18 (sub)urban primary care centers. The care manager and GP were supported by a web-based tracking and decision aid system that advised targeted treatment actions to achieve rapid response and if possible remission, and that warned the consultant psychiatrist if such treatment advice was not followed up. Eligible patients had a score of 10 or higher on the PHQ9, and met diagnostic criteria for major depression at the subsequent MINI Neuropsychiatric interview. A total of 93 patients were identified by screening. They received either collaborative care (CC) or CAU. Another 56 patients received collaborative care after identification by the GP. The outcome measures were response to treatment (50% or greater reduction of the PHQ9-total score from baseline) at three, six, nine and twelve months, and remission (a score of 0–4 on the PHQ9 at follow-up). Results: Treatment response and remission in CAU were low. Collaborative care was more effective on achieving treatment response than CAU at three months for the total group of patients who received collaborative care [OR 5.2 ((1.41–16.09), NNT 2] and at nine months [OR 5.6 ((1.40–22.58)), NNT 3]. The effect was not statistically significant at 6 and 12 months. Limitations: A relatively high percentage of patients (36.5%) did not return one or more follow-up questionnaires. There was no evidence for selective non response. Conclusions: Our adapted target driven CC was considerably more effective than CAU for MDD in primary care in the Netherlands. The Numbers Needed To Treat (NNT) to achieve response in one additional patient were low (2–3), which suggest that introducing CC at a larger scale may be beneficial. The relatively large effects may be due to our focus on reducing practice variation through the introduction of easy to use web based tracking and decision aids. The findings are highly relevant for the application of the model in areas Contents lists available at SciVerse ScienceDirect journal homepage: www.elsevier.com/locate/jad Journal of Affective Disorders 0165-0327/$ - see front matter & 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jad.2012.09.015 n Corresponding author at: Department of Diagnosis and Treatment, Netherlands Institute of Mental Health and Addiction (Trimbos Institute), PO Box 725, 3500 AS Utrecht, The Netherlands. Tel.: þ31 30 2971185; fax: þ31 30 2971111. E-mail addresses: [email protected], [email protected](K.M.L. Huijbregts). Journal of Affective Disorders 146 (2013) 328–337
Transcript
Journal of Affective Disorders 146 (2013) 328–337
Contents lists available at SciVerse ScienceDirect
Journal of Affective Disorders
0165-03
http://d
n Corr
Institute
Utrecht
E-m
(K.M.L.
journal homepage: www.elsevier.com/locate/jad
Research report
A target-driven collaborative care model for Major DepressiveDisorder is effective in primary care in the Netherlands. A randomizedclinical trial from the depression initiative
Klaas M.L. Huijbregts a,b,c,n, Fransina J. de Jong a,c, Harm W.J. van Marwijk c, AartjanT.F. Beekman d, Herman J. Ad�er e, Leona Hakkaart-van Roijen f, Jurgen Unutzer g, ChristinaM. van der Feltz-Cornelis a,h,i
a Department of Diagnosis and Treatment, Netherlands Institute of Mental Health and Addiction (Trimbos Institute), Utrecht, The Netherlandsb GGZinGeest, partner VU University Medical Center, Amsterdam, The Netherlandsc Department of General Practice and the EMGOþ Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlandsd Department of Psychiatry and the EMGOþ Institute for Health and Care Research, VU University Medical Center, Amsterdam, The Netherlandse Johannes van Kessel Advising, Huizen, The Netherlandsf Institute for Medical Technology Assessment, Erasmus Medical Centre, Rotterdam, The Netherlandsg University of Washington, Psychiatry and Behavioral Sciences, Seattle, Washington, USAh Department of Tranzo, University of Tilburg, Tilburg, The Netherlandsi Clinical Center for Body, Mind and Health, Academic Psychiatry Department, GGZBreburg, Tilburg, The Netherlands
a r t i c l e i n f o
Article history:
Received 25 May 2012
Received in revised form
12 September 2012
Accepted 15 September 2012Available online 13 October 2012
Although there remains enormous variation across the globe,recent reports have shown that the access to treatment of affectivedisorders, such as Major Depressive Disorder (MDD), is improving inmany places in the world (Wang et al., 2007, 2005). This isimportant as numerous reports have demonstrated that depressionhas enormous public health damage (Mathers and Loncar, 2006),while being a highly treatable disorder Nierenberg (2001). Eventhough access has improved, it was established that patients withdepressive and anxiety symptoms presenting themselves in theprimary care setting with the request for treatment receive guide-line based care in less than half of cases in the NetherlandsPrins et al. (2010). The alternative, referring all patients asking fortreatment of depression to specialty mental health settings, may notbe preferable due to the heterogeneity and prevalence of depressivedisorders in the primary care setting. Some may be first timedepressive disorders, and may benefit from treatment in primarycare; some may be related to comorbid somatic disorders (Nuttinget al., 2000) that may be treated by the GP as well, allowing forintegrated treatment in the primary care setting; (Neumeyer-Gromen et al., 2004) some depressive disorders may be treatmentresistant, recurrent or chronic (Spijker et al., 2002). Establishingeffective treatment models in the primary care setting for at leastsome of these patients, and establishing clear rules for referral ofpatients who may not benefit from treatment in the primary caresetting, is needed to solve this problem. A disease managementapproach for depression, such as the collaborative care model maybe suitable for this Neumeyer-Gromen et al. (2004).
