Blackwell Science, LtdOxford, UKADDAddiction0965-2140 2004 Society for the Study of Addiction

99
Original Article
Prevalence of personality disorder
Owen Bowden-Jones
et al.

RESEARCH REPORT

Prevalence of personality disorder in alcohol and drug

services and associated comorbidity
Owen Bowden-Jones1, Muhammad Z. Iqbal2, Peter Tyrer3, Nicholas Seivewright2, Sylvia Cooper4,
Ali Judd5 & Tim Weaver5 on behalf of the COSMIC study team
Central and North-west London Mental Health NHS Trust, London,1 Community Health Sheffield NHS Trust, Sheffield,2 Department of Psychological Medicine,
Imperial College, London,3 Nottinghamshire Healthcare NHS Trust, Nottingham4 and Department of Social Science and Medicine/Centre for Research on Drugs
and Health Behaviour, Imperial College Faculty of Medicine, London, UK5

Correspondence to:
Dr Owen Bowden-Jones
Drug Treatment Centre
Chelsea and Westminster Hospital
369 Fulham Road
London SW10 9NH
UK
Tel: 0208 8466111
Fax: 0208 8466112
E-mail: owen.bowdenjones@nhs.net
Submitted 20 May 2003;
initial review completed 21 July 2003;
final version accepted 17 May 2004

RESEARCH REPORT

ABSTRACT
Aims To compare the prevalence of personality disorder in alcohol and drug
populations with special attention to its impact on psychopathology and service
characteristics.
Design Cross-sectional survey.
Setting Three alcohol and four drug services in four urban UK centres.
Participants Two hundred and sixteen drug and 64 alcohol service patients
randomly sampled from current treatment populations.
Measurements A treatment population census recorded demographic and
diagnostic data. Patient interviews assessed the presence, cluster type and
severity of personality disorder using the Quick Personality Assessment Schedule (PAS-Q). Other psychopathology was measured using the Comprehensive
Psychopathological Rating Scale (CPRS). A case-note audit recorded psychotic
psychopathology using the OPCRIT schedule and data regarding social
morbidity.
Findings The overall prevalence of personality disorder was 37% in the drug
service sample and 53% in the alcohol service sample. The distribution of
severity and clusters differed markedly between the two samples. There was a
significant association between the severity of personality disorder and psychopathology in both samples. Levels of morbidity associated with clusters B and C
were similar. Clinical diagnosis of personality disorder showed high specificity
but low sensitivity when compared to PAS-Q.
Conclusions In both alcohol and drug service populations, personality disorder is associated with significantly increased rates of psychopathology and
social morbidity that worsens with increasing severity of the disorder. Despite
this, personality disorder is poorly identified by clinical staff. The PAS-Q may be
useful as a clinical assessment tool in the substance misuse population for the
early identification and management of patients with personality disorder.
KEYWORDS
substance use.

Alcohol, comorbidity, personality disorder, prevalence,

INTRODUCTION
Studies of substance misuse service populations have
shown prevalence estimates for personality disorder
ranging between 40% and 100% (Seivewright & Daly
2004 Society for the Study of Addiction

1997; Nadeau et al. 1999; George & Krystal 2000). This

wide variation in prevalence estimates has been attributed to differences in study method, rating instruments
and treatment population (Verheul 2001). Drug service
populations generally have higher rates of personality

doi:10.1111/j.1360-0443.2004.00813.x

Addiction, 99, 13061314

Prevalence of personality disorder

disorder than alcohol service populations (DeJong et al.

1993; Verheul et al. 1995).
The importance of accurate identification of personality disorder in routine substance misuse service populations has been highlighted by epidemiological studies
that have demonstrated associations between personality
disorder and a range of negative outcomes. These include
higher rates of dropout from treatment (Reich & Vasile
1993), poor social function (Darke et al. 1994), increased
rates of crime (Hernandez-Avila et al. 2000) and
increased prevalence of comorbid Axis I psychiatric disorder (Rousar et al. 1994). This epidemiological evidence
has led to a widely held clinical view that substance misuse patients with personality disorder have poorer prognoses in routine practice (Links et al. 1995). However, the
clinical picture is complicated by evidence from a number
of treatment studies which suggest that personality disorder may not have negative effects on outcome, particularly with respect to therapeutic community treatment
(Clopton et al. 1993; Brooner et al. 1998; Messina et al.
2002) and that schizoid personality disorder may even be
associated with a better outcome (Griggs & Tyrer 1981).
In mental health treatment populations the assessment of personality disorder often shows marked differences between clinical and research assessments, with
the latter showing higher prevalence (Westen 1997).
There is a major problem in disentangling comorbidity
of mental state and personality disorder, particularly for
the anxious and fearful group of personalities (cluster C)
whose features overlap considerably with anxiety and
depressed mood (Hassiotis et al. 1997). In substance
misuse populations the validity of routine clinical diagnosis of personality disorder has not been well
researched. This is a significant omission, given that
there are particular difficulties in interpreting personality pathology in substance-misusing patients. Even
patients without personality disorder may often exhibit
enduring patterns of maladaptive behaviour such as
repeated criminal acts to obