Study sample

The study was a large multicentre randomised trial conducted in MHCs in The Netherlands. The MHCs are specialised mental healthcare institutes to which patients must be referred by a general practitioner (GP). Seven MHCs (Altrecht, Eleos, Emergis, Gelderse Roos, Mentrum, Zaanstreek/Waterland and Adhesie) participated in the study and were representative of the patient population. Patient selection was undertaken in two steps. Enrolment took place between February 2000 and October 2001, and follow-up data were collected until March 2003. First, during this period a total of 6095 new clients presented to the MHCs and were screened with regard to inclusion criteria (age 18–65 years, eligible for out-patient mental healthcare, not treated by the same MHC in the past year) and exclusion criteria (psychotic or bipolar disorder, cognitive impairment, high suicide risk, poor command of the Dutch language, hard drug misuse/dependence). Comorbidity associated with other psychiatric diagnoses (with the exception of psychotic or bipolar disorder) was allowed, and this included personality disorders, alcohol misuse/dependence and somatic disorders. Second, all of the remaining patients were assessed with the Composite International Diagnostic Interview (CIDI; Reference WittchenWittchen, 1994). Patients with the following DSM–IV diagnoses (American Psychiatric Association, 1994) were included: major depressive disorder (single episode or recurrent); dysthymic disorder; panic disorder (with or without agoraphobia); social phobia; or generalised anxiety disorder. Over a fifth of the interviewed patients were excluded (n=1338; 22%) or else failed to meet the inclusion criteria (n=1257; 21%). Nearly a third of the patients (n=1995; 33%) refused to participate, and 803 (13%) were not included for other reasons (e.g. not contactable because telephone number was incorrect, on holiday). In total, 702 patients gave their informed consent and were included in the study. This sample was somewhat smaller than the pre-set sample size of 750 patients.

A follow-up period of 1.5 years was chosen to establish the long-term effects on health status and costs. Measurements were recorded at baseline and then every 3 months. Patients were permitted to switch to an alternative treatment arm if their medical condition rendered this necessary. We applied intention-to-treat analyses.

Interventions

Brief therapy is a formalised ‘stepped-care’ approach. The therapy focuses mainly on the present, and emphasises abilities rather than disabilities. Brief therapy is expected to reduce costs and increase efficiency in the short term. Its claim to long-term cost-effectiveness in the Dutch mental healthcare setting for a broad range of psychiatric problems has not yet been confirmed by scientific evidence. Problems may recur at a later date, or patients may require additional treatment in mental healthcare services or other parts of the healthcare sector and consequently generate additional costs. In the present study, patients who were undergoing brief therapy for depression or anxiety were compared with those who were receiving CBT, which had been shown by earlier studies to be an effective approach (Reference Black, Wesner and BowersBlack et al, 1993; Cloaguan et al, 1998; Reference Dorrepaal, van Nieuwenhuizen and ScheneDorrepaal et al, 1998). In this study, CBT was formalised and the main focus was on altering irrational cognitions by challenging them. The maximum number of sessions was 15. In the same way as for brief therapy, a ‘stepped-care’ approach was adopted for CBT, in which all patients initially received a first-line treatment, and they were only switched to another therapy if the first therapy proved to be inadequate. In The Netherlands, care as usual is not formalised, and a multidisciplinary team is free to assign a therapy from a wide range of therapeutic options. The treatment decision is based on professional experience, taking into account the specific problems and characteristics of the individual patient. The number of sessions depends on the therapy that is offered. Evidence on the (cost-)effectiveness of care as usual is lacking. However, this type of ‘matched care’ is claimed to be more effective than a ‘stepped-care’ approach, in which one form of (relatively brief) therapy is started for all patients indiscriminately, and patients are switched to other options if there is no subsequent improvement or if side-effects occur.

Measures

At baseline, several demographic characteristics (e.g. age, gender, educational level and employment status) were assessed. In all interviews that were conducted during treatment, we assessed health-related quality of life, use of medical resources and productivity loss using the EuroQol Questionnaire (EQ–5D) (Reference Essink-Bot, Stouthard and BonselEssink-Bot et al, 1993) and the Trimbos and iMTA Questionnaire on Costs Associated with Psychiatric Illness (TiC–P; Reference Hakkaart-van RoijenHakkaart-van Roijen, 2002).