A well studied collaborative care model is the IMPACT-protocol in which a depression care manager (DCM) collaborateswith a GP and a liaison psychiatrist in order to provide depressiontreatment in primary care Unutzer et al. (2002). Effectiveness hasbeen established in the US (Unutzer et al., 2002; Bower et al.,2006), mostly within a combined insurance company and healthcare provider setting such as Kaiser Permanente (KP). Collabora-tive care has so far been researched most extensively in countrieswith central steering and relatively large GP practice sizes. In theUK, with its nationalized health service and a larger averagenumber of GPs per primary care practice than for instance inthe Netherlands, France, and Germany (Schoen et al., 2009), themodel was effective as well Richards et al., (2008). In Chile, wherecollaborative care was also found to be effective (Araya et al.,2003), practices are even larger with up to 60 physicians perpractice Jara (2011). In the USA, Chile and the UK, primary carepractices are also generally centrally organized units that arerelatively large, and where some form of central regulation interms of availability of treatment and reimbursement exists. Inthe Netherlands, every primary care practice is a small businessunit with its own culture and rules. The number of GPs in onepractice ranges from 1 to 5, 2 GPs on average (Karssen et al., 2009;Hingstman and Kenens, 2010), and the patients in one practiceoften receive reimbursement from a variety of health insurancecompanies, that have different reimbursement policies in termsof depression treatment. In the USA, an ongoing discussion iswhether the IMPACT collaborative care model could be applied insuch small, individual practices, which in the USA are mostlyfound in rural areas. The Dutch healthcare system provides anopportunity to evaluate these issues in mixed healthcare systems,where there is no single ownership of input and health and costoutcomes.
For this purpose, we adapted the collaborative care model forapplication in such a system. We maintained the architecture of theIMPACT CC model: a DCM, GP and a liaison psychiatrist; and the useof Problem Solving Treatment (PST) as intervention provided by theDCM. However, we made the CC model target driven, that is, aimed atfast treatment response or if possible remission in 18–24 weeks, asindicated by the PHQ9 score. An efficient way to enhancethe efforts of health care professionals in small scale primary carepractices in a target-driven collaborative care model, may be the useof a decision aid that facilitates close following of a clearly definedstepped care protocol. Because in the Netherlands, primary carepractices are numerous and because they use different electronicpatient file systems that may not be compatible, this patient trackingsystem decision aid was made web-based and could be accessed fromany practice on a secured website. The web-based decision aid mighttherefore not only be an important facilitator for the implementationof collaborative care in the Netherlands, but it might also be a highlyrelevant innovation with respect to the implementation of collabora-tive care in rural areas with small practices (such as, for instance, inthe United States). Furthermore, providing decision support by way ofa detailed algorithm might be an efficient way of reducing practicevariation, which is likely to be disruptive in situations where there aresmall practices, operating without close cooperation or control—suchas primary care in the Netherlands.
A stepwise algorithm was incorporated in the decision aid. Notonly was the progress of the patient monitored, but also theadherence of the health care professional to the treatment protocol.Advice was provided on how and when to take a next treatmentstep, or to start or switch antidepressants, in order to improveadherence to antidepressant treatment by GPs in this decentrallyorganized primary care setting. Also, the web-based tracking systemcontains a protocol for handling suicidality as assessed by item 9 ofthe PHQ9, in order to enhance patient safety. It monitors if treatmentin the primary care setting can be expected to still provide progress,or if referral to the specialty mental health setting is needed, andwhen a psychiatrist should be consulted. Consultation by a psychia-trist occurs in cases where this decision aid does seem too rigid forthe particular patient, i.e. in case of comorbidity. Another reason forpsychiatric consultation would be non-adherence to the treatmentprotocol. The decision aid signals if a prescribed action was not takenby the health care professional, and the consultant psychiatristwould in that case receive an email that this was the case, andcontact the professional and provide consultation advice.
The primary objective of this paper was to establish effectivenessof this target driven collaborative care model for MDD with a web-based decision aid, and consultant psychiatrist availability if treat-ment was not followed as intended, in the mixed healthcare settingin the Netherlands with small primary care practices, compared tousual care.
2. Methods
The design of the study has been described in more detailelsewhere, but is summarized below IJff et al. (2007). The researchproject this manuscript is based on was approved by the medicalethical committee of the VU medical center in Amsterdam (protocolno. 2006/158) and the trial was registered in the Dutch trial register(ISRCTN15266438).
Using cluster randomization (Van der Feltz-Cornelis and Ader,2000), 18 primary care centers (with a total of 82 GPs) were
K.M.L. Huijbregts et al. / Journal of Affective Disorders 146 (2013) 328–337330
randomly assigned to either the collaborative care (CC) or the careas usual condition (CAU) by an independent statistician using acomputer algorithm for allocation. We chose to deviate from ourinitial method of patient randomization, as described in the designpaper (IJff et al., 2007), in order to prevent contamination of theeffect Van der Feltz-Cornelis and Ader (2000). A study by Richardset al. showed that there is a substantial risk of spill over of the effectto the control group when the effect of collaborative care isassessed in a patient randomized design Richards et al. (2008).
GPs in Primary care centers randomized to the CC conditionreceived training in the collaborative care model and the use ofthe web-based tracking system and got acquainted with theconsultant psychiatrist. Patients of the respective practices couldenter the trial in two ways: either by screening or after identifica-tion by their GP. Screening was done as follows: all patients whohad consulted their GP during the past 6 months received thePHQ9 (Kroenke et al., 2001) by mail, regardless of the reason forprior consultation. The reason for this broad approach is thatliterature suggests that patients suffering from Major DepressiveDisorder (MDD) are more likely to visit their GP with miscella-neous somatic symptoms instead of the core symptoms of MDD(Nutting et al., 2000; Simon et al., 1999).
After informed consent, patients were asked to return thequestionnaire regardless of whether or not they still had anysymptoms. At the moment of screening, patients were unaware oftheir allocation. Inclusion criteria were age 417 years, PHQ9score Z10 and a classification of MDD according to the MINIneuropsychiatric interview (Sheehan et al., 1998), which wasadministered by telephone. The MINI interview was applied toavoid false positive diagnoses of MDD based on the PHQ9-score.Exclusion criteria were high risk of suicide, psychosis, dementia,drug or alcohol dependence, already being under specialty mentalhealth treatment, or insufficient knowledge of Dutch to fill in thequestionnaires.
