In the Dutch mental health care sector, an attempt is underway to integrate the different treatment and care components for patients with psychiatric disorders. The most important mental health care institutions have already merged into regional centres for integrated mental health care [1]. Disciplines, too, are making efforts to collaborate in care programmes and to construct a collective sense of profession. So far, most of these attempts have resulted in controversies in the domains of the different disciplines [2]. It is generally acknowledged, however, that bringing together different professional knowledge domains is a necessary step towards an integrated system of health care. Only by bridging the gap between disciplines can adequate answers be given to the following question: What care components—administered by whom—does this patient need at this particular moment?

In this article, I shall report on a project of five major professional groups in Dutch mental health care, which aims to assemble their collective knowledge in the form of multidisciplinary guidelines. In the first section, I shall briefly discuss the history of the project and describe its ideological framework. Next, I shall focus on the main problems the project is currently facing, and show that those problems can be attributed at least in part to the ‘evidence-based’ ideology of the project. In the third section, I shall argue that these problems can only be resolved by accepting non-empirical arguments, and by incorporating into the theoretical framework an ideological statement on what is good integrated care.

Clinical guidelines are “systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances” [3]. Clinical guidelines are generally produced by medical societies [4] and supported by special institutes, such as the National Institutes of Health (NIH) in the United States ( http://www.nih.gov), the National Health and Medical Research Council (NHMRC) in Australia ( http://www.health.gov.au/nhmrc/), the New Zealand Guidelines Group ( http://www.nzgg.org.nz), the National Institute for Clinical Excellence (NICE) in England ( http://www.nice.org.uk), and the Scottish Intercollegiate Guidelines Network (SIGN) ( http://www.show.scot.nhs.uk/). In the Netherlands this task is performed by the Quality Institute for Health Care, the CBO ( http://www.cbo.nl).

The Dutch mental health care sector was comparatively late in developing guidelines. The first guidelines on mental health topics such as depression, anxiety and sleeping disorders were established in primary care, i.e. outside the world of specialised mental health care [5]. Not until the early 1990s did psychiatrists start formulating a guideline on depression [6], which was eventually published three years later, in 1997 [7]. Even though these guidelines were formulated by specialists from different disciplines, they were essentially monodisciplinary in nature, in the sense that they were only accepted and settled by the relevant professional association. These guidelines were heterogeneous; they used different formats, different diagnostic criteria and different treatment options. More importantly, they disagreed on essential features concerning treatment options. The guideline for general practitioners proposed a TCA as the first medicine, the guideline for psychiatrists an SSRI, and the guideline for psychotherapists brief psychotherapy such as CBT or interpersonal therapy [8]. In the late 1990s, psychiatric nurses set out to develop guidelines according to their own insights. Unlike the doctors and psychotherapists, they took problematic care situations (such as handling aggressive patients), rather than clinical syndromes, as their starting point. When the clinical psychologists announced their intention of developing their own guidelines, the whole effort threatened to end in chaos.

In 1998, the Dutch government (Ministry of Health) staged a conference to enable all parties involved to discuss the situation. The general sentiment was that the development of autonomous guidelines was undesirable. The conference resulted in the launch of a joint project towards new integrated multidisciplinary guidelines, to be formulated by representatives of the five core disciplines in mental health care (psychiatrists, general practitioners, psychotherapists (clinical), psychologists and psychiatric nurses). The presidents of the five professional associations formed a Steering Committee, chaired by the General Inspector for mental health care.[1] The CBO and the Trimbos-institute (Netherlands Institute for Mental Health and Addiction, http://www.trimbos.nl) share responsibility for technical support and for the Secretariat.

In its Vision Paper of 2000, the Steering Committee presented its view on both the organisational aspects of guideline development and the concept of multidisciplinary guidelines itself [9]. Each topic was to be supported by its own organisational structure, comprising expert study groups in each of which at least the five core disciplines were to be represented; if desired, other disciplines could participate. The Vision Paper also proposed to establish two permanent commissions: the first to prepare new forms of patient participation in the development of guidelines, the second to ensure the proper implementation of the guidelines.

