Abstract
The changing landscape of psychiatry
The last quarter of the century has been a time of dramatic change for at least two areas of psychiatry: the technology of research into mental disorders and the delivery of mental health services. Previously dissociated from one another, these two aspects of the discipline now begin to appear connected as the ‘upstream’ and ‘downstream’ regions of a common flow of development. ‘Upstream’, the application of molecular and cognitive neuroscience to the study of brain development, structure and function has a far-reaching impact on the way psychiatric disorders will be defined, diagnosed and managed over the next decades. Functional brain imaging is already proving to be a powerful tool for unravelling the cerebral basis of at least some of the psychopathological symptoms. Molecular genetics has been successful in identifying genes that play a role in the susceptibility to Alzheimer's disease [4]. Notwithstanding the difficulties associated with disentangling genetic complexity, there is a reasonable prospect that genes predisposing to disorders such as schizophrenia, bipolar affective illness, autism and other developmental disorders, as well as genes contributing to temperament and character traits, will eventually be identified [5].
‘Downstream’, the trend in mental health service delivery in the economically affluent part of the world has been one of a major shift from hospital-based to community psychiatry. This development is accompanied by a growing pressure for ‘evidence-based’ treatment and management modalities, cost containment and accountability. In other parts of the world, which include the developing countries and many of the former socialist economies, the issue at stake is how to provide basic services and an affordable minimum of treatment to all those in need who constitute the world's majority of people with mental disorders. The outcome of these global processes will determine whether the benefits of new knowledge and treatments generated ‘upstream’ will be distributed equitably according to need or will only reach the privileged few.
More than ever, diagnosis and classification of psychiatric disorders is expected to play a linking role between psychiatric research and mental health service delivery. As a consequence, largely unintended or unexpected by their authors, DSM-IV and ICD-10 are being perceived outside the clinical context as documents conferring scientific credibility to the definition of mental illness and legitimacy to the practice of psychiatry [6].
Significance of the ‘common language’ in psychiatric classification
These developments resulted from fundamental changes in the area that had been the Achilles’ heel of the discipline: the scientific status of psychiatric diagnosis. In the late 1960s, the World Health Organization (WHO) initiated a critical review of diagnosis and classification in psychiatry which resulted in a glossary of mental disorders to accompany ICD-8 [7]; a multiaxial classification of childhood disorders [8]; and a standardised diagnostic interview linked to a computer algorithm [9]. These developments were reinforced by the emergence of a ‘neo-Kraepelinian’ school of thought in the United States [10] and the bold innovations introduced with DSM-III. In turn, WHO developed ICD-10 [11] which was radically different from the earlier, rather conservative versions of the international classification. The work on ICD-10 was accomplished in close consultation with the designers of DSM-IV [12], and the two classifications converged on many essential features.
Two decades after the introduction of operational diagnostic criteria in psychiatry, it is time to examine the impact of these tools in a broader perspective. The introduction of an internationally shared framework of concepts, a rule-based classification, and explicit diagnostic criteria made it possible: (i) to increase diagnostic agreement among clinicians and improve the statistical reporting on psychiatric morbidity, services, treatments and outcomes; (ii) to introduce rigorous diagnostic standards in psychiatric research; (iii) to reduce the scope for idiosyncrasies in the teaching of psychiatry by providing an international reference system; (iv) to improve communication with consumers, carers, and the public by demystifying psychiatric diagnosis and making its logic transparent to non-professionals. Although the majority of the current diagnostic criteria remain provisional (and some of them are frankly arbitrary), they have been stated in a testable form and can be modified or rejected using the normal procedures of hypothesis testing in science.
While acknowledging such gains, we must examine critically the current versions of standardised diagnostic criteria and rule-based classification systems in psychiatry for conceptual shortcomings and adverse practical side effects. Two difficulties that restrict the benefits to be gained from the current diagnostic systems need to be especially considered. First, the introduction of explicit diagnostic criteria and new classification categories in psychiatry took place in the context of a discipline that still lacks sufficient conceptual coherence and hence remains easily influenced by ideological, political and market forces. Second, there are inherent shortcomings in the design of these classification systems which limit their usefulness and make them liable to misinterpretation or misuse.
In the wake of the global success of the new classifications, such problems have not received sufficient attention. Yet they must be stated and discussed if further progress in consolidating the scientific base of the discipline is to be achieved.
The social and cultural context of psychiatric classification
More than any other medical discipline, psychiatry is characterised by the historicism of its key concepts [13]. Its theoretical framework has changed surprisingly little since the turn of the century, although novel concepts and methods borrowed from fields ranging from genetics to epidemiology feature prominently in current psychiatric research and practice. Whether overtly or implicitly, psychiatrists today tend to invoke in clinical discourse concepts formulated in the 19th century. In contrast to other medical disciplines where a reference to Laennec or Ossler would be considered appropriate in the context of the history of medicine but hardly in the diagnostic formulation of a current case, it is not unusual for psychiatrists to refer to the authority of Kraepelin, Bleuler or Freud in their daily clinical work. The conceptual continuity with the past rests on paradigms such as descriptive psychopathology, phenomenology, psychodynamic theory and the disease model of mental illness. These paradigms have been linked to differing assumptions about the nature and classification of mental disorders that persist into the present but remain poorly integrated. Both ICD-10 and DSM-IV mirror, to an extent, this lack of internal conceptual coherence of the discipline. As a result, psychiatry tends to oscillate, pendulum-like, between contrasting views on the nature of mental disorders aptly described by Eisenberg [14] as ‘mindless’ versus ‘brainless’ psychiatry. Such conceptual instability may again make psychiatry vulnerable to political ideologies, market forces and various forms of abuse. A poignant reminder from the not too distant past should be the collusion between eugenics, psychiatric genetics and clinical psychiatry in Nazi Germany, which ultimately resulted in the ‘euthanasia’ of over 70 000 psychiatric patients [15]. It is sobering to remember that the culling of the mental hospital population was supported methodologically by a categorical classification of the mental disorders with a strong genetic tint [16]. More recently, the notorious cases of compulsory psychiatric treatment of political dissidents in the former Soviet Union also involved attempts at justification by reference to particular diagnostic categories [17].
The risk of misuse of diagnostic categories and classifications for political or economic purposes is not buried with the past. The de-institutionalisation of the mentally ill, initially driven by the ideology of social psychiatry, was subsequently highjacked by economic rationalism. Where no adequate community-based alternatives have been put in place, the result has been pauperisation and homelessness for a large number of people impaired by psychiatric illness [18]. Current and future calls for managed or rationed psychiatric care will also seek the ‘evidence-based’ imprimatur of psychiatric classification.
Such aberrations certainly do not portray the essence of psychiatry. However, it would be a mistake to dismiss them as mere accidents. Concepts about the nature and classification of psychiatric illness will always attract ideological and political attention that can translate into laws, policy, or other action with unforeseen consequences. To quote Eisenberg again [19], ‘the planets will move as they always have, whether we adopt a geocentric or a heliocentric view of the heavens…But the behaviour of men is not independent of the theories of human behaviour that men adopt’. Psychiatric diagnosis and classification are theories that cannot be de-contextualised.
Classification and the ascendancy of the ‘medical model’ in psychiatry
The introduction of the new classifications occurred concurrently and synergistically with the trend of ‘medicalisation’ of psychiatry which aims to remove the boundaries that separate mental illness from physical disease. Most of the funded psychiatric research today is being conducted within a biomedical paradigm. In the words of an influential protagonist of this school of thought, ‘psychiatry is a branch of medicine, which in turn is a form of applied biology…I have elsewhere characterized this approach as the Medical Model emphasizing thereby that psychiatry, as a medical discipline, should be based as the rest of medicine on modern biological science’ [20]. However, despite the advances in neuroscience, clinical diagnosis in psychiatry remains intrinsically dependent on the clinician's ability to elicit, and the patient's readiness to communicate, subjective experience. Except for rare instances, hardly any ICD-10 or DSM-IV diagnosis relies entirely or primarily on objective signs or tests. The evidence required for diagnosis is essentially phenomenological and behavioural–descriptive, and this remains as true today as it was a century ago. Consequently, communication and semiotic analysis in which the clinician's trained introspection plays an important role remain key components of the diagnostic process. Intersubjectivity, therefore, is intrinsic to the discipline [21]. It is debatable whether this ‘subjective element’ in psychiatric diagnosis could ever be replaced by instrumental measurement or, if it could, at what cost.