3. Outcome measures
Outcome measures were assessed by self report questionnairethat was sent to the participants by mail three (T1), six (T2), nine(T3), and twelve (T4) months after their inclusion in the trial. Theprimary predefined outcome measure was whether or not apatient had reached a clinically relevant response on the PHQ9Kroenke et al. (2001). The PHQ9 consists of 9 items referring tothe DSM IV criteria for MDD. Each item is scored from 0 (not atall) to 3 (nearly every day). The total score thus varies from 0 to27. A score of 10 or higher indicates depressive symptoms of atleast moderate severity Kroenke et al. (2001). A decrease of atleast 50% on the PHQ9 compared with baseline on a follow upquestionnaire was defined as a clinically relevant response totreatment. The secondary predefined outcome was remission,defined as a score of lower than 5 on the PHQ9 Kroenke et al.(2001).
4. Intervention
4.1. Target driven collaborative care
The target driven collaborative care protocol used during thisstudy has been described in detail elsewhere, (IJff et al., 2007) andis also described in the Appendix. Target was to achieve remission(PHQ-9o5) within 18–24 weeks of treatment. If this was notachieved, referral to specialty mental health care, was advised. Incase of suicidality, a suicidality protocol was followed, includingconsultation by the psychiatrist if needed. An antidepressant
algorithm was provided by the decision aid tracking system, asdescribed in the Appendix. The specific tasks of the DepressionCare Manager (DCM), the GP, and the consultant psychiatrist areexplained in the Appendix as well.
4.2. Usual care
Patients recruited in the usual care practices were informed asto their diagnosis and advised to seek treatment from their GP.There were no restrictions to treatment in any way. The GPs in thecontrol condition were not informed about the presence of MDDin screened patients. Treatment in the usual care group wasmonitored during the course of the study.
5. Statistical analyses
Multi Level Analysis (MLA) was performed using MLwiN 2.0 mul-tilevel software (Goldstein et al., 1998) on the basis of the intentionto treat analysis. Propensity scores were calculated to correct forselection bias due to the cluster randomization Ader et al. (2008). Theamount of variance in the effect was established at the level ofprimary care centers, GPs and patients. Multilevel logistic regressionanalysis was used to assess the treatment response as well as theremission outcome at each administration after baseline. The effectof the intervention for the collaborative care group as a whole(CCtotal) was compared to CAU. We also estimated the effect for thescreened collaborative care group compared to CAU (CCscreen vs.CAU), because these patients were included the same way (afterscreening instead of selection by GP). Based on this model, wecalculated odds ratios and Numbers Needed to Treat (NNT) regardingthe primary outcome measure ‘response to treatment’. The NumberNeeded to Treat can be defined as the number of patients that haveto be treated according to a new intervention (in this case collabora-tive care) compared to care as usual in order to achieve a clinicallyrelevant outcome (in this case response to treatment) for one extrapatient.
A composite score was computed based on the first time pointto achieve treatment response. Differences between treatmentgroups on this score were also analyzed using MultiLevel AnalysisAder et al. (2008); Ader, (2012).
Results are presented according to the CONSORT statement forcluster randomized trials Schulz et al. (2010).
6. Study oversight
This RCT was part of the Depression Initiative, a nationalinitiative to improve depression management in the Netherlands(De Jong et al., 2009; van der Feltz-Cornelis 2009). The studyprogress was monitored by a steering group and advisory boardon a 3 monthly basis.
7. Results
7.1. Participant flow
The inclusion phase of the trial started in September 2007 andended in August 2009. 18 Primary care clinics agreed to partici-pate in the trial. They were situated both in major urban areassuch as Amsterdam and Haarlem and in smaller urban areas andmore rural communities. 150 patients gave informed consent forthe treatment phase, 49 patients participated in the CAU condi-tion and 101 participated in one of the two collaborative caregroups: 45 in the screened collaborative care group (CCscr) and101 in the total collaborative care group (CCtotal, the 45 patients
26.580 patients approached with depression screener(patients blinded for treatment allocation)
6.500 patients returned the screener (24.5%)
PHQ9 ≥ 10; N=917 (14.1%)
MINI-interview (N=627) Diagnosis of MDD40.0% (N=251)
9 primary care centers randomly allocatedto CC
Patients who received CC after identi-fication by their GP: N=56 (actually re-ceived the intervention: N=45;80%)
CCscr: N=45(actually received allocatedintervention: N=36;80%)
CAU: N=49 (1 by the GP)(actually received allocatedintervention: N=49;100%)
3 months: N=30 (53.6%)Loss to follow-up: N=26Reasons:Did not return questionnaire thistime:16Drop out:10
3 months: N=31 (68.9%)Loss to follow-up: N=19Reasons:Did not return questionnairethis time:14Drop out:0
3 months: N=38 (77.6%)Loss to follow-up: N=11Reasons:Did not return questionnaire thistime:8Drop out:3
100 Patients were identifiedby their GP. They also
received the PHQ9
Randomised: 18 primary care centres with82 GPs who agreed to participate in the
trial
6 months: N=32 (57.1%)Loss to follow-up: N=24Reasons:Did not return questionnaire thistime:10Cumulative drop out:14
6 months: N=31 (68.9%)Loss to follow-up: N=14Reasons:Did not return questionnairethis time:12Cumulative drop out:2
6 months: N=39 (79.6%)Loss to follow-up: N=10Reasons:Did not return questionnaire thistime:6Cumulative drop out:4
9 months: N=31 (55.4%)Loss to follow-up: N=25Reasons:Did not return questionnaire thistime:4Cumulative drop out:21
9 months: N=28 (62.2%)Loss to follow-up: N=17Reasons:Did not return questionnairethis time:10Cumulative drop out:7
9 months: N=31 (63.3%)Loss to follow-up: N=18Reasons:Did not return questionnaire thistime:10Cumulative drop out:8
12 months: N=25 (44.6%)Loss to follow-up: N=31Reasons:Cumulative drop out:31
12 months: N=33 (73.3%)Loss to follow-up: N=12Reasons:Cumulative drop out:12
12 months: N=32 (65.3%)Loss to follow-up: N=17Reasons:Cumulative drop out:17
Excluded (for all 3 groups):-No MDD: 364-Severe mental health problems: 8-Primary substance dependence: 12-Already receiving treatment in specialized mental health care: 77-Refused treatment: 61-Language problemsduring MINI: 8-Other reason: 25 (for instancemoved to another city)
Fig. 1. Flowchart of participants in the trial in accordance with consort statement (Schulz et al., 2010).
K.M.L. Huijbregts et al. / Journal of Affective Disorders 146 (2013) 328–337 331
in the CCscreen group plus the 56 patients who received colla-borative care after identification by their GP). Only one patientwas included by the GP in the CAU group. Inclusion and trial flowaccording to the cluster-randomized CONSORT statement (Schulzet al., 2010) is shown in Fig. 1. The 56 patients who received
collaborative care after identification by their GP were treated asa separate group in the flowchart.
Of the 82 GPs that agreed to participate in the trial, 50included patients in the final cohort. 18 GPs only included onepatient. As mentioned in the method section we applied Multi
Table 1Characteristics of the patients in the collaborative care groups, and in the care as usual group.
Characteristics Care as usual(CAU; N¼49)
Collaborative care, total group(CC total; N¼101)
Collaborative care screened(CCscr; N¼45)
p-value CAU vs.CCtotal
Age (sd) 52.1 (14.8) 47.0 (13.5) 52.0 (13.0) .023*
Gender (% female) 73.5 72.3 64.4 .979 (Chisq)
Living alone (%) 50.9 49.5 53.3 .773 (Chisq)
Non Dutch origin(%) 25.0 30.0 22.7 .528 (Chisq)
Level of education** 5.2 (2.9) 5.7 (2.6) 5.4 (2.5) .304
Prior episode of depression (%) 56.2 56.2 58.5 .994 (Chisq)
Chisq¼w 2 test.n Statistically significant: po .05.nn 5 corresponds to only secondary school; 6 corresponds to a few years of education following secondary school.
Table 2Results for response to treatment and remission.
Outcome variable CAU CCtotal CCscreen
Three months (T1)
N 38 61 31
Response to treatment (%) T1 10.5% 45.9% 41.9%
OR Ref. Cat. 5.20* 4.59*
(95% CI) (1.41–16.09) (1.15–18.34)
NNT (only when Po .05) 2 2
Remission (%) T1 7.9% 26.2% 16.1%
OR Ref. Cat. 2.53 1.73
(95% CI) (.61–10.48) (.32–9.46)
Six months (T2)
N 39 63 31
Response to treatment (%) T2 25.6% 39.7% 29.0%
OR Ref. Cat. 1.99 1.27
(95% CI) (.70–5.67) (.38–4.17)
Remission (%) T2 10.3% 23.8% 19.4%
OR Ref. Cat. 3.91* 2.87
(95% CI) (1.01–15.17) (.42–19.83)
Nine months (T3)
N 31 59 28
Response to treatment (%) T3 25.8% 61.0% 50.0%
OR Ref. Cat. 5.62* 4.06
(95% CI) (1.40–22.58) (.92–17.80)
NNT (only when Po .05) 3
Remission (%) T3 12.9% 37.3% 32.1%
OR Ref. Cat. 3.67 3.04
(95% CI) (.96–14.04) (.71–13.00)
Twelve months (T4)
N 32 58 33
Response to treatment (%) T4 25.0% 39.7% 33.3%
OR Ref. Cat. 1.82 1.47
(95% CI) (.61–5.42) (.45–4.81)
Remission (%) T4 6.3% 20.7% 15.2%
OR Ref. Cat. 3.38 2.07
(95% CI) (.62–18.49) (.32–13.18)
CAU¼Care as usual; CI¼Confidence interval; CCtotal¼Collaborative care, total
group; CCscreen¼Collaborative care, indentified through screening; OR¼Odds
Ratio; NNT¼Number Needed to Treat; Ref. cat.¼Reference category.n po .05.
K.M.L. Huijbregts et al. / Journal of Affective Disorders 146 (2013) 328–337332
Level Analysis (MLA) to correct for any disturbances in the datadue to cluster effects. This method provides correct estimateseven if the patient-to-site-ratio is small. Moreover the IntraCluster Correlation (ICC), a measure for the variance in outcomesexplained by the different levels, was .00 at both the level ofHealth Care Centers and at the level of individual GPs. This meansthat the level of variation that can be attributed to Health carecenters and GPs is very low.
The average loss to follow up percentage (patients who did notreturn a questionnaire at one of the time points) was 36.5%.
Patients who dropped out three and six months after inclusiondid not have higher PHQ9-scores at baseline (T1, p¼ .799; T2,p¼ .577). Patients who dropped out at nine and twelve monthsafter inclusion had on average a higher PHQ9-score at baseline,but this trend was the same for the patients in the collaborativecare group and the patients who received CAU (T3, p¼ .575; T4,p¼ .547). This implies that there is no evidence for selectivedropout in the trial, and because of this we decided not to imputemissing data. Multilevel analysis has been shown to be veryflexible when handling missing data. Multilevel analysis on anincomplete dataset might even be preferable compared withimputation methods Twisk (2006).
7.2. Baseline data
Baseline characteristics and PHQ9 scores are displayed inTable 1 and 2.
Most variables did not differ significantly between the fourgroups at baseline, except age. The patients in the CCtotal groupwere younger (p¼ .023). The patients in the CCscr group did notdiffer from the patients who received CAU with respect to anyof the baseline variables. Adjusting the models for age did notinfluence the results.
