Introduction

Personality disorder (PD) and its management present a clinical dilemma that community mental health teams (CMHTs) have struggled to deal with since their introduction into mainstream service provision for patients with mental illness in the early 1980s [21, 30]. Personality pathology is often under diagnosed, resulting in minimal management and a relative dearth of services directed at this problem, and within the UK only one country (i.e. England) attempts to provide a national personality disorder service [16]. This paucity of provision persists despite emerging evidence that personality disorder contributes greatly to long-term morbidity [12, 18, 33, 39]. The importance of this is accentuated by the high prevalence rates of personality disorder in small studies in secondary care, 40–52% in CMHTs [8, 13], and 92% in assertive community teams [20], compared with 5–12% in community sample [5, 6, 25]. Despite some scepticism over the significance of research diagnoses of personality disorder [38], these figures cannot be ignored.

But in ordinary community practice these findings are generally disregarded. This could be justified if the association of mental state and personality disorders was an artefact of common pathology, consanguinity rather than comorbidity [27], or if personality assessment was inaccurate in the presence of florid mental state abnormalities and improved pari passu with them [7, 23]. An alternative view is that the high prevalences reported are real and that personality status is an under recognised part of community team practice. If this is true, then the personality pathology identified in a secondary care population may be associated with an increased burden of psychopathology for a subset of CMHT patients. Our study therefore examined the extent to which personality pathology identified using a rapid assessment suitable for clinical practice could be identified in CMHT patients and whether such pathology was independently associated with total psychopathology irrespective of mental state disorder.

Method

Study population

The data used for this analysis were derived from the comorbidity of substance misuse and mental illness in community mental health and substance misuse services (COSMIC) study, a cross-sectional study of patients in secondary care in urban England. Only the data collected from CMHTs were analysed. Data were collected between January 2001 and February 2002 in four urban centres in England, the details of which are published elsewhere [37]. All four of the CMHTs studies conformed to the defined national standard practice of CMHTs [11]. Importantly, all patients were under the care of a consultant psychiatrist and had a care co-ordinator assigned to them. Ethical approval for the collection of data was provided by the LREC’s covering each of the four study centres.

Study sampling

A two-phase methodology was used to identify the patients included for the analysis. In phase one, a case-load census identified the sampling frame that was used to identify all patients attached to psychiatric services. All key workers were identified and sent a questionnaire to complete. In phase two, 400 patients from the CMHTs were selected at random to be interviewed by trained field-workers. These patients were selected at random using the Statistical Package for Social Services (SPSS) software [24] from which patient and key workers’ interviews were used to assess the patients’ psychiatric diagnoses, substance misuse and social situation. The data from the CMHT clients interviewed are reported here.

Research tools

A range of clinical assessment and research instruments were used to confirm mental state disorders, personality disorder diagnoses and psychopathology. All patients had a primary mental state and/or personality disorder diagnosis, confirmed by psychiatric assessment, recorded at interview. These were confirmed and quantified using standard peer reviewed instruments, the severity of which was assessed by the use of a rating instruments.

Depression severity was measured using the Montgomery-Åsberg Depression Rating Scale (MADRS) [14] and anxiety disorder impact measured using the Brief Scale for Anxiety (BSA) [34]. Overall psychopathology was measured using the Comprehensive Psychopathological Rating Scale (CPRS) [1]. The presence of any psychotic illness was confirmed using a standard approach involving review of the case notes (OPCRIT) [10] in a sub-sample that confirmed clinical diagnosis.

Personality disorder diagnoses were generated using the Quick form of the Personality Assessment Schedule (PAS-Q) [26], a rapid and easily applied form of the original schedule (PAS) [29], which records individual personality disorders but aggregates them into the three main clusters, the odd eccentric group (cluster A), the flamboyant or dramatic group (cluster B) or anxious and fearful group (cluster C). This involves the assessment of 24 personality traits, with the results recorded using a dimensional scale ranging from no personality disorder through to personality difficulty, simple and complex (diffuse) personality disorders, the latter comprising those patients who had personality disorders in two or more clusters. Training for assessing with the PAS-Q was carried out by one of the authors (SC), an experienced PAS trainer, as this short instrument requires training with the original Personality Assessment Schedule [28] before the data can be regarded as reliable. The PAS has been shown to have good psychometric properties [28], high cross-national inter-rater reliability [29] and independent support for its validity in patients with mental state disorders longitudinally [35, 36].

These instruments were chosen as they were fast and reliable to administer, valid and had been published in peer review journals.

Statistical analysis

The primary aims of the analyses were threefold:

  1. 1.

    To describe the prevalence of personality pathology within a secondary care setting.

  2. 2.

    To examine the extent to which personality and mental state disorder contribute to global pathology.

  3. 3.

    To examine the extend to which the severity of personality disorder was accounted for by comorbid mental state disorder.

To examine the first aim, the data set was analysed comparing personality disordered patients both generally and by cluster to non-personality disordered patients. These comparisons assessed for unadjusted differences by location, age, sex, ethnicity and mental state disorder. Two-tailed Chi-squared tests were used to compare the proportions of patients living in London to those outside London, age, sex and ethnicity. Odds ratios were generated to assess the unadjusted size of the correlation between mental state disorder and personality disorder.

Linear regression modelling was used to compare the impact on global psychopathology of mental state disorders and personality severity as defined by Tyrer and Johnson [32]. Linear regression was used to take into account the effects of each diagnosis on the other(s). CPRS was entered as the dependent variable and each diagnosis as independent variables. The assumptions of linear regression were tested to ensure the model was appropriately applied and a goodness of fit of the data to the model checked. The degree of variance of the CPRS was assessed using R 2.

To assess the impact of mental state diagnosis on personality severity, ordinal regression modelling was used. This is similar to linear regression but makes no assumption about the statistical ‘size’ of the differences in the dependent variables steps. In this case, the clinical order of personality severity is known (i.e. a score of 2 is greater than 1) but the size of the difference is not (i.e. a score of 2 is not necessarily twice as severe as a score of 1). This prevents the use of simple linear regression; however, this is taken into account by ordinal regression. Personality severity was entered as the dependant variable and mental state disorders entered as the independent variables. The assumptions of ordinal regression were tested to ensure the model was appropriately applied and a goodness of fit of the data to the model checked. Furthermore, an estimation of the degree of variance in the model was assessed using Nagelkerke pseudo-R 2. This allows the variability of the severity of personality pathology accounted for by mental state disorder to be assessed. A Nagelkerke pseudo-R 2 approach was chosen to assess variance as the total variability of the model was considered to be clinically less relevant than the variability compared to the best fit model.

The data set was analysed using the SPSS v14.0. Due to the nature of descriptive studies and the large data set available to examine, the risk of a type one error is significantly increased. To overcome this, data other than general prevalence data were only analysed by clustered groups and a level of statistical significance of at least P = 0.01 level was regarded as necessary to be significant.

Results

Number and characteristics of subjects

We obtained demographic and casemix data on 2,528 of 2,567 psychiatric patients (98.5%) meeting census eligibility criteria. Interviews were completed in 282 of 400 cases (70.5%) randomly sampled from this population. Table 1 shows that the study population was predominantly male (57%) with a median age of 40. Psychotic disorder was present in 76.6% (n = 216) of cases. There were no clinically or statistically significant differences between the demographic and casemix profile of the achieved sample and sampled non-respondents, or between the achieved sample and the total treatment population from which the sample was drawn.

Table 1 Demographic characteristics of the secondary care sample
Full size table

Prevalence of personality disorder

The general prevalence of personality disorder (i.e. a diagnosis of at least one personality disorder) in the total sample was 39.4% (95% CI 34–45%) with no difference between London and non-London centres (Table 2). The commonest personality cluster was cluster C (anxious/avoidant), which represented over a quarter of the entire sample (25.5%, 95% CI 20–31%), whilst cluster B pathology constituted almost a one-fifth of the sample (18.4%, 95% CI 14–23%). There was a significantly higher prevalence of anxious personality disorder in non-London centres and a higher prevalence of emotionally unstable (impulsive) personality pathology in London centres. However, personality disorder by cluster or severity did not vary between London and non-London centres.

Table 2 General prevalence of personality pathology in London and non-London cohorts [n and (% of total sample by area)]
Full size table

The association of personality disorder to age, sex or ethnicity did not show any significant variation to the non-personality disordered population. This remained true for the individual clusters also (Table 3).

Table 3 Demographic data by personality status
Full size table

Association of personality disorder and mental state disorders

The distribution of personality disorder was correlated with mental state diagnoses (see Table 4). In this sample patients with a personality disorders were five times less likely to have a psychotic disorder than the non-personality disordered patients in this secondary care sample (OR 0.23, 95% CI 0.13–0.41). When patients were examined by cluster, this negative correlation was not found in the cluster A group (OR 0.85, 95% CI 0.29–2.44), but remained for clusters B and C disordered patients. Unlike the psychotic patient group, personality disorder increased the probability of co-morbid severe depression fivefold (OR 6.15, 95% CI 3.10–12.22), although this association was again not found in cluster A disorder. Personality pathology was also strongly correlated with anxiety disorders, this association being true of clusters A and B patients as well as those with cluster C disorder. A diagnosis of drug or alcohol dependence was not associated with a personality disorder in general or by cluster. More general drug and alcohol diagnoses were not examined to minimise the possibility of Type I error.

Table 4 Relationship between personality disorder by cluster and Axis I diagnosis
Full size table

The hypothesis that personality psychopathology lacked independence from mental state disorders (i.e. is a confounder) was assessed by linear regression modelling. This modelling accounted for 83% of the variance in overall psychopathology severity with an excellent fit to the data set (P < 0.001). Personality disorder, depression and anxiety independently contributed to global psychopathology (P < 0.01). Psychosis (P = 0.039) and alcohol dependence (P = 0.011) did not reach the pre-determined levels of statistical significance to contribute to psychopathology independent of other diagnosis. Personality disorder appeared to act independently of mental state disorders leading to a greater degree of global psychopathology than either psychosis or alcohol or drug dependence in isolation (Table 5).

Table 5 Linear regression modelling of the effect of axis I and II diagnoses on global psychopathology
Full size table

Impact of mental state disorder on personality severity

To further clarify the relationship between personality and mental state diagnosis, ordinal regression modelling was used to assess whether having a specific mental state diagnosis was associated with greater personality disturbance. Psychotic disorder, depression, anxiety disorder, alcohol dependence and drug dependence were entered as independent variables for the severity of recorded personality pathology. Although the model fitted the data set well (goodness-of-fit Pearsons χ 2 = 40.78, P = 0.57) only 18.2% of the variation in the severity of personality disturbance was accounted for (Nagelkerke pseudo R 2 = 0.182). Only an anxiety disorder (P = 0.01) and psychosis (P = 0.002) acted as a predictors of personality disorder severity. These effects were positive and negative, respectively. Severity of personality disturbance is not accounted for by mental state disorder and acts independently of it (Table 6).

Table 6 Ordinal regression modelling of the impact of Axis I pathology on Axis II severity
Full size table

Discussion

Although there are many papers reporting on individual aspects of the rates and impact of personality pathology in the literature, this is the first to our knowledge that provides information combining prevalence, comorbidity, and the effects of mental state and personality disorder in a single large interviewed UK sample. The evidence that personality pathology largely contributes to total pathology independently of mental state disorder is strong and suggests that a significant proportion of patients in secondary care settings need treatment and management directly targeting their personality difficulties.

Prevalence of personality disorder

The overall prevalence of personality disorder was similar to that found previously in community health teams [8]. There was little difference between London and non-London centres apart from somewhat higher rates of borderline PD in the London centres compared with non-London ones (P = 0.085). The higher prevalence of cluster B diagnosis in London may be due to the transient nature of the population and we already have evidence that these personalities are more common in urban areas and seek more treatment [3, 22]. The reasons for the association between cluster C personality disorder and non-London centres are not entirely clear, but the higher rate of substance misuse in London skew services towards the cluster B group which is particularly prevalent in this group [15]. The relative instability of some personality diagnoses within individual patients weakens these cross-sectional findings, but nonetheless the findings appear to be robust.

Within this clinical population there is no indication that the rates of personality pathology vary with age, gender or ethnicity. Taken as a whole, this suggests that the distribution of personality disorder is evenly spread across age, sex, ethnicity and type in secondary care. Further analysis, for example, by individual categorical disorder, may highlight variability within particular categorical diagnosis (e.g. borderline PD) but is unwarranted as this would lead to an abundance of data and significantly increase the risk of type I error (“data trawling”).