How good is the fit between current classifications and clinical reality?
The use of the current classifications in clinical research and practice raises a number of issues concerning the ‘goodness of fit’ between diagnostic concepts and certain aspects of clinical practice and research. The following is an overview of some of the issues.
Classification of diseases or classification of people?
Both DSM-IV and ICD-10 are ambiguous with regard to the basic unit of classification. While the ‘misconception that a classification of mental disorders classifies people, when actually what are being classified are disorders that people have’ [12] is being emphatically rejected, presumably on grounds of avoiding stigma and labelling, the implications of the fact that the primary material out of which the diagnostic entities in psychiatry are constructed consists of patterns of human behaviour, are circumvented by using the term ‘disorder’. The all-embracing category of ‘disorder’ has no clear correspondence with either the concept of disease or the concept of syndrome in medical classifications. Some of those patterns of behaviour constitute syndromes in the medical sense, and some probably do not; however, in DSM-IV and ICD-10 ‘disorders’ have assumed de facto the status of quasi-disease entities. Whereas true disease entities in psychiatry remain ‘an idea in Kant's sense of the word’ [22], it may be wise to reinstate the concept of the syndrome in psychiatric classification as an heuristic tool enabling a sharper delineation of the taxonomic status of many of the diagnostic entities. This was proposed many years ago by Essen-Möller [23], who advocated the separate classification of the clinical syndrome and the presumed aetiology: ‘At the present state of knowledge, there appears to be a much closer connection between aetiology and syndrome in somatic medicine than in psychiatry …while in somatic medicine it is an advantage that aetiologic diagnoses take the place of syndromes, …in psychiatric classification, aetiology can never be allowed to replace syndrome…a system of double diagnosis, one of aetiology and one of syndrome, has to be used’.
Categories versus dimensions
Although the possibility of a dimensional classification is acknowledged, and Appendix B to DSM-IV provides tentative dimensional descriptors and additional axes, the current practice of diagnosis and classification is based on the assumption that psychiatric disorders form discrete categories. A thoughtful clinician or researcher would detect in this a ‘fallacious belief that psychopathological processes constitute discrete entities, even medical diseases, when in fact they are merely concepts that help focus and coordinate our observations’ [24]. It could be argued that the multiaxial recording format introduced with DSM-III and its successors does, in fact, provide a dimensional structure that reflects clinical reality. However, a contradiction in the internal structure of the current multiaxial classifications is often over-looked. While the five axes of DSM-IV usefully highlight the multiaspect nature of ‘caseness’ in psychiatry, the presentation of the diagnostic rubrics remains traditionally categorical. Moreover, the majority of the DSM-IV diagnostic categories contain, as inclusion or exclusion criteria, aetiological assumptions. For example, to diagnose schizophrenia, one must rule out certain aetiological factors such as organic brain disease or substance abuse. The criteria also require the ascertainment of social behaviour or dysfunction. There is an inherent circularity in using features of the axes as criteria defining the entities that are to be classified according to those same axes. This is in conflict with the intended purpose of the multiaxial system which has been designed for an independent evaluation of syndromes, personality traits, possible aetiologies, and social functioning. The problem was anticipated by Essen-Möller [23]: ‘…when using two different systems of classification simultaneously,…one of syndromes and one of aetiology, it is of course important to avoid contamination between the two systems. It is therefore necessary to exclude, as far as ever possible, any aspects of aetiology from the concept of syndromes’. Such contamination has not been avoided in present systems, and the goal of designing a classification model that would be isomorphic to the nature of the majority of psychiatric conditions remains remote.
Symptoms versus course and outcome
The problem of drawing boundaries between the diagnostic entities in psychiatry has so far defeated all attempts at finding the ‘final solution’ by various rearrangements of symptoms and signs. The inclusion, since DSM-III, of duration, course and outcome among the criteria that define disorders, has probably created more problems than it purports to solve. An editorial in the American Journal of Psychiatry, a journal that had played a leading role in promoting the ‘criterion-based definitions of DSM-III and its successors’, aptly commented that ‘narrowing the concept of schizophrenia by including the requirement of established chronicity, while broadening affective illnesses by eliminating the concept of spontaneous remission, has resliced the same “pie” of patients, with the net effect that the prognosis of both groups of illnesses appears worse’ [25].
‘Atypical’ disorders
The problem of patients and clinical states that refuse to fit into any of the standard diagnostic categories has always accompanied psychiatric classification but it seems that the current, rule-based diagnostic concepts are contributing to an increasing proportion of patients who are simply undiagnosable or have to be allocated to the residual categories of ‘other’ or ‘not otherwise specified’ disorders. As a by-product of the rigour of the operational definitions and criteria, the present classifications have a poor ability to account for mixed or atypical states. Estimates from the clinical trials of the new classifications suggests that the ‘goodness of fit’ between the diagnostic criteria and the actual features of clinical cases is unsatisfactory in 18–22% of cases [26]. This phenomenon is perhaps less troublesome to clinicians (whose experience has taught them tolerance for ambiguity) than for researchers who require clear definitions and crisp boundaries. The exclusion of cases from study samples because of a lack of unequivocal diagnostic classification probably results in considerable loss of information (e.g. in genetic research where such individuals are either not counted or classified as having ‘unknown’ affection status).
Comorbidity
Another by-product of the current classification systems is the proliferation of comorbid diagnoses. Since the abandonment of the hierarchical rules (which existed in DSM-III) it has become increasingly difficult to decide what to do with the large number of individuals who simultaneously meet the diagnostic criteria of several disorders of the same diagnostic rank (e.g. major depression, panic disorder and generalised anxiety). It has become common practice to list all of them as ‘comorbid’ without attempting to establish any hierarchical (temporal or causal) relationships among them. According to a recent review of the problem, ‘at least a third of all current cases in general population samples fulfil diagnostic criteria for more than one disorder’ [27]. If a ‘disorder’ is construed as a quasidisease entity (see above), these findings strain credibility. Either the nature of psychiatric illnesses is such that, as a rule, they tend to occur in clusters, or the diagnostic classification fails to discriminate between spurious comorbidity (mistaking the various aspects of the same clinical entity for independent disorders) and true coexistence of clinically independent conditions which is the original meaning of the term comorbidity [28]. An individual can, of course, be the host of two or more diseases and, consequently, be a member of two or more categories in medical classifications, but such multiple membership should be based on distinct and independent sets of characteristics. Current psychiatric classifications do not preclude multiple category membership on the basis of the same set of data since their categories are not mutually exclusive. For example, it is possible to code several ‘comorbid’ DSM-IV personality disorders on Axis II; or an individual can simultaneously meet the diagnostic criteria of both dysthymia and major depressive disorder on the basis of essentially the same symptoms, depending on their intensity, duration and sequence.
Cultural diversity
Current classifications tend to obscure the complex relationships between culture and mental disorder. Although both ICD-10 and DSM-IV acknowledge the existence of cultural variation in psychopathology (and the inclusion of a gloss on ‘specific culture features’ with many of the DSM-IV rubrics is a step forward), they regard it essentially as a pathoplastic influence that distorts or otherwise modifies the presentation of the ‘disorders’ as defined in the classification. Both systems ignore the existence of ‘indigenous’ languages in mental health [29].
