The categorical approach

Both ICD–10 and DSM–IV identify categories of personality disorder. In keeping with their medical origins, the two schemes promulgate a system for diagnosing personality disorder that is categorical in nature: people are thought either to have a personality disorder or not to have one. The categorical approach has a two-component structure – generic criteria to make a diagnosis of personality disorder, and specific criteria for the different types of the disorder. The generic criteria (Box 1) seek to separate personality disorder (DSM Axis II disorders) from other mental disorders (DSM Axis I disorders), whereas the disorder-specific criteria attempt to distinguish different types of personality disorder (e.g. borderline and narcissistic) from one another.

Critique of categorical systems

There are major problems with our current nosology – especially with the DSM system, which will be the main focus of the remainder of this article. Establishing diagnoses and clusters in a categorical manner may lead to greater agreement and communication between clinicians (increased reliability), but it does not enhance the fundamental understanding of disorders (no increase in validity).

One of the supposed advances of DSM–III (American Psychiatric Association 1980) was the introduction of the multiaxial system, which separated the newly classified Axis I disorders (which were considered to be transient disorders of state) from Axis II disorders (deemed to be more enduring and dependent on the abnormal traits that the individual possessed). Part of the rationale for this system was that it would force clinicians to consider assessing personality disorder, even in patients with an Axis I condition. By so doing, it was hoped that clinicians would take personality disorder more seriously in their clinical practice and that research into personality disorder would also be promoted (Reference Millon and FrancesMillon 1986). However, many conceptual difficulties remained and have yet to be resolved.

First, the distinction between Axis I and Axis II disorders is not borne out by empiricism. This is because the level of comorbidity is so high, especially for some disorders (e.g. borderline personality disorder with depression, antisocial personality disorder with substance misuse), that the distinction is vitiated.

Second, symptoms specific to personality disorders are continuously distributed across both clinical and healthy samples (Reference Livesley, Schroeder and JacksonLivesley 1994). Consequently, diagnostic ‘disorders’ reflect an arbitrary threshold and not true disorders (Reference BlackburnBlackburn 2000).

Third, personality disorder diagnoses often show poor psychometric properties such as validity and reliability (Reference Blais, Benedict and NormanBlais 1998) because current criteria have been selected by clinical consensus rather than empirical analysis (Reference LivesleyLivesley 2007). The ICD has a lower threshold for making a diagnosis (Reference Tyrer and JohnsonTyrer 1996) but the DSM, a more rigid system advocating a checklist approach, identifies a higher number of personality disorders: 11 by DSM–IV as opposed to 8 by ICD–10.

Fourth, and related, the diagnoses have limited clinical utility, not helping practitioners to choose between pharmacological or psychotherapeutic interventions (Reference Sanderson, Clarkin, Costa and WidigerSanderson 2002; Reference Tyrer and BatemanTyrer 2004).

Fifth, owing to ‘loose’ taxonomic criteria, the nomenclature and number of personality disorders have changed with each new edition of the DSM and ICD, further undermining practitioner confidence. With the revision of DSM–III–R to DSM–IV, some personality disorders (e.g. sadistic and self-defeating) disappeared entirely, whereas others (e.g. passive–aggressive) were removed to the appendix. The disappearance of sadistic personality disorder was largely a consequence of political pressure from feminists who wished to remove what they considered to be a psychiatric loop-hole that might exculpate some extreme (male) offenders (Reference Stone and SkodolStone 1998).

The categorical approach has been popular because it is simple to operate and fits with a medical model of disease, establishing clear boundaries between normal and abnormal functioning. In a social welfare system of democratic governance this is important in terms of resource allocation, prioritising of services and identifying suitable individuals for receipt of interventions. The deficits of the categorical system are probably central to the belief among many clinicians in the UK that personality disorder is not a ‘real’ disorder and that those with personality disorder are so different from people with mental illness that generic mental health services can and should exclude them. Such beliefs and attitudes have led to a crude form of resource allocation within mental health services in the UK that has often tended to ‘reserve’ services for people with chronic and severe forms of psychoses and mood disorders, most often excluding as untreatable those with personality disorder (Department of Health 2003).