8. Response to treatment and remission
Treatment response for CAU was low at first (10.5%), and tooklonger to come into effect than for CC. It increased at follow-up toa maximum of 25.8% at nine months. The response percentageswere higher in both collaborative care groups. After twelvemonths, the effect of both care as usual and collaborative caresubsided to a certain extent. At 3 months both the screened(CCscreen) and the total group of patients who received colla-borative care (CCtotal) had a strong and statistically significantfavourable odds ratio on response to treatment: 5.20 (CI: 1.41–16.09) for CCtotal and 4.59 (CI: 1.15–18.34) for CCscreen. Thestrongest effect of collaborative care on response to treatmentcompared with CAU was found after nine months, with astatistically significant favourable odds ratio on response of 5.62(CI: 1.40–22.58) for the CCtotal group compared with CAU.Numbers Needed to Treat (NNT) calculated for the responsepercentages were 2 for the total CC group at 3 months and 3 forthe total CC group at nine months. In the screened patients whoshowed response, the difference in time to first response betweenCCscr versus CAU was three months (4.4 versus 7.4 months).
Striking are the low remission percentages in CAU. The max-imum remission percentage in CAU was 12.9% after nine months.The strongest effect of collaborative care on remission compared
Table 3Actual care given.
Care as usual Collaborative care, total group Collaborative care screened
Mean number
contacts
Percentage of
people (%)
Mean number
contacts
Percentage of
people (%)
Mean number
contacts
Percentage of
people (%)
GP 4.7(4.9) 85.7 4.8 (4.4) 76.2 5.3 (4.9) 83.7
Mental Health Care Institute 3.2 (11.7) 26.5 1.8 (4.4) 23.8 1.3 (2.9) 23.3
Counselling center for drugs alcohol .1 (.4) 2.0 .0 .0 .0 .0
Admission to (parttime) psychiatric day care 1.9 (13.2) 4.1 .06 (.5) 2.0 .02 (.2) 2.3
Admission to a psychiatric hospital .6 (2.2) 14.3 2.3 (22.4) 3.0 5.4 (34.3) 4.7
Psychiatric nurse No information – 1.7 (3.2) 35.7 2.0 (3.7) 39.0
Mental health care practice nurse .1 (.6) 2.0 .8 (2.4) 12.9 1.0 (2.5) 15.6
Medication (not necessarily an
antidepressant)
– 79.6 – 73.3 – 88.4
K.M.L. Huijbregts et al. / Journal of Affective Disorders 146 (2013) 328–337 333
with CAU was found after six months, with a statistically significantfavourable odds ratio on remission of 3.9 for CCtotal compared withCAU. As is shown in Table 3 the patients in the CAU-group didreceive care. It is therefore not likely that absence of care was animportant reason for the low remission percentages in this group.We can also provide some additional data on adherence to thecollaborative care protocol by the caremanagers. 90.9% of thepatients in the total collaborative care group visited the caremana-ger at least once. The average number of visits was 5.8 (sd 3.6). Thepsychiatrist was consulted for 8.8% of the patients. During the firstweeks of the intervention an antidepressant was prescribed (orprescription was continued) to 26.4% of the patients.
9. Discussion
This trial shows that our adapted target driven collaborative care(CC) model for the treatment of Major Depression in primary care iseffective. We adapted the CC model to cope with practice variationthat is inherent in a mixed health care model with small primarycare practices, such as in the Netherlands. CC was most effective inthe short term: the OR for response to treatment at three monthswas about 5, and the Number Needed to Treat (NNT) was 2–3. Thestrongest effect of collaborative care on response to treatmentcompared with CAU was found after nine months, with a statisti-cally significant favourable odds ratio for response to treatment of5.6 (CI: 1.40–22.58) for the collaborative care group in totalcompared with CAU. The average time to first response was threemonths shorter in the screened CC group compared to CAU.
The effect of collaborative care, particularly for the screenedgroup, subsided to a certain extent after twelve months, whichmight be explained by the fact that the intensity of the interventionwas toned down at this point, and that the care as usual groupstarted to catch up, although without surpassing the interventiongroup at any point. A stronger focus on maintaining response andremission may be needed in future research, e.g. a stronger focus onrelapse prevention in the collaborative care algorithm.
It is striking that treatment response and remission in the CAUgroup are very low (the response percentages ranged between10.5% and 25.8%). The results are comparable though to those inthe IMPACT-study where treatment response in the CAU-groupranged between 14.8% and 30.9% Unutzer et al. (2002). In ourcurrent study, ‘depression’ may not have been on the patients’agenda in their contacts with primary care in the CAU-condition,as these cases were detected by means of screening (although thepatients in the CAU-group may have been more motivated to seektreatment because they knew they were suffering from MDD).
In the time restricted primary care setting GPs and patients areoften confronted with competing demands, e.g. hypertension ordiabetes may ‘take away’ the focus from depression Nutting et al.(2000). Moreover, patients’ perceived needs for care are alsostrongly associated with the delivery of guideline-concordantcare for depression Prins et al. (2010). Hence, care for depressionmight not have been high enough on the patients’ agenda duringtheir appointments with their GPs (but this probably does notcorrespond to a major difference between care as usual in thisstudy and care as usual in everyday care). Some of the patientswho were included in the trial had been taking an antidepressantin the same dosage for years, without any monitoring or adjust-ment. In the collaborative care group monitoring and contactingpatients was a task for the caremanager, which may havediminished the burden on the GP.
In our collaborative care model, more emphasis is put onfollowing a precise algorithm and on treatment adherence andmonitoring of the health care professional by the tracking system(decision aid). This is a new aspect in our current study that mayfacilitate implementation of target driven collaborative care in amixed health care system and in a variety of primary carepractices. It may also reduce practice variation, which is inherentto a primary care setting with relatively small practices actingindependently to a large extent. Implementation efforts withregard to the webbased decision aid are needed though, in orderto facilitate the uptake of this promising tool in everyday care.GPs in our study indicated that it would be easier for them to usethe decision aid if compatibility with their current electronicpatient files could be improved Jellema and Licht (2009).
10. Limitations
An important limitation of the current study is the relativelyhigh percentage of patients who did not return one or morefollow-up questionnaires. However, no indication for selectivenon-response exists.