Diagnoses as sampling criteria for clinical and epidemiological research
The use of restrictive DSM-IV or ICD-10 definitions, rather than broader clinical concepts, as sampling criteria in recruiting subjects for clinical or epidemiological research, carries the risk of replacing random error (due to diagnostic inconsistencies) with systematic error (due to a consistent exclusion of segments of the syndrome). For example, the DSM-IVrequirement of at least 6 months duration of symptoms plus presence of social or occupational dysfunction for a diagnosis of schizophrenia is likely to introduce a serious bias in the selection of populations for biological, therapeutic or epidemiological longitudinal studies. It would certainly make little sense to study the variation in course and outcome of a clinical sample that had been selected for prior chronicity in the first place, by applying the 6-month duration criterion.
Diagnoses as phenotypes for genetic research
Both the DSM family and, to a lesser extent, ICD-10, were welcomed and quickly adopted by researchers as rigorous diagnostic standards. Major studies on the molecular genetics of psychoses, usually involving collaborative consortia of investigators and a considerable investment of resources, are predicated on the validity of DSM-III-R or DSM-IV criteria. However, no susceptibility genes have been identified to date and very few of the weak positive linkage findings reported by some groups have been replicated by others [30]. In addition to the likely genetic heterogeneity of the psychiatric disorders across and within populations, it appears possible that ‘current nosology, now embodied in DSM-IV, although useful for other purposes, does not define phenotypes for genetic study’ [31]. In the absence of genes of major effect, the chances of detecting multiple genes of small or moderate effect depend critically on the availability of phenotypes defining a ‘pathognomonic lesion’ [32]. The polythetic nature of the present diagnostic criteria makes them unlikely to meet this requirement. Although restrictive in the sense of providing fixed cut-offs for the individual items on the list of attributes, the criteria are too broad in another, more fundamental sense (e.g. it is not impossible for patients to meet the diagnostic criteria for DSM-IV s c h i z o p h r e n i a without having a single symptom in common). Thus, the appearance of diagnostic standardisation, conferred by the use of DSM-IV or ICD-10 criteria in selecting patient samples for genetic and other biological research may be masking an unknown amount of phenotypic variation at the symptom and course levels. Such variation would appear as noise in genetic analysis and can nullify the potential power of the sample to generate high-resolution data. As a result, researchers are now looking for ways of dissecting the diagnostic entity into more homogeneous ‘correlated phenotypes’ or supplementing the clinical diagnosis with quantitative measures of phenotypic traits such as cognitive or neurophysiological dysfunction [33].
Adverse ‘side effects’ of current classifications
Notwithstanding the unresolved technical problems outlined in the preceding section, the advantages of having explicit diagnostic criteria and a rule-based classification outweigh substantially the alternative of having no such universal frame of reference in psychiatry. However, injudicious use of the current diagnostic and classification systems can have unintended negative educational and social consequences.
Such effects are already apparent. Medical students and specialist trainees may internalise the ‘reification’ fallacy [34] consisting in the uncritical belief that DSM-IV or ICD-10 represent the only and true ‘system’ of diseases in psychiatry. The provisional nature of the majority of the ‘disorders’ whose definitions are based on expert consensus and best estimate rather than on unequivocal evidence would thus remain insufficiently understood. The consequences are even more serious when this belief is adopted by teachers, editors of journals and peer reviewers, granting agencies and ‘purchasers’ of mental health services. The dogmatic ‘reification’ of DSM-IV and ICD-10 and the exclusion of alternative ways of looking at the psychiatric disorders can lead to a general impoverishment of the psychiatric culture and stifle conceptual innovation. A dulling of the natural interest in exploring variations in the clinical phenomena, studying atypical forms of illness, or describing new syndromes may already be present.
In other practical areas, the adoption of the current diagnostic and classification system of ‘disorders’ by the pharmaceutical industry in clinical trials, registration and marketing of products has resulted in a blurring of the distinction between symptomatic and pathogenetic therapies which may be detrimental to research. At the service provision level, the authority of internationally sanctioned classifications is already being invoked to justify, as ‘evidence-based’, essentially economic or political decisions about managed care, cost reimbursement and rationing of treatment [35].