The dimensional approach

If the individual, interpersonal and group aspects of personality functioning are emphasised in a diagnostic system, personality will be seen to involve a number of different capacities – dimensions, domains or traits – operating at different times and in different settings. Allport first emphasised the role of ‘traits’ in the makeup of personality as the ‘dynamic organization … of those psychophysical systems that determine characteristics of behaviour and thought’ (Reference AllportAllport 1955). The most influential model of normal personality was proposed by Eysenck. It focuses on the individual's intra-personal characteristics rather than social interactions and describes personality in terms of three dimensions of higher-order traits or ‘superfactors’: psychoticism, extraversion and neuroticism – the PEN model (Reference Eysenck, Pervin and JohnEysenck 1990). A modern extension of this model has five dimensions: neuroticism, extraversion, agreeableness, openness and conscientiousness – the so-called ‘big five’ or the five-factor model of personality (Reference McCrae and CostaMcCrae 1987). These dimensions can be measured reliably and, with the exception of openness, all the factors have been replicated across cultures and shown to be moderately heritable (Reference Bouchard and LoehlinBouchard 2001).

There is overwhelming empirical support for a dimensional representation of normal personality (Reference Widiger, Frances, Costa and WidigerWidiger 1994; Reference Clark, Livesley and MoreyClark 1997; Reference LivesleyLivesley 2007). Many dimensional or trait models of personality exist but most of these collapse into three (extraversion, neuroticism, psychoticism: Table 1), four (emotional dysregulation, dissocial behaviour, inhibitedness, compulsivity: discussed below) or five (neuroticism, extraversion, agreeableness, openness and conscientiousness) basic structures.

Personality disorder within a dimensional model of understanding might represent an extreme position on a personality continuum (Reference BlackburnBlackburn 2000). Although this has an intuitive appeal, Reference LivesleyLivesley (2007) has argued that for a disorder to be present, then more is required than an extreme position on a continuum. Instead, he argues that it is the failure to accomplish one or more of certain life tasks that needs to be present for personality to be regarded as disordered. We will return to this when we come to discuss the general features of personality disorder.

Critique of dimensional systems

There are advantages to using a dimensional approach. First, it would fit with other accounts of chronic developmental disorders, which assess both vulnerability and resilience factors, and reframe personality disorder as a disability rather than a disease (Reference FulfordFulford 1989; Reference AdsheadAdshead 2001). Second, it would help to limit the reductionist and rather stigmatising approach to personality disorder, whereby those with the disorder are seen as having a ‘lifelong’ condition that is impervious to change. Third, treatment selection would be informed by existing evidence base: some degrees of disordered personality dimensions will, and clearly do, ameliorate both with time and the appropriate interventions. However, dimensional schemes are simply too complex for everyday use, as substantial knowledge and clinical ability are required to identify the wide range of traits, many of which fall below a threshold for determining abnormality (Reference FirstFirst 2005). When assessment of abnormal personality is required, much time and effort might be spent in assessing the normal aspects while clinically useful constructs such as suspiciousness, insecure attachment, self-harm and narcissism are overlooked.

An integrated diagnostic system for personality disorders

John Livesley has been one of the most influential critics of the current psychiatric nosology and of the DSM in particular. What we find particularly attractive about his suggested revisions is his attempt to integrate into the existing system solutions to many of the criticisms targeted at it. This integration is crucial for two reasons. First, if one were to replace DSM–IV (or ICD–10) with a completely different classificatory system, it would be impossible to draw inferences from research knowledge which is based on the current systems. We would in effect know nothing about the epidemiology of personality disorders, their naturalistic course or their treatment. Second, at a pragmatic level, there would be no reason for the body of practitioners to switch suddenly from one classificatory approach to another, especially as many of the advantages of any new system would be largely theoretical and await empirical verification. Therefore, why would anyone wish to change? Hence, it would be far better to integrate any new system into the existing DSM, as there is too much now invested in the latter to allow its complete replacement.

Here, we provide a brief distillation of the work of John Livesley and his colleagues, but we would strongly recommend the interested reader to refer to their many original contributions (Reference Livesley, Schroeder and JacksonLivesley 1994, Reference Livesley, Jang and Vernon1998, Reference Livesley, Phillips, First and Pincus2003a ,Reference Livesley, Jang, Vernon, Millon, Lerner and Weiner b , Reference Livesley2005, Reference Livesley2007). Although much of Livesley's work focuses on the arguments for a dimensional rather than a categorical classification, this will not be our main focus. Rather, we wish to concentrate on Livesley's general approach and on certain of his crucial changes of emphasis that address many of the criticisms described above.