In this cluster randomized trial, the screened patients wereblinded for the intervention and thus the comparison betweenscreened collaborative care versus screened CAU patients can beconsidered nonbiased. Obviously the findings for the group ofpatients who were identified by their GP are hard to interpret, asonly one patient was included in the trial by the GP in the CAUgroup. And thus there were almost no GP identified patients inthe CAU group to compare the GP identified CC patients with. Themore positive findings for the group of patients identified for CCby their GP may have to do with more motivation for treatment
K.M.L. Huijbregts et al. / Journal of Affective Disorders 146 (2013) 328–337334
and with the capability of the GPs to select a group of patients forwhom collaborative care is most likely to be beneficial. In anycase, that the GP identified CC group outperformed the screenedCC group is a promising result because these patients may closelyresemble the patients GPs would select in everyday care. Theresults for the CCtotal group (in which the GP identified patientswere included) imply that collaborative care may also be effectivein every day care, because the patients who were identified bytheir GP may closely resemble the patients they would refer forcollaborative care in everyday practice. These patients have a‘perceived need for care’ which might be more strongly associatedwith successful treatment than a positive score on a screener.More research into this is necessary. In any case, we werealso able to compare a screened collaborative care group to ascreened CAU group in the current trial, which allowed for a validcomparison.
Another limitation might be the strikingly low remissionpercentages in CAU. One could argue that this is due to recruit-ment bias, with patients in the CAU group told that they aredepressed and that it was advisable to seek treatment, whereaspatients in the CC group were contacted by the caremanager.However, this is not supported by the finding that when weexcluded the CAU-patients who were not seen by their GP duringthe first three months of the intervention (39.4% in CAU asopposed to 21.8% in CC), the results for CAU did not improve:10.0% of the CAU-patients who were seen by their GP achievedresponse, as opposed to 15.4% of those were not seen. The resultsremained strongly in favour of collaborative care (Po .01).
11. Implications for research and clinical practice
We conclude that target driven collaborative care for majordepression in small primary care practices in a mixed healthcaresystem such as in the Netherlands can be very effective. The web-based tracking system and the precise decision aid based on analgorithm with follow up by the consultant psychiatrist, is apromising development in that respect. This way, an algorithm iscombined with expert advice in case the algorithm is insufficientfor a particular patient. The uptake of the decision aid by GPs couldstill be improved, which may even increase the effectivenessJellema and Licht (2009). Future research efforts might furthermorefocus on expanding the model to other mental health problems,fine-tuning the model and on the identification of subgroups ofpatients for whom the intervention is less effective or particularlyeffective. What, for instance, are the outcomes for patients sufferingfrom comorbid medical disorders and for patients suffering con-comitant medically unexplained physical symptoms? Recent stu-dies have shown that this is an important issue Huijbregts et al.(2010; Huijbregts et al., 2010).
The target driven collaborative care model with a web-baseddecision aid may be particularly useful for implementation in i.e.rural areas or areas with a large percentage of small primary carepractices where collaborative care was so far considered not to befeasible, and in other mixed healthcare systems where clinicalpractice and reimbursement of treatment are not in one hand.Collaborative care may help to meet the need for effective,efficient and easy to use interventions to treat large numbers ofpatients with MDD in primary care. In places with a welldeveloped, but relatively unregulated primary care, such as manyparts of Europe, our adapted target driven model may be fruitfullyintroduced.
Role of funding sourceThe Innovation fund health care insurers (Innovatiefonds zorgverzekeraars,
formerly known as RVVZ: http://www.rvvz.nl/) in the Netherlands funded the
research project this manuscript is based on. The Innovation fund health care
insurers did not state any limitations regarding the methods of the research
project or regarding publication of the results.
Conflict of interestAll authors declare that they received no support from any organization for
the submitted work except for The Innovation fund health care insurers (Innova-
tiefonds zorgverzekeraars, formerly known as RVVZ) in the Netherlands that
funded the research project this manuscript is based on; no financial relationships
with any organization that might have an interest in the submitted work in the
previous three years, no other relationships or activities that could appear to have
influenced the submitted work. The Innovation fund health care insurers did not
state any limitations regarding the methods of the research project or regarding
publication of the results.
This is also stated in our cover letter that was signed by all authors.
AcknowledgementWe thank Prof. Dr. Richard Roberts, Family Physician, Wisconsin University
and President of WONCA, for his useful comments on this article.
Appendix A. Algorithm for target driven collaborative care
Target was to achieve remission (PHQ-9o5). If this was notachieved within 18 to 24 weeks of treatment, referral to specialtymental health care was advised. Subtargets were to achieve adecrease of at least 5 points on the PHQ9 within each 6 weeks, ina succession of 18–24 weeks (the target was remission if a step inthe treatment algorithm had already resulted in a decrease of atleast 5 points after six weeks). If this was not the case, a switchwould be advised to a more intensive treatment step, eitherpsychotherapy or antidepressant medication, as indicated in Fig. 1of this Appendix. In case of suicidality, a suicidality protocol wasfollowed, and also an antidepressant algorithm was monitored bythe decision aid tracking system, as described below. Monitoring oftreatment progress and providing Problem Solving Treatment (PST)were the tasks of the Depression Care Manager (DCM), assisted bythe decision aid. He or she discussed progress with the patients andthe GP. The GP was responsible for treatment with antidepressantmedication and the drawing up of a treatment plan together withthe patient and the DCM. In case of adverse events, nonadherenceor suicidality a consultant psychiatrist was to be consulted.
The DCM, the GP, and the consultant psychiatrist had access tothe online patient tracking system that contained the decision aidthat would indicate—based on the progress of the patientmeasured with the PHQ9-score—if a subsequent treatment stepor psychiatric consultation was deemed necessary. Also, thetracking system could send out a signal to the consultantpsychiatrist if proposed algorithm steps (such as the suicidalityprotocol and an agreed on treatment plan) were not followed,allowing the consultant psychiatrist to contact the DCM or GP forexploration of the problem and advice.
The process of the recruiting and the training of caremanagers forour study, as well as the issue of the impact of either approach (CAU orCC) on the organization, staffing, and resources of the GP’s practices,has been decribed extensively elsewhere De Jong et al. (2009).
