Towards ICD–11 and DSM–V: issues beyond ‘harmonisation’

Summary

In the revision process leading up to DSM–V and ICD–11, it is important to highlight both similarities and differences between the two classifications. Where irreconcilable conceptual differences are involved, these should be stated in a manner inviting future research to elucidate the advantages and disadvantages of alternative concepts or definitions. Eventually, both DSM–V and ICD–11 will need to be assessed against a set of benchmarks of validity and utility.

Assen Jablensky is Winthrop Professor in the School of Psychiatry and Clinical Neurosciences, the University of Western Australia, and Director of the Centre for Clinical Research in Neuropsychiatry at Graylands Hospital in Perth, Australia.

In an editorial published in the Journal nearly 15 years ago at the official release of DSM–IV, ¹ Professor John E. Cooper ² wondered whether future DSM and ICD versions would ‘serve, stimulate or torment the next generation of mental health workers, depending on their viewpoint’. With only a few years before the expected delivery of DSM–V (in 2012) and ICD–11 (2014), this question is as germane as ever.

In this issue, Michael First, one of the architects of DSM–IV, addresses the troublesome issue of ‘harmonising’ ICD–11 and DSM–V, a goal that is as worthy as it is hard to attain, considering the differences in the origins, purposes and scope of the two classifications. ³ Although both DSM–IV and ICD–10 ⁴ are widely regarded as documents purveying evidence-based credibility to the definitions and diagnoses of mental disorders, they have evolved in different contexts and address partially overlapping but different constituencies. The ICD, of which Chapter V, Mental and Behavioural Disorders is a part, is a statutory responsibility of the World Health Organization (WHO) aiming to provide a common language for the reliable reporting of all known diseases and health states across the world's populations. In contrast, the DSM is essentially a national diagnostic classification of mental disorders, developed by a non-governmental professional body – the American Psychiatric Association (APA) – but widely adopted by USA government agencies such as the Food and Drugs Administration and the Social Security Administration, as well as by the health insurance industry and the USA legal system. Since the introduction of explicit diagnostic criteria and a vastly expanded list of diagnostic rubrics, DSM–III ⁵ and its successors have gained international influence. In addition to a strong expert input, the DSM owes much of its success to a sustained financial support and worldwide marketing strategy. ⁶ Although DSM–III provided an important stimulus, the process involved in the development of ICD–10 was different. Wide consultations were held by WHO with clinicians and researchers representing clinical traditions and schools of thought from diverse cultures and socioeconomic environments. The process also involved collaboration with the designers of DSM–IV, which culminated in a major international conference. ⁷ Although ICD–10 was published 2 years earlier than DSM–IV, this collaboration achieved a convergence on many features of substance between the two classifications. Yet sameness was never intended and differences were inevitable, as highlighted by three further Journal editorials. ^8–10

The current process of preparation of the respective revisions of the two classifications is unique in its synchronicity and offers an opportunity to resolve lingering discordances through the establishment of a formal ICD–DSM Harmonization Coordination Group. First's overview of ‘opportunities and challenges’ should provide a helpful road map for the process.

Similarities and differences

First's analysis distinguishes between three groups of definitions and diagnostic criteria within the two classifications: a minority group of disorders where ICD–10 and DSM–IV essentially concur; a group of 23 disorders (36 sets of criteria) where differences are substantial and result from divergent conceptual perspectives; and a group of some 50 disorders where differences are mainly a matter of wording and therefore relatively trivial. Although the only disorder with a literally identical definition in ICD–10 and DSM–IV is transient tic disorder, the list of conditions with minor definitional differences reassuringly includes dementia, delirium, substance dependence, subtypes of schizophrenia, recurrent depressive disorder, panic disorder, generalised anxiety disorder, post-traumatic stress disorder and many others. However, the disorders where conceptual differences appear to raise important questions merit a comment.

In contrast to DSM–IV, which provides a single set of ‘operational’ diagnostic criteria for all users, ICD–10 was designed as a ‘family’ of interrelated versions addressing different users. Whereas the ICD–10 Clinical Descriptions and Diagnostic Guidelines ⁴ is the conceptual ‘core’ of the system, the Diagnostic Criteria for Research, ¹¹ and the WHO Guide to Mental Health in Primary Care ¹² are derivatives for use in specific contexts. First's analysis compares the DSM–IV criteria with the similarly structured ICD–10 Diagnostic Criteria for Research, rather than with the more flexible ICD–10 Clinical Descriptions and Diagnostic Guidelines. As a result, the diagnostic discordance between the two systems may appear to be greater than it actually is in everyday clinical practice.

One classification or two?

Both DSM and ICD will continue to produce new revisions and in some respects to compete with one another. It is, of course, confusing to have two rival classifications, particularly because many of the differences between them are trivial or accidental, as First points out. On the other hand, having two parallel classifications with some explicitly stated conceptual differences between them does help to highlight the provisional nature of many nosological concepts and their arbitrary definitions. Where irreconcilable conceptual differences are involved, these should be stated in a manner inviting future research to elucidate the advantages and disadvantages of alternative concepts or definitions. Eventually, both DSM–V and ICD–11 will be assessed against a set of benchmarks of validity and utility. ¹³

Criteria for assessing the ‘goodness of fit’ of future classifications ¹⁴

Clinical relevance and cognitive ease of use

The clinical relevance of a classification refers to its representative scope (or coverage), its capacity to describe attributes of individuals (such as clinical severity, impairments and disabilities), and its ease of application in a variety of clinical and cultural settings. By and large, the classification should allow ‘individualising’ the diagnostic assessment sufficiently to satisfy most clinicians and researchers. The system should be adaptable to the differing cognitive styles of its users, allowing clinicians to use the type of knowledge usually described as clinical experience or judgement, and to enable decisions under conditions of uncertainty, incomplete data and time pressure.

Reliability

DSM–III and its successors, as well as ICD–10, have undergone clinical trials that have shown them to be reliable. The diagnostic criteria of future classifications will be similarly field tested, but it is unlikely that reliability will remain a major goal – in contrast to issues of validity. However, reliability imposes a ceiling on the evaluation of validity in the sense that validity would be extremely difficult to determine if the diagnostic category were unreliable.

Categories and dimensions

Categorical models are firmly entrenched forms of representation for medical diagnoses, and have many practical and conceptual advantages – they are easy to use under conditions of uncertainty or incomplete clinical information. The disadvantage of the categorical model is its propensity to encourage a ‘discrete entity’ view of the nature of psychiatric disorders. Dimensional models, on the other hand, introduce quantitative variation and graded transitions between forms of disorder, as well as between ‘normality’ and pathology. Whether psychiatric disorders can be better represented dimensionally or categorically remains an open, researchable question. The likely difficulty with dimensional models stems from the lack of agreement on the number or nature of dimensions required to account for clinically relevant variation; the absence of an empirically grounded metric for evaluating severity or change; and their cumbersomeness in everyday clinical practice. These considerations may preclude, at least for the near future, a radical restructuring of psychiatric classification from a predominantly categorical to a predominantly dimensional model. However, categorical and dimensional models are not mutually exclusive, and it should be possible to supplement a refined categorical description of the clinical syndromes with selected quantitative traits assessing cognitive function, personality predispositions or tentative biomarkers of brain structure and function.

Meeting the needs of researchers

The performance of a classification as a research tool needs to be evaluated against requirements that are not always compatible. For example, the type of diagnostic criteria for biological research may not be the same as those suitable for epidemiological surveys. Studies of the molecular genetics of psychoses, usually involving collaborative consortia and costly resources, are typically predicated on a belief in the biological validity of DSM criteria. However, genome-wide association studies (GWAS), now involving over 10 000 individuals and an even larger number of healthy controls, have so far contributed little to the understanding of the genetic basis of common psychotic disorders. In addition to the genetic heterogeneity characterising many psychiatric disorders, it is possible that the current nosology does not define phenotypes optimal for genetic study. The disorders of current classifications may be phenotypic end-points for multiple pathogenetic pathways and pathoplastic interactions, masking substantial variation in the underlying neurobiological and cognitive endophenotypes. Future developments of diagnostic systems for research should supplement the clinical diagnoses with selected quantitative measures of brain structure, cognitive and neurophysiological dysfunction.

Meeting the needs of users of mental health services

A diagnostic classification has important implications for the well-being and human rights of those who are being diagnosed or classified. Avoidance of stigma is a crucial concern that needs to be taken into account when developing diagnostic classifications. This means that the reasoning behind psychiatric diagnoses and treatment decisions should be amenable to presentation in lay terms that are meaningful within the particular setting. A key requirement concerns the social needs and self-esteem of those who are diagnosed and their families.