Association of personality disorder with mental state disorder

The strong association between mental state and personality disorder may lead to the view that the CPRS score, as a measure of Axis I pathology, should not be regarded as independent of personality pathology. It is well established that there is significant comorbidity between pathology in these groups [31] but there is much additional evidence that they are independent and can be separately recorded [4]. This is also illustrated by the variation in PD prevalence in different diagnostic groups. Personality disorder was found to be more common in patients with an anxiety disorder and much less common in patients with psychosis, and this would not be predicted if personality pathology was regarded as an epiphenomenon associated with greater severity of illness.

The variation in personality pathology is worthy of separate comment. Brief psychotic episodes are a feature of some cluster B (mainly borderline) presentations and cluster A disorders share similar phenomenological and pharmacotherapeutic features as mental state psychosis [17]. This finding is, perhaps a reflection on the heterogeneous nature of reported personality pathology in psychosis and the difficulties in accurate diagnosis in this population [19]. Similarly, the overlap between mental state anxiety disorders and cluster C personality disorders is expected due to the similarity between some of the features of each. Nevertheless, a mental state anxiety disorder also increases the odds of clusters A and B disorder approximately fivefold. This again raises the question as to whether personality disorder leads to psychopathology independently of major mental disorder or if, in fact, it is a confounding factor. We are confident that the regression modelling contradicts the latter view. First, it can be seen that when personality is considered as a component of overall psychopathology, it acts as an independent factor to mental state disorder. Secondly, despite modelling showing a good fit to the data set, only 18% of the variance in personality disorder severity is accounted for by mental state disorders. Therefore, 83% of personality disorder severity is due to other causes.

Limitations

We acknowledge that, as with all studies, this paper has a number of limitations. First, the sample used is a pragmatic one generated from urban England and as such may not be representative of non-England or rural settings. It is, however, a real life “snap-shot” of patients seen in routine care and as such provides information about ‘everyday’ (as opposed to tertiary/research or clinical trial) patients. Secondly, this data-set analysed is a cross-sectional survey. As such, although associations can be made, putative causative or correlative factors are speculative at best. Thirdly, this is a secondary analysis and therefore more susceptible to type one error (finding associations by chance alone). To minimise the risk of this, only a priori analyses were undertaken and the limit for significance raised to P = 0.01. Finally, there remains debate about the diagnostic validity of personality disorder and the tools used to assess it. This remains problematic for many mental disorders but is perhaps more pronounced in the personality disorder diagnoses. For this study, a well-validated tool was used and all interviewers received formal training in its use. Although all formal personality disorder tools yield higher rates of personality disorder than clinical assessment, the PAS is more conservative than many. Despite this nosological debate, this study shows us that personality disorder (and severity) as assessed by PAS does provide useful clinical information on psychopathology and its impact is not accounted for by mental state disorder. This perhaps is an argument for its inclusion in everyday clinical practice.

Conclusions

What is the relevance of these results for the practising psychiatrist? First, and perhaps most importantly, two in every five patients engaged in routine secondary care are likely to suffer from personality disorder which produces significant inter and intrapersonal distress and disability. Personality disorder is therefore common in this setting and carries its own burden of psychopathology. These effects are independent of the mental state disorders usually earmarked as the key diagnoses of patients in CMHTs. Secondly, personality disorder, although common, is rarely very severe and is therefore more likely to go undiagnosed. By missing the diagnosis, it is impossible to account for its effects in the management and prognosis of the patient, although research shows us the impact it has in day-to-day practise, such as admissions to hospital [9]. Thirdly, personality disorders contribute inexorably to overall psychopathology independently of mental state diagnosis, a message which all clinicians must heed. This may also explain the higher rates of personality disorder in secondary care settings, as patient’s increased levels of distress lead to increased referral rates. Finally, if ignored, personality disorder is not adequately assessed and managed and may account for the poor response rates of many patients in secondary care to apparently optimal management for their mental state disorder. It is already known that personality abnormality is associated with the emergence of mental state disorder in community settings, and to affect outcome [2]. Therefore, the assessment and management of personality disorder within the secondary care setting is likely to significantly improve overall prognosis for a substantial proportion of patients.