Guidelines should offer practical suggestions and instructions for professionals and patients concerning preventive, diagnostic, therapeutic and organisational procedures [10]. The development of guidelines can be regarded as a process with three dimensions: height, width and depth (see Figure 1). The width (horizontal axis) presents the different phases in the care process (from mono to multi-phases); the height (vertical axis) gives the number of disciplines involved (from monodisciplinary to multidisciplinary) and the depth (diagonal axis) indicates the level of elaboration of the guideline (from general to specific). Each activity in guideline development can be represented on these three axes.

The phases in the care process are represented on the horizontal axis: prevention, diagnostics, indication and care allocation, specialised diagnostics, formulation of the treatment plan, negotiations with the patient, treatment, nursing, caring, coaching, evaluation, follow-up care, back to a former treatment phase or conclusion of the care process. Not every phase needs to be described in every guideline, but each guideline should prescribe the relevant phases and their tuning. The horizontal axis relates multidisciplinary guidelines to existing programmes in mental health care. The priority in target groups (adults) and topics follows the development of programmes: anxiety disorders, mood disorders, psychotic disorders and schizophrenia, personality disorders, organic psycho disorders, somatoform disorders and substance abuse disorders.

The vertical axis represents the number of disciplines involved in guideline development. The spectrum begins with the monodisciplinary guideline. Ideally, a guideline should include all disciplines concerned in the care process. However, multidisciplinary guidelines do not indicate specific tasks for the disciplines, but suggest appropriate interventions; what must be done, instead of who must do what. Multidisciplinary guidelines should describe the following interventions: diagnosis, biological interventions, psychological interventions, non-verbal interventions, practical and social interventions, nursing, caring and protective interventions, and, finally, coordination and fine-tuning activities. It is important to stress that multidisciplinary guidelines do not describe what the specific disciplines should do. For this reason, the interventions must be translated into tasks for specific disciplines; the multidisciplinary guideline can then be used as a ‘master guideline’ for the establishment of monodisciplinary guidelines. This distinction between interventions and tasks for disciplines also implies that multidisciplinary guidelines always need regional or institutional translations (programmes) in which interventions are allocated to disciplines and institutions. Hence, regional characteristics do not influence the guideline itself, but they do affect the distribution of responsibilities in the care process.

The third, diagonal axis proposes the chronology of the guidelines and their degree of elaboration. Four different products or steps are distinguished in the development of a guideline. The first is an overview of the evidence available. Guidelines should be ‘evidence-based’: statements should be substantiated by the best available knowledge. It is proposed to follow the principles of evidence-based knowledge [11], which distinguish five levels of evidence. The overview paper contains the standard data concerning the topic of the guideline (the characteristics of the problem/disease and of the patients), the treatment options (indications) and possible relations between topic and indications. This overview serves as the basis for the second step: the consensus document. Experts of the participating disciplines will confront the evidence-based knowledge with their collective sense of profession. This process should result in judgements on good clinical practice. The consensus document is then translated into a decision document, which contains so-called decision trees in which the arguments for the different treatment options are represented in relation to the relevant patient characteristics. This document will also make the possible choices for the patient explicit. Finally, the decision document should be translated into an individualised expert system. The system to be developed will probably be a computer program that supports patient and caregiver in the selection of the appropriate intervention, given a specific set of patient characteristics and preferences. The guideline might eventually be translated into an electronic disease management system.

To a large degree this Dutch approach follows (inter) national trends in guideline development, such as the shift from consensus-based to evidence-based guidelines [10]. This method relies on the Cochrane tradition ( http://www.cochrane.org). In addition, its multidisciplinary character is consistent with the international standard (e.g. see http://www.show.scot.nhs.uk/sign/methodology/). However, what makes this approach unique is the number of professions involved. Note also that the collaboration between primary care and specialised mental health care in a single guideline programme is quite exceptional. Indeed, this points to another novelty: the explicit aim of describing the whole care process, from initial registration to chronic care. Another remarkable feature is the degree of client participation; even if such participation is not unique to this method, the weight assigned to patient contribution is exceptional. Finally, the implementation component merits special mention, since a special commission has been established to focus on this aspect. The process is directed by specialists from the primary care sector ( http://www.wokresearch.nl). We may conclude, therefore, that this guideline project is unique in its class.