Suicidality protocol
Every time the PHQ9 was assessed, suicidality was assessed aswell by a subprotocol in case of a score ofZ2 in the 9th PHQ9 item,as shown in the box in Fig. A1 of the appendix. In caseof suicidality according to this item, the depression care manager(DCM) was instructed to contact the GP and ask for assessment andmanagement of the suicidality by the GP. If the DCM would not actaccordingly within 24 h, the tracking system would alert theconsultant psychiatrist who would contact the DCM at once in
K.M.L. Huijbregts et al. / Journal of Affective Disorders 146 (2013) 328–337 335
order to perform psychiatric consultation and give advice to DCMand GP concerning the suicidality. If the GP assessed that thesuicidality was indeed of high risk, urgent referral to crisis specialtymental health care was advised. In case of low suicide risk, thepatient could proceed with treatment in the primary care practice.
Antidepressant protocol
The antidepressant algorithm was in accordance with theDutch guidelines for depression (Van Marwijk et al., 2003;Trimbos-instituut., 2005), but it provided more specific
K.M.L. Huijbregts et al. / Journal of Affective Disorders 146 (2013) 328–337336
information about dosage and management of three antidepres-sants, from which the GP could make a choice. The antidepressantprotocol was based on the following principles:
Principles
For the on-target collaborative care intervention an antide-pressant algorithm was developed that fitted in the nationalguidelines for depression treatment (Van Marwijk et al., 2003;Trimbos-instituut, 2005) but that gave more specific instructionsto GPs and the DCMs for treatment and follow up based onmonitoring with the PHQ9. It included a choice of antidepressantsthat could be chosen and recommendations for switching if after6 weeks the medication did not produce enough improvement onthe PHQ9. This 6 week evaluation was repeated every 6 weeksuntil 18–24 weeks.
Principle for treatment was that pharmacologic treatment ofthe depression should include treatment with an antidepressantrather than with a benzodiazepine. The GP should choose anantidepressant together with the patient. As correct dosage wasdeemed important, the GP received detailed suggestions fordosage for three antidepressants to choose from. The threeantidepressants mentioned below were selected as
�
Bo
they could be prescribed on a generic basis;
� there was no patent of a pharmaceutical company on the
suggested antidepressant;
� the antidepressant was used frequently in the primary care
practice setting;
� the antidepressant was easy to start and easy to stop without
adverse effects;
� and should be applicable for the elderly as well; � the medication should be safe in case of frequently occurring
somatic comorbidity such as diabetes and cardiovasculardisease.
According to these principles, as first step an SSRI was advisedthat was safe in case of somatic cardiovascular comorbidity and inthe elderly. As second step a more broad spectrum antidepres-sant, namely a SNRI, was advised, or a TCA that was well toleratedin the elderly. The algorithm is shown in Box 1.
x 1–Antidepressant algorithm.
Step 1: Citalopram, 1st week 20 mg, second week 40 mg dailyresponse of at least 5 points drop in PHQ9 score after
6 weeks?Yes: continuation of treatmentNo: step 2Step 2: Venlafaxine,1st week 75 mg, Second week 150 mg,response of at least 5 points drop in PHQ9 score after 12
weeks?Yes: continuation of treatmentNo: step 34th week possibility to increase to 225 mg daily.Alternative step 2: Nortriptyline,1st week 50 mg2nd week 100 mg3rd week 150 mg dailyTarget was remission (PHQ9 score o5) after 12 weeks of
an antidepressant that produced response.After 18–24 weeks, treatment was only continued in
primary care if optimization in primary care was deemedpossible after consultation by a psychiatrist.
References
Araya, R., Rojas, G., Fritsch, R., et al., 2003. Treating depression in primary care inlow-income women in Santiago, Chile: a randomised controlled trial. Lancet361, 995–1000.
Ader, H.J., Mellenbergh, G.J., Hand, D.J., 2008. Advising on Research Methods:A Consultant’s Companion. Johannes van Kessel Publishing, Huizen, theNetherlands.
Ader, H., 2012. A composite score that combines response and remission. Availablefrom: /http://www.jvank.nl/Trimbos/Remres/Composite.pdfS.
Bower, P., Gilbody, S., Richards, D., Fletcher, J., Sutton, A., 2006. Collaborative carefor depression in primary care: Making sense of a complex intervention:systematic review and meta-regression. British Journal of Psychiatry 189,484–493.
C.B.O., Trimbos-instituut, 2005. Multidisciplinaire Richtlijn Depressie. (Multidisci-plinairy Guideline Depressive disorder) In Dutch]. Trimbos-insituut; Utrecht,the Netherlands (in Dutch).
De Jong, F.J., van Steenbergen-Weijenburg, K.M., Huijbregts, K.M., et al. 2009. TheDepression Initiative. Description of a collaborative care model for depressionand of the factors influencing its implementation in the primary care setting inthe Netherlands. Projects and Developments. International Journal of Inte-grated Care; ISSN 1568–4156.
Goldstein, H., Rasbash, J., Plewis, I., et al., 1998. A Users Guide to MlwinN.Multilevel Models Project. University of London, London.
Hingstman, L., Kenens, R.J.,2010. Cijfers uit de registratie van huisartsen (Figures fromthe general practice registration). Nivel; Utrecht, the Netherlands (in Dutch).
Huijbregts, K.M., van der Feltz-Cornelis, C.M., van Marwijk, H.W., de Jong, F.J., vander Windt, D.A., Beekman, A.T., 2010. Negative association of concomitantphysical symptoms with the course of major depressive disorder: a systematicreview. Journal of Psychosomatic Research 68, 511–519.
Huijbregts, K.M., van Marwijk, H.W., de Jong, F.J., Schreuders, B., Beekman, A.T.,van der Feltz-Cornelis, C.M., 2010. Adverse effects of multiple physicalsymptoms on the course of depressive and anxiety symptoms in primarycare. Psychotherapy and Psychosomatics 79, 389–391.