Livesley's work re-directs the clinician to the two-step approach that the DSM recommends – application first of the general criteria (Box 1) and then of the specific criteria – which appears to have been lost in current clinical practice. What happens currently is a ‘bottom-up’ approach: first, individual personality traits are assessed and then these are grouped into the various categorical disorders. In contrast, Livesley recommends a ‘top-down’ approach whereby the first decision to be made is whether the individual has a personality disorder or not. If the answer to this question is in the affirmative, two further tiers of investigation of increasing specificity may be applied if required. The important point is that the process is hierarchical, proceeding from the general to the specific, rather than the other way round. We will now briefly expand on this process, as described by Reference LivesleyLivesley (2007).

A three-step, top-down approach

In defining the general features of personality (and personality disorder), Livesley takes an evolutionary perspective and suggests that there are three life tasks that individuals need to carry out as evidence that they are adapted to their environment. Failure or difficulty in meeting one or more of these tasks is a general sign of personality disorder. The three areas and corresponding potential failures are:

1. • achieving a coherent sense of self (intrapersonal failure)

2. • developing intimacy in interpersonal relationships (interpersonal failure)

3. • behaving prosocially (social group failure).

It is important to recognise that these general features of personality disorder, unlike the secondary domains and primary traits that we shall describe further below, are purely social constructs. Livesley proposes that with a few simple screening questions focused on each of these three ‘general’ areas, it ought to be possible to decide whether or not someone has a personality disorder (Fig. 1, step 1). These questions might be along the lines of ‘Do you have a clear sense of yourself and what you wish to accomplish in your life?’ (the intrapersonal domain); ‘Do you find it difficult either being too close or being very detached from important people in your life, so that relationships are inevitably problematic?’ (the interpersonal domain); ‘Do you find it difficult to conform to the expectations that your family, friends or society at large have of you, so that you are quite often at odds with them?’ (the prosocial domain). It is possible that answers to the last two questions may not be honest, but it is likely that other sources of information (such as key informants or documentary evidence from statutory agencies) may be available to inform one's conclusions.

Proceeding from these general features of personality to the two lower levels in the hierarchy, Livesley commences his discussion of secondary domains by drawing attention to one of the most robust findings in the field of personality disorder. That is, when individual personality traits are subjected to factor analysis, four domains invariably emerge, so that ‘the robustness of the 4 factor model across clinical and nonclinical samples, cultures, and measurement instruments suggests that it reflects the biological organisation of personality’ (Reference Livesley, Jang and VernonLivesley 1998). This is where ‘nature is carved at its joints’, as these four domains of phenotypes are closely correlated to four genetic factors. Indeed, Livesley defines a secondary domain as ‘a cluster of traits influenced by the same general genetic factor’ (Reference LivesleyLivesley 2007). Livesley labels these four domains emotional dysregulation, dissocial behaviour, inhibitedness and compulsivity (Table 2). These dimensions correspond to Reference Mulder and JoyceMulder & Joyce's (1997) nomenclature of the four ‘As’ of personality: asthenic, antisocial, asocial and anankastic.

It is important to recognise that some of these higher-order categories/domains comprise entities that encompass a broader array of traits than is implied by the domain name. For instance, Mulder & Joyce's antisocial secondary domain includes not only simple rule-breaking and criminal behaviour but also features of suspiciousness, paranoia and narcissism. The asthenic domain includes not only those with anxious dependent traits but also traits of emotional dysregulation. The asocial domain similarly includes both anxious avoidance and schizoid traits. Finally, the anankastic domain comprises obsessive–compulsive personality traits. We recognise that these four higher-order categories or domains do not map easily onto the three clusters in DSM–IV (Box 2). The antisocial domain, for instance, includes features not only of personality disorders in Cluster B but also of those in Cluster A. The asthenic domain includes features of both Cluster B (emotional dysregulation) and Cluster C (anxious avoidant). The asocial domain comprises traits that occur in both Cluster A (schizoid) and Cluster C (avoidant). Step 2 of the evaluation (Fig. 1) would involve categorising someone with a personality disorder into one (or more) major domain of dysfunctions. This would allow clinical ‘clustering’.

Reference LivesleyLivesley (2007) argues that the four secondary domains are composed in turn of a number of primary traits, which he defines as ‘a cluster of behaviours influenced by the same general and specific factors’. These primary traits are ‘the fundamental building blocks of personality and hence the basic unit for describing and explaining personality disorder’. Through factor and behavioural genetic analysis, Livesley identifies 30 such primary traits, divided unequally between the four secondary domains (Table 2). Step 3 of the evaluation would involve detailed assessment of the particular dysfunctional domain and would provide behavioural and phenomenological markers (Table 3).