Is the classification of psychiatric disorders a biological classification?
Classifying involves forming categories, or taxa, for ordering natural objects or entities, and assigning names to such categories. Ideally, the categories should be jointly exhaustive, in the sense of accounting for all possible individual entities, and also mutually exclusive, in the sense that the allocation of an entity to a particular category would preclude the allocation of the same entity to another category of the same rank. Regardless of continuing arguments between evolutionary systematics, numerical taxonomy and cladistics, their proponents agree that biological classifications reflect fundamental properties of biological systems and can be said to be ‘natural’ classifications. Psychiatric classifications and biological classifications are dissimilar in many important aspects. For example, the taxonomic units of ‘disorders’ in DSM-IV and ICD-10 do not form hierarchies and the classifications contain no supraordinate, higher-level organising concepts.
But what are thens the current psychiatric classifications and where would they belong in a ‘classification of the classifications’? Perhaps the closest analogue to current psychiatric classifications can be found in the so-called indigenous or ‘folk’ classifications of living things (e.g. animals in traditional rural cultures) or other material objects [36]. ‘Folk’ classifications do not consist of mutually exclusive categories and have no single rule of hierarchy (but may have many rules that can be used ad hoc). Such naturalistic systems seem to retain their usefulness alongside the more rigorous scientific classifications because they are pragmatic and well adapted to the needs of everyday life [37].
It is almost certain that DSM-IVand ICD-10 do not represent classifications in the usual sense in which the term is applied in biology. Essentially, they are augmented nomenclatures (i.e. lists of names for conditions and behaviours, supplied with explicit rules as to how these names should be assigned). As such, they are useful tools of communication and can play an important role in psychiatric research, clinical management and teaching. However, it is essential that neither DSM-IV nor ICD-10 be perceived as a complete, closed system of validated diagnostic entities in psychiatry.
Options for future classifications
It has been suggested that clinical neuroscience will eventually replace psychopathology in the diagnosis of mental disorders, and that the phenomenological study of subjective experience of people with psychiatric illnesses will lose its importance. Such transformation of clinical psychiatry would reproduce developments in other medical disciplines where molecular, imaging and computational tools have largely replaced the traditional clinical skills in making a diagnosis. In time, such developments would result in a completely redesigned classification of mental disorders, based on genetic aetiology [38]. The categories of such a classification and their hierarchical ordering may disaggregate and recombine the present clinical diagnostic entities in quite unexpected ways. Such a classification would finally approximate the ‘natural’ system of psychiatric disorders.
This, indeed, is already happening in general medicine where molecular biology and genetics are transforming medical classifications. New organising principles are producing new classes of disorders, and major chapters of neurology are being re-written to reflect novel taxonomic groupings such as diseases due to nucleotide triplet repeat expansion or mitochondrial diseases [39, [40]]. The potential of molecular genetic diagnosis in various medical disorders is increasing by leaps and bounds and it is unlikely to bypass psychiatric disorders.
Although the majority of psychiatric disorders appear to be far more complex from a genetic point of view than it was assumed until recently, molecular genetics and neuroscience will play an increasing role in the understanding of their aetiology and pathogenesis. However, the extent of their impact on the diagnostic process and the classification of psychiatric disorders is difficult to predict. The eventual outcome is less likely to depend on the knowledge base of psychiatry per se, than on the social, cultural and economic forces that shape the public perception of mental illness and determine the nature of the clinical practice of psychiatry. Apossible but unlikely scenario is the advent of an eliminativist ‘mindless psychiatry’ [14] which will be driven by biological models and jettison psychopathology. It is much more likely that clinical psychiatry will retain psychopathology as its core, and its relationship with subjective experience and behaviour. It is also likely that classification will evolve towards a dual system with two major axes: one aetiological, using neurobiological and genetic organising concepts, and another behavioural–dimensional or syndromal, which would be isomorphic to clinical reality. The mapping of two such axes onto one another would provide a stimulating research agenda for psychiatry for quite some time to come.