Since the Steering Committee was launched, six study groups have been installed. In 2000, the first two set out on their task to develop a guideline for anxiety and mood disorders. The third group started in 2001 on the topic of eating disorders, the fourth group in 2002 on substance use disorders. The other two groups deal with schizophrenia and personality disorders, respectively. The first two are now finalising their first overview paper and preparing the second consensus paper. Both groups, after a splendid start, went through a period of serious doubt concerning the usefulness of multidisciplinary guidelines. In the next section I will discuss the arguments.

The first two study groups are currently preparing a consensus document about anxiety and mood disorders. Initially they were quite enthusiastic, but the process now seems to be stagnating. This may be just a period of hesitation, or the calm before the storm of actually establishing the guideline. Some critics, however, claim that more fundamental problems have arisen.

The overview papers for both disorders list a great many interventions. Each discipline offers several treatment or care options, resulting in a description of a large number of interventions for each disorder. It is not clear to what extent these interventions differ from one another. Obviously, SSRI treatment is different from TCA treatment, but it is less certain to what extent improvement of coping should always be distinguished from social skills training. We do not have clear criteria to determine whether interventions are identical or not. The overview papers thus give a huge amalgam of treatment options that are not easily comparable.

From a conceptual point of view, it is almost impossible to list the interventions in a meaningful way. They are described in vocabularies that belong to different bodies of knowledge. Each discipline uses its own concepts. Interventions described in different terms may well be the same, and interventions described in identical terms may well be different. Social workers and psychotherapists both use the intervention known as ‘improving coping’, but their interpretations are quite different. Only the ‘collective sense of profession’ could resolve this conceptual swamp. Or could the ‘evidence-based’ approach offer a solution, as proposed in the Vision Paper?

The paradigm of ‘evidence-based’ medicine or mental health care suggests that empirical arguments could bring more order. The five levels of evidence can be used as markers to categorise interventions; interventions are different if their level of evidence and thus their effects are different. If we had enough proof to corroborate this solution, it would indeed be viable. But most of the research done on the majority of interventions does not exceed level C (non-competitive design). So, we must conclude that the available evidence does not create much transparency.

The same argument can be used to show that the method of evidence-based medicine is even less helpful to prescribe the appropriateness of interventions, which, after all, is the main purpose of a guideline. According to the evidence-based approach an intervention is preferable to another if it has proved to be more effective. The more level A evidence, the more an intervention will be regarded as appropriate. A brief survey of the literature reveals that level A evidence mainly concerns psychopharmacological interventions and brief psychotherapy (cognitive behavioural therapy or interpersonal therapy). In general, both types of treatment seem quite effective, regardless of the kind of disease treated [12]. Other interventions have not yet been so thoroughly studied, and without denying the existing research results we must admit we do not have much level A or B evidence for any of them. In accordance with the evidence-based ideology, it must then be concluded that brief psychotherapy and psychopharmacological interventions should be preferred as first treatment options. In this way, guidelines are basically reduced to protocols indicating what psychopharmacological intervention should be preferred in what dose, or to recommendations on what form of brief psychotherapy is suitable as a first or second treatment option.