Data analysis

Statistical analysis

Multiple imputation analysis was performed using PROC ‘MI’ in SAS for Windows (available in version 8.02 and higher). All other statistical analyses were performed using SPSS version 10.1 for Windows.

Direct and indirect costs

Table 2 shows the estimated total mean direct medical costs and indirect costs per patient over a period of 1.5 years. As was expected, the mean number of contacts and the associated costs in the MHCs were significantly lower in the brief therapy group than in the CBT group (95% CI €169 – €741) or the usual care group (95% CI €14 – €464). However, no significant difference was found between the three treatment groups with regard to the mean total direct medical costs per patient (i.e. including the costs due to resource utilisation in other parts of the healthcare service), nor was there a significant difference in indirect costs between the three treatment groups.

Table 2

Mean direct medical costs and indirect costs¹

	Care as usual	CBT	Brief therapy
Number of patients	245	241	187
Number of sessions: mean (s.d.)	12 (14)	10 (7)	8 (11)
Costs (€): mean (s.e.)
MHC	1396 (107)	1180 (58)	941 (99)
Other medical care	1964 (235)	1947 (239)	2738 (395)
Total direct medical costs	3360 (251)	3127 (255)	3679 (417)
Indirect costs	6151 (551)	6621 (683)	6537 (675)
Total costs (direct and indirect)	9511 (607)	9748 (747)	10216 (816)

Over time, for all treatment groups the percentage of patients in paid employment who had long-term absence from work declined, and conversely the percentage of patients who had no impediments increased (Table 3).

Table 3

Percentage of respondents in paid employment with impediments, and short-term and long-term absence from work by treatment group¹

	Care as usual	CBT	Brief therapy	Total
Baseline
n	149	141	130	420
No impediments at work (%)	13	12	16	14
Impediments at work (%)	30	30	33	31
Absent from work for <2 weeks (%)	15	11	12	13
Absent from work for ≥2 weeks (%)	56	57	50	55
One-year follow-up
n	95	80	90	265
No impediments at work (%)	46	46	48	47
Impediments at work (%)	29	38	30	32
Absent from work for <2 weeks (%)	8	15	11	11
Absent from work for ≥2 weeks (%)	20	16	19	18
End of study (1.5 years)
n	86	80	77	243
No impediments at work (%)	52	63	43	53
Impediments at work (%)	33	30	40	35
Absent from work for <2 weeks (%)	17	14	16	16
Absent from work for ≥2 weeks (%)	12	8	17	12

Quality of life

The utility scores for the three treatment groups also showed significant improvement during the study period and did not differ significantly between the groups (Table 4). The improvement in utility scores was moderate during the first year, but was low during the final 6 months of follow-up (Reference CohenCohen, 1988). At the end of the study period, the patients’ quality of life was still significantly lower than the average utility score of the general population (0.88) (Table 4).

Table 4

Utility scores by treatment arm at baseline, at 1 year and at the end of the study (1.5 years), and QALYs per patient at the end of the study

	Care as usual	CBT	Brief therapy
Number of patients	245	214	187
Baseline
Utility score (s.e.)	0.51 (0.02)	0.52 (0.02)	0.53 (0.02)
Number of patients	165	127	131
One-year follow-up
Utility score (s.e.)	0.64 (0.02)	0.68 (0.02)	0.64 (0.03)
Number of patients	147	111	126
End of study (1.5 years)
Utility score (s.e.)	0.69 (0.04)	0.65 (0.04)	0.71 (0.04)
QALYs (s.e.)	0.909 (0.03)	0.936 (0.03)	0.939 (0.03)

Cost utility

Table 5 shows the incremental cost-effectiveness ratios both for direct medical costs per QALY gained and for total costs per QALY gained. As we are comparing three treatment options, we have reported the results in the format suggested by Karlsson & Johannesson (Reference Karlsson and Johannesson1996). The treatments are ordered from least to most effective. Comparison of the direct medical costs of care as usual and CBT shows that CBT is superior to usual care (with lower costs and better outcomes), so an incremental cost-effectiveness ratio is only calculated for brief therapy ν. CBT, yielding a value of €262 857 per QALY gained. With regard to total costs the same relationship applies (CBT could achieve the same number of QALYs at lower costs), so again only the incremental cost-effectiveness ratio for brief therapy ν. CBT is calculated, yielding a value of €222 956 per QALY gained.

Table 5

Costs, effects and incremental cost-effectiveness ratio for the three treatment groups

Treatment	QALYs	Direct medical costs (€)	ICER	Total costs (€)	ICER
Care as usual	0.91	3360	CBT could achieve the same number of QALYs at lower cost than treatment as usual	9511	CBT could achieve the same number of QALYs at lower cost than treatment as usual
CBT	0.94	3127		9748
Brief therapy	0.94	3679	262 857	10 216	222 956

To assess the uncertainty, we constructed acceptability curves for each treatment by calculating the proportion of bootstrap replicates for which that treatment is optimal, for a number of threshold incremental cost-effectiveness ratios (Fig. 1) (Reference Fenwick, Claxton and SculpherFenwick et al, 2001). Figure 1 relates to the direct medical costs per QALY, and it indicates that, taking uncertainty into account, CBT is optimal. However, the preference for CBT becomes less strong as the threshold incremental cost-effectiveness ratio increases. The same is true if all costs are considered (data not shown).

Fig. 1

Acceptability curves for the three treatment options, for direct medical costs per QALY gained. CBT, cognitive–behavioural therapy; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life-year.

Sensitivity analyses

In the sensitivity analyses of the missing data, we recalculated the costs and QALYs by performing linear extrapolation and complete case analysis (Table 6). Again no significant difference was found between the groups in overall mean costs and QALYs. When only 48% of the participants were used for the complete case analysis this resulted in comparable findings to those of the linear extrapolation analysis, but with much larger standard errors.

Table 6

Cost per QALY after applying multiple imputation, linear extrapolation and complete cases analysis of the missing data for the three treatment groups

	Care as usual	CBT	Brief therapy
Multiple imputation
Direct medical costs (€): mean (s.e.)	3360 (251)	3127 (255)	3679 (417)
Indirect costs per patient (€): mean (s.e.)	6151 (551)	6621 (683)	6537 (675)
QALY: mean (s.e.)	0.909 (0.03)	0.936 (0.03)	0.939 (0.03)
Linear extrapolation
Direct medical costs (€): mean (s.e.)	3573 (295)	3191 (240)	3605 (327)
Indirect costs per patient (€): mean (s.e.)	6331 (567)	6383 (672)	6061 (693)
QALY: mean (s.e.)	0.895 (0.03)	0.931 (0.03)	0.906 (0.03)
Complete cases analysis
Direct medical costs (€): mean (s.e.)	3770 (392)	2924 (360)	3792 (594)
Indirect costs per patient (€): mean (s.e.)	5767 (717)	6135 (925)	6263 (928)
QALY: mean (s.e.)	0.914 (0.03)	1.009 (0.03)	0.920 (0.04)

Sample selection

We conducted a large-scale multicentre randomised trial in the setting of general mental healthcare services in The Netherlands. External validity may have been affected by selection bias resulting from patients either refusing to enter the study or being lost to follow-up. Our data on refusal and our missing data seem to be comparable to those of other important trials (Reference Lambert, Ogles and LambertLambert & Ogles, 2004). Moreover, treatment drop-out in this study was comparable to that in the general population of MHC patients (Reference van der Sande, van Hoof and Hutschemaekersvan der Sande et al, 1992). We recruited patients with a wide range of mood and anxiety disorders, comparable to a normal population of out-patient mental healthcare patients. Furthermore, the patients in our trial reported more (severe) psychological symptoms at baseline (mean score of 223 on the 90-item Symptom Checklist (SCL–90) (Reference DerogatisDerogatis, 1977)) than did a reference population of arbitrarily selected Dutch psychiatric out-patients (mean SCL–90 score=204) (Reference Arrindell and EttemaArrindell & Ettema, 2003). There was no indication of any selection bias leading to the enrolment of only ‘mild’ cases in the study. Furthermore, we found no significant differences in baseline characteristics or mental health status between patients who completed the study and those who did not. Therefore there is no compelling reason to believe that selection bias affected the external validity of the study.