IJff, M.A., Huijbregts, K.M., van Marwijk, H.W., et al., 2007. Cost-effectiveness ofcollaborative care including PST and an antidepressant treatment algorithmfor the treatment of major depressive disorder in primary care; a randomisedclinical trial. BMC Health Services Research 7, 34.
Jara C.G.P., 2011. Practice size in Chile. Personal communication by Carolina Jara,Family Doctor Medical Staff East Metropolitan Health Services and AssistantProfessor University of Chile, Santiago de Chile. Oral communication WHOConference on Connecting Health and Labour, 29th November–1st December2011, The Hague, The Netherlands.
Jellema, J.L., Licht, E., 2009. Barri�eres voor invoering van een collaborative careprogramma voor depressie in de huisartsenpraktijk. Een kwalitatief onderzoek(Barriers with regard to the implementation of a collaborative care pro-gramme for depression in general practice centers. A qualitative study)EMGO-Instituut VUMC; Amsterdam, the Netherlands, (in Dutch).
Karssen B., Schipper M., Jurling B., 2009. Praktijkkosten en opbrengsten vanhuisartsenpraktijken, eindrapportage van een onderzoek naar de praktijkkos-ten en opbrengsten van huisartsenpraktijken in Nederland in 2006. (Costs andturnover of general practice centers, final report of a study into the costs andturnover of general practice centers in the Netherlands in 2006), NederlandseZorgautoriteit (NZA)/Significant; Barneveld, the Netherlands, (in Dutch).
Kroenke, K., Spitzer, R.L., Williams, J.B., 2001. The PHQ-9: validity of a briefdepression severity measure. Journal of general internal medicine : officialjournal of the Society for Research and Education in Primary Care InternalMedicine 16, 606–613.
Mathers, C.D., Loncar, D., 2006. Projections of Global Mortality and burden ofDisease from 2002 to 2030. Plos Medicine 3, 2011–2029.
Nierenberg, A.A., 2001. Current perspectives on the diagnosis and treatmentof major depressive disorder. The American journal of managed care 7, S355–S366.
Nutting, P.A., Rost, K., Smith, J., Werner, J.J., Elliot, C., 2000. Competing demandsfrom physical problems: effect on initiating and completing depression careover 6 months. Archives of family medicine 9, 1059–1064.
Neumeyer-Gromen, A., Lampert, T., Stark, K., Kallischnigg, G., 2004. Diseasemanagement programs for depression: a systematic review and meta-analysis of randomized controlled trials. Medical care 42, 1211–1221.
Prins, M.A., Verhaak, P.F., Smolders, M., et al., 2010. Patient Factors Associated withGuideline-concordant Treatment of Anxiety and Depression in Primary Care.Journal of general internal medicine : official journal of the Society forResearch and Education in Primary Care Internal Medicine 27, 648–655.
Richards, D.A., Lovell, K., Gilbody, S., et al., 2008. Collaborative care for depressionin UK primary care: a randomized controlled trial. Psychological Medicine 38,279–287.
Spijker, J., de Graaf, R., Bijl, R.V., Beekman, A.T., Ormel, J., Nolen, W.A., 2002.Duration of major depressive episodes in the general population: results fromThe Netherlands Mental Health Survey and Incidence Study (NEMESIS). TheBritish journal of psychiatry : the journal of mental science 181, 208–213.
Schoen, C., Osborn, R., Doty, M., Squires, D., Peugh, J., Applebaum, S., 2009. ASurvey of primary care physicians in eleven countries, 2009: perspectives oncare, costs, and experiences. Health Aff (Millwood) ;28:1171–1183.
Simon, G.E., Vonkorff, M., Piccinelli, M., Fullerton, C., Ormel, J., 1999. An interna-tional study of the relation between somatic symptoms and depression. TheNew England journal of medicine 341, 1329–1335.
K.M.L. Huijbregts et al. / Journal of Affective Disorders 146 (2013) 328–337 337
Sheehan, D.V., Lecrubier, Y., Sheehan, K.H., et al., 1998. The Mini-InternationalNeuropsychiatric Interview (M.I.N.I.): the development and validation of astructured diagnostic psychiatric interview for DSM-IV and ICD-10. TheJournal of clinical psychiatry 59 (Suppl 20), 22–33.
Schulz, K.F., Altman, D.G., Moher D., The Consort Group. CONSORT 2010 State-ment: updated guidelines for reporting parallel group randomised trials. Trials2010;11:32.
Twisk, J.W.R., 2006. Applied Multilevel Analysis. Cambridge University Press,Cambridge; United Kingdom.
Unutzer, J., Katon, W., Callahan, C.M., et al., 2002. Collaborative care managementof late-life depression in the primary care setting: a randomized controlledtrial. JAMA : the journal of the American Medical Association 288, 2836–2845.
van der Feltz-Cornelis, C.M., 2009. Towards integrated primary health care fordepressive disorder in the Netherlands. The depression initiative. Internationaljournal of integrated care 9, e83.
van der Feltz-Cornelis, C.M., Ader, H.J., 2000. Randomization in psychiatricintervention research in the general practice setting. Int J Methods in PsychiatrRes 9, 134–142.
Van Marwijk, H.W.J., Grundmeijer, H.G.L.M., Van Gelderen, M.G., et al., 2003. NHG-Standaard Depressie. Huisarts en Wetenschap 46, 614–633.
Wang, P.S., Angermeyer, M., Borges, G., et al., 2007. Delay and failure in treatmentseeking after first onset of mental disorders in the World Health Organiza-
tion’s World Mental Health Survey Initiative. World Psychiatry 6, 177–185.Wang, P.S., Berglund, P.A., Olfson, M., Pincus, H.A., Wells, K.B., Kessler, R.C., 2005.
Failure and delay in initial treatment contact after first onset of mental
disorders in the National Comorbidity Survey Replication. Archives of generalpsychiatry 62, 603–613.