Treatments differ in their level of evidence; in the guidelines these differences are recognized, which has serious consequences. Level A evidence needs meta-analyses or randomised clinical trials (RCT), which are generally accepted in medical sciences and suitable to study the effects of specific interventions. However, the RCT paradigm is less accepted or implemented in other disciplines such as nursing, social work or non-verbal therapies. The adherence to the RCT paradigm seems partly related to the level of research: interventions by academic disciplines are more often systematically studied than interventions by non-academic disciplines. But, the quality and quantity of interventions are also important: the more discrete an intervention and the better it can be described in a protocol, the easier it can be studied in an RCT design. Differences with regard to ethical issues also affect the use of the RCT as a golden standard. In acute psychiatry, for example in the case of serious risk of injury, randomisation of the patients in a treatment group and a control group may not be impossible, but it is certainly problematic. Adherence to the evidence-based ideology entails acceptance of the notion that formal differences between disciplines and interventions do imply an a priori hierarchy. In other words: the evidence-based ideology necessarily implies that academic disciplines are favoured over non-academic disciplines and that short, discrete interventions have priority over long-term, continuous or ethically debatable interventions.

Apart from the problem of implicit preferences, there are other reasons to doubt whether an uncensored evidence-based approach is suitable as a foundation for multidisciplinary guidelines. These reasons have to do with the actual body of knowledge in mental health care.

One general problem, which is not specific to mental health care, has to do with the difference between efficacy and effectiveness. High-quality studies are almost always concerned with efficacy, whereas the question of effectiveness seems more important in clinical practice. Efficacy studies in psychotherapeutic interventions reveal the best results for cognitive behavioural therapy [13]. We have no insight into results on effectiveness trials (pragmatic trials) [14]. The results available on the effectiveness of psychotherapy do not indicate much difference between different psychotherapeutic interventions [14]. Given the EBM priority on levels of evidence, the guidelines should prescribe cognitive behavioural therapy. We do not know, however, if this therapy also achieves better results in every-day practice.

A second problem, which is related to the first one, concerns the lack of knowledge on consumer preferences. Existing figures are based on efficacy studies in which patients are randomised across interventions. Hence, RCT studies only marginally provide information about patient preferences on the effects of interventions. The few results available originate from general health care and show only a limited influence of consumer preferences [15]. In mental health care, the figures available derive from naturalistic studies: psychotherapeutic research shows that the motivation of the patient, as well as the agreement between therapist and patient on what should be done, are good predictors of drop-out events and probably also of treatment effect [16]. Again the question is: what should the guideline prescribe? The answer is after all also a political issue.

Two other arguments may raise further questions on the unconditional use of evidence-based knowledge in guidelines. The first has to do with the demand to carry out psychological interventions in a double-blind design, which is virtually impossible. The researcher always knows which patient is receiving what care and his expectations may influence the outcome. The caregiver always knows what treatment he is offering, and within an experiment he generally knows whether his treatment belongs to the experimental group or is ‘just’ treatment as usual. Meta-analysis has revealed a direct link between this knowledge and the results of the study: the outcome of a study can be very well predicted by the beliefs of the researcher or research group [17,18].

The second argument essentially concerns the role of the professional in the care process [19]. In an RCT the role of the professional is that of an expert who carries out the intervention. Interventions are defined by protocols that must be carried out in the most skilful way, with as little interpersonal variance as possible. The effects of studies cannot be studied systematically unless differences between professionals are reduced to an absolute minimum. The ideal situation is thus to ‘throw experts out of the process’ [20]; it is the intervention that is responsible for the results, not the professional. If he has any influence at all it is likely to be negative. In adhering to the RCT paradigm, evidence-based guidelines favour a uniform role for the professional [20]. This position is questionable, if only because it contradicts research in psychotherapy, which suggests that, the therapist is more important than his interventions. This phenomenon is known as the influence of non-specific factors. It accounts for at least forty percent of variance, whereas specific interventions only account for ten to twenty percent of total variance [16,21]. Evidence-based knowledge, therefore, is not as evident as it is supposed to be.

What are the consequences of these arguments for the development of evidence-based guidelines in mental health care? In our view, these critical remarks on the evidence-based paradigm do not necessarily lead to the conclusion that the process of formulating evidence-based guidelines should be stopped. We should realise, however, that the evidence-based approach cannot give final answers to all questions. The problems described above cannot simply be resolved by a more consistent use of the evidence-based paradigm. On the contrary, interventions cannot be compared on the basis of the level of evidence alone. Other criteria are necessary. The same is true for the whole process of developing guidelines: criteria from outside are needed. Although this conclusion is not new (see, for example, the discussion between Klerman and Stone [10], it could have far-reaching consequences for future directions.

The development of multidisciplinary guidelines was triggered by the observation that monodisciplinary guidelines disagree on essential features of the treatment of patients with a depressive disorder.

The disciplines decided to bring together their bodies of knowledge in the hope of developing an integrated guideline. The traditional consensus-based methodology seemed inappropriate to bridge the gap between the disciplines. The more recent evidence-based methodology of developing guidelines offered one important advantage: it created the possibility to compare interventions using an independent and objective vocabulary. The different levels of evidence were used as a first criterion, suggesting that the results of scientific research could solve old controversies. This, of course, was naive.

The problem with monodisciplinary guidelines has moved the discussion to what disciplines have in common. Indeed psychiatrists, psychotherapists, GPs, psychologists, social workers and social psychiatric nurses sometimes offer the same interventions, such as brief psychotherapy, problem solving, bibliotherapy, counselling, etc. Even if drug treatment is reserved to doctors, other disciplines also have some knowledge about it. All these interventions are aimed at reducing complaints and are most likely prescribed in the first or second phase of the treatment process. The monodisciplinary guidelines focused on that phase and those interventions, and it was there that disagreement came to light. Because all those interventions more or less shared the same objectives, it was clear that research outcomes could help in solving the controversies. The EBM paradigm offered a useful basis for the development of new multidisciplinary guidelines.

Disciplines in mental health care, however, do more than reduce complaints. Many interventions have other goals. For example, psychiatric nurses who support the patient during admission carry out interventions directed towards the creation of a context in which improvement becomes possible: helping the patient to get up in the morning, to have his breakfast, to attend different therapies, to resolve conflicts with his family, to realise how his behaviour is reinforcing feelings of misery, etc. In general, it is safe to say that as the psychopathology becomes more complex, so does the care process. Accordingly, that process gradually loses its firm orientation towards symptom reduction.

The recognition of different goals sheds another light on the problems mentioned earlier. The best method to achieve more order in the amalgam of interventions is not to compare interventions as regards their level of evidence but, first and foremost, to compare their goals. This first step seems easy to realise: if disciplines know what they are doing, they can describe why they are doing so. Unfortunately, professional work is not always as transparent as that. Reflective practitioners are needed to draw up guidelines [22]. Moreover, it should be noted that setting intervention objectives can be a highly complex process. The more discretely an intervention can be described, the easier it is to identify and define its purpose. However, as mentioned earlier, many activities in mental health care cannot easily be described as discrete interventions. Clinical care comprises a variety of activities; at first glance the grouping of activities into interventions seems more or less arbitrary. It emerges that disciplines differ considerably in their experience and tradition of describing interventions and setting goals. Disciplines that mainly involve diagnostic and healing tasks generally have fewer problems in defining their goals then disciplines that involve more care tasks. Similar differences exist between the more medical and the more social disciplines. Psychiatric nurses have more experience with this process than social workers or group leaders (pedagogic workers).

The next step in guideline development should be to establish a taxonomy of goals. Interventions should be catalogued: for each intervention goals should be made more explicit. As most interventions can probably be used to reach several goals, a distinction should be made between primary and secondary goals. The challenge then is to construct a taxonomy of goals that is independent of the specific vocabularies of disciplines. It is not clear whether such a taxonomy exists, or indeed whether it can be developed. Goals are related to the way problems are perceived and defined. The literature shows that definitions of problems are essential features of the specific disciplinary bodies of knowledge [23]. As long as disciplines, in the process of making care objectives explicit, adhere only to their own bodies of knowledge, the development of a multidisciplinary guideline will remain difficult. Even so, this does not preclude a multidisciplinary taxonomy. Multidisciplinary taxonomies do exist; one example is the DSM. It is unlikely, however, that the DSM could also be used for the classification of goals, because it has the disadvantage of reducing problems to complaints and disorders. As a result, goals would then too easily be reduced to symptom reduction.

Let us consider in more detail the possibilities of international classification as proposed by the WHO.

In 2002, the World Health Organization (WHO) published an International Classification of Functioning, Disability and Health (ICF) [24]. The ICF is not a ready-made instrument to classify disorders, but it can help to describe human functioning and health problems in relation to external and personal factors. The ICF offers possibilities for combining taxonomies of disorders, such as the DSM, with taxonomies of disabilities, levels of social functioning, etc.

The anxiety disorders working group has applied the ICF to the care process for patients that suffer from anxiety disorders, and has described the various goals of that process [13]. Four different components are distinguished: disorder, disability, participation, and inhibiting factors. Within each component, different categories are identified. The ‘disorder’ component is divided into mental functions, psychomotor functions and cognitive functions. The ‘disability’ component comprises four categories: communication, self-care, housekeeping, interactions, and social relations. Next, the working group fixed the desired results (intervention objectives) for each component or category. The objective for the ‘disorder’ component, for instance, is to reduce symptoms, the objective for the ‘self-care’ category is to achieve an adequate level of daily self-care. Finally, the working group identified the interventions that claim to lead to the attainment of such goals. For example, for symptom reduction they identified the following interventions: psycho-education, cognitive behavioural therapy, drug treatment, combination treatment, supportive interventions, relaxation therapy and, finally, movement therapy. Sociotherapy is mentioned for adequate daily self-care.

The ICF taxonomy elaborated by the anxiety disorders working group seems promising. It offers a framework for the classification of interventions, with reference mainly to goals, and also offers perspectives for the development of guidelines. But, the proof of the pudding is in the eating. Other working groups are likely to opt for different frameworks. In each case a framework is needed to determine the intervention objectives. Classification is necessary in order to compare interventions. In the development of guidelines, it is only in this phase that a comparison between interventions becomes fruitful. Once agreement has been reached on a taxonomy of goals and interventions have been classified, it is possible to proceed to comparing interventions that share the same goals. Obviously, in that phase levels of evidence can be used as a standard for comparison.

Guidelines should describe interventions and their goals, and indicate which interventions are suitable for achieving a specific goal. In this paragraph, we will argue that guidelines should also deal with the question of how to determine appropriate goals during the treatment process.

Why not leave the choice to the patient himself? After a diagnosis, caregivers could present a taxonomy of goals and explain to the patient how these different goals could be realised. The patient could then opt for a specific goal, and the guideline would prescribe the appropriate interventions to achieve it. Even if this idea sounds sympathetic to the patient, it is slightly nanve. Some goals are inherently difficult to achieve, others are only attainable after other goals have been achieved. This means that at least some professional knowledge is required. Caregivers and patients should jointly select the goals for treatment. Given that professionals always have the final responsibility for the treatment they offer, it is their task to determine the treatment goals together with their patients. However, this is not to say that the guidelines should leave the selection of goals totally open. Guidelines should support professionals and patients in their selection of specific goals.

Other solutions are conceivable. For instance, multidisciplinary guidelines could leave room for regional or institutional preferences. Another possibility is to leave the job to monodisciplinary guidelines. However, these solutions, too, would fail to incorporate the real choices into the guidelines. What is needed, therefore, is a hierarchy of goals.

Multidisciplinary guidelines should combine the evidence, the collective sense of profession and a vision of the care process, i.e. a hierarchy of goals. A guideline should indicate which goals need to be realised first, what comes next, and how to handle complex interactions between different interventions.

Although several models for the care process are available, in the Dutch mental health care sector we see an increasing interest in the principles of stepped care. Stepped care provides a framework for the care of patients that uses limited resources to their greatest effect on a population basis. In stepped care, the intensity of professional care is augmented for patients who do not achieve an acceptable outcome with lower levels of care [25]. Stepped care proposes to opt first for the less intrusive forms of care that offer a chance of success. Only if these do not lead to improvement, more intrusive care is prescribed. In other words: the first step is to choose the intervention that is most effective in facilitating the patient's capacity to cope. Stepped care maximises the patient's autonomy and empowerment.

The stepped care model offers several advantages. First, it provides clear criteria for choosing among interventions that are equally effective. Second, it can be used to construct a hierarchy of goals. Figure 2 presents an example of such a hierarchy.

The hierarchy proposed is constructed on the basis of evidence in combination with a judgement on the level of intrusiveness. In the absence of clear-cut contraindications (crisis, psychosis, etc.), the first step is a psychosocial intervention (re-labelling the context, problem solving psycho-education, watchful waiting, etc.), which should take no more than a few weeks. If after this period no amelioration can be observed, the patient is guided towards the second step, which focuses on the reduction of symptoms. Common methods are drug treatment, cognitive behavioural therapy, interpersonal therapy, etc. The second step takes three months or less, and after this period the complaints should have subsided. If the problems persist, the patient is referred to the third step, where the interventions are directed towards the transformation of adaptation mechanisms. Psychodynamic psychotherapy, partner relation therapy and experiential therapy are examples of this type of intervention. The third step tends to take a fair amount of time. The fourth step provides a combination of the first three steps, and is generally offered in specialised transmural care centres. These centres have two points of entry: from the third step, or from a crisis intervention stage. If the fourth step is not successful either, it is sometimes possible to refer the patient to a fifth step in the form of specialised top referent care.

Each step is characterised by a unique set of goals, to be realised by several interventions. These interventions can subsequently be compared as regards their level of evidence. Sometimes we do have some additional evidence concerning their order in the treatment process (within the same step). For example, drug treatment should precede cognitive behavioural therapy. A guideline based on the principle of stepped care should first present diagnostic criteria for inclusion (What kinds of patients?) and exclusion (Which patients should be directly referred to crisis intervention or specialised care?). Next, the guideline should present the main goals of each step and describe the interventions needed to realise these goals. This should be followed by a description of the available evidence for each intervention, as well as of possible contraindications and (adverse) side effects. Finally, the guideline should give information on how these interventions interrelate. By linking this information to the course of the disease, the guideline can be given the characteristics of a disease management system.

In this article I have described a Dutch experiment aimed at the development of multidisciplinary guidelines in mental health care. I have argued that this project is unique in many respects. However, I have also had to admit that the results have so far fallen short of expectations. Indeed, the new guidelines merely seem to replicate existing guidelines. In my view this can be attributed at least in part to an unduly narrow interpretation of the EBM methodology. We cannot develop guidelines simply by scanning the literature and weighting the levels of evidence. Empirical arguments alone are simply not enough to draft multidisciplinary guidelines. Moreover, there is a need to take account of collective sense of professions, in weighting the literature as well as in setting care objectives and reaching agreement on their hierarchy. The stepped care model, which is common in general health care, is an example of such an explicit set of goals in an explicit hierarchy.

As long as guidelines remain restricted to evidence-based interventions, without explicit goals and without hierarchy, integration of care will only become more problematic. In the short term I foresee that several disciplines will terminate their collaboration with the project and develop their own monodisciplinary guidelines, which will probably be mutually contradictory. Even more problematic will be the implicit shift in the central question, from: What kind of care is needed in what phase?, to: Who should provide care? This shift will refuel strong controversies and rivalries between disciplines. I do not consider those power games to be conducive to the integration of care.

However, if a set of goals is introduced and a hierarchy between these goals is established, the new multidisciplinary guidelines may pave the way towards the further integration of mental health care; more specifically, towards the integration of content-related aspects, in the wake of organisational integration. They may even lead to the integration of care without the need for institutions to actually merge. In that situation, guidelines could serve as care programmes relating tasks of disciplines in primary and specialised health care.