Study design

Unfortunately, this study did not provide an opportunity to evaluate the cost-effectiveness of psychological treatment compared with natural recovery of the patient. The potential effect of antidepressant medication, which was allowed in addition to the treatment to which patients were randomised, is also unclear. At baseline, 36% of the respondents used antidepressant medication, and there were no significant differences between the treatment groups.

Direct costs

It was not possible to distinguish between healthcare utilisation for depression and/or anxiety disorders and that for eventual other general healthcare problems which relied on other areas of the healthcare system. However, this was equally true for all three treatment groups. Furthermore, data were collected on healthcare utilisation, which it was expected would be relevant to the treatment of psychological problems. Data on healthcare utilisation other than sessions at the MHCs were collected by self-report. A previous study has indicated that such patients’ self-reports are a valid source of data on days of hospitalisation and out-patient visits. However, costs of medication may be underestimated (Reference van den Brink, van den Hout and Stiggelboutvan den Brink et al, 2004).

Indirect costs

Our study indicated that inclusion of the indirect costs for patients with depression and/or anxiety was highly relevant. We did not assess productivity losses resulting from reduced efficiency at work and from unpaid work (e.g. household work), because of the practical and methodological difficulties involved in measuring these losses. Consequently, the actual productivity losses to society were probably underestimated.

In a population that has social insurance it is unlikely that respondents include the societal impact of ill health in quality-of-life measures, because they do not bear the full consequences of their reduced productivity (Reference Brouwer, Koopmanschap and RuttenBrouwer et al, 1997; Reference Meltzer and JohannessonMeltzer & Johannesson, 1999). Recently, an empirical study by Sendi & Brouwer (Reference Sendi and Brouwer2005) indicated that respondents do not include the effect of ill health on income if the instrument used does not explicitly ask about this effect (as is the case for the EQ–5D).

Cost-effectiveness

The results of our study are consistent with the findings of a systematic review on the effectiveness and cost-effectiveness of brief psychological treatments for depression (Reference Churchill, Hunot and CorneyChurchill et al, 2001). The review suggested that some forms of brief psychological treatment, particularly those derived from cognitive–behavioural models, are beneficial in the treatment of people with depression who are being managed outside the hospital setting (Reference Churchill, Hunot and CorneyChurchill et al, 2001).

A cost-effectiveness study by Pyne et al (Reference Pyne, Smith and Fortney2003) estimated the cost per QALY of a primary care intervention for women with depression. Its findings suggested that enhanced care for women with depression was more expensive and more effective than usual primary care, the additional cost being US$5244 per QALY.

Overall, we found no cost savings of brief therapy over CBT or care as usual. However, in terms of the MHCs, brief therapy was a cost-effective treatment and may help to reduce waiting lists. In routine practice, ‘stepped care’ is characterised not only by fewer sessions, but also by an earlier start after intake. Subsequently, the MHCs may increase the quality of care that they provide by a formalised approach, and be more successful in meeting the preferences of patients. This might allow redistribution of some resources to the group of patients who are not effectively treated. This treatment policy should be supported by a monitoring system for detecting inadequately treated patients to ensure that there is a timely switch to a more appropriate treatment option.

CLINICAL IMPLICATIONS

1. ▪ Care as usual (‘matched care’) was not more cost-effetive than brief psychological therapy or cognitive–behavioural therapy (‘stepped care’).

2. ▪ Brief psychological therapy may help to reduce the waiting lists of mental health care centres.

3. ▪ Brief psychological therapy reduces the costs of mental healthcare centres and may allow redistribution of resources to the group of patients who are ineffectively treated.

LIMITATIONS

1. ▪ Data on the use of other healthcare services and on short-term absence from work were collected from patients with a limited recall period.

2. ▪ Complete service use data for other healthcare services and data on absence from work for the total follow-up period (1.5 years) were available for less than half of the respondents.

3. ▪ The study provided no opportunity to evaluate the cost-effectiveness of psychological treatment compared with natural recovery of the patient.