Advantages of an integrated system

As already stated, we believe that an integrated classification system such as that offered by the three-step, top-down evaluation model outlined in Fig. 1 meets many of the objections levelled at the DSM system in particular.

First, both the primary traits and the secondary domains are empirically derived and so can be tested with a rigour that is currently impossible.

Second, they make clinical sense, with a focus on the personality traits rather than on behaviour. An obvious example is the antisocial higher-order factor, which includes many traits that are recognised by clinicians, such as ‘suspiciousness’, ‘narcissism’ and ‘hostile dominance’ in the presentation in addition to rule-breaking behaviour and conduct disorder. This moves the description of antisocial personality disorder away from simple criminality to encompass broader features of dissocial personality disorder.

Third, the top-down and hierarchical structure provides the evaluation model with a flexibility in application that current systems lack. For instance, if the question is ‘Does the individual have a personality disorder?’, the answer is provided by screening for three general criteria of personality disorder (i.e. the ability to form an intimate relationship, to act prosocially and to have a sense of identity). If the answer is ‘No’ to any one or more of these criteria, the four secondary domains can be examined with a few further screening questions to discern the predominant features in each (Table 3). It is only when more detailed information is required to identify the specific primary traits that a detailed inquiry has to be made. Even then, Livesley's proposal is parsimonious, as it requires the assessment of only 30 traits, compared with the 79 in DSM–IV–TR: a reduction of 62% (Reference LivesleyLivesley 2007).

Fourth, the three-step, top-down model is arguably more comprehensive than the current system, so that fewer individuals are placed in the ‘not otherwise specified’ category, a problem with the current system (Reference Verheul and WidigerVerheul 2004).

Finally, and perhaps most importantly, as the four secondary domains differ aetiologically, they should have differing courses and implications for treatment. Detailed consideration of treatment implications is beyond the scope of this article and will not be addressed here.

Implications of an integrated diagnostic system

Conclusions

There are problems with the diagnosis of personality disorders using current classificatory systems which, in our view, neither an entirely categorical nor an entirely dimensional approach will be able to rectify. We believe that the best way forward is to incorporate aspects of both approaches into the integrated diagnostic system that we have outlined here. The strength of this approach would be to align prototypical data (descriptive behavioural and phenomenological information) with genotypically grounded empirical data. We have adapted Livesley's dimensional approach and revealed how this can explain certain seemingly irreconcilable difficulties within current classificatory systems related to severity of personality disorder, adult onset, the ‘not otherwise specified’ category and multiple comorbidities of personality disorders. It remains to be seen whether the DSM–V and ICD–11 task groups take up these challenges. We are confident that adopting such an approach by the busy clinician will repay the time and effort invested in it.

MCQs

Select the single best option for each question stem

1. 1 With regard to personality disorders, difficulties with current classificatory systems include:

   1. a inadequate capture of core personality pathology

   2. b good discrimination between different personality disorders

   3. c good reliability

   4. d clear influence on treatment selection

   5. e all of the above.

2. 2 The categorical approach to diagnosing personality disorders:

   1. a adopts a three-component structure

   2. b identifies people who also have neurotic disorders

   3. c offers a severity rating

   4. d is based on psychological foundations in terms of nosology

   5. e is adopted by both ICD–10 and DSM–IV.

3. 3 Personality disorders:

   1. a are developmental in origin and relatively inflexible

   2. b are not markedly different from cultural norms of thinking, feeling and behaving

   3. c can be due to other medical, mental or substance use related disorders

   4. d are never maladaptive patterns of relating with others

   5. e are episodic in nature and easily managed.

4. 4 There are:

   1. a eleven personality disorder diagnoses in DSM–IV

   2. b ten personality disorder diagnoses in ICD–10

   3. c lower thresholds for making a diagnosis in the DSM system

   4. d higher thresholds for making a diagnosis in the ICD system

   5. e many personality disorders, but the total number has remained constant over time.

5. 5 Dimensional diagnostic systems:

   1. a have their origins in medical traditions

   2. b do not provide for a range of severity

   3. c are meant for everyday use

   4. d are favoured by the current international classificatory systems

   5. e do not reflect the reality of clinical presentations any better than categorical systems.

MCQ answers

1

a

2

e

3

a

4

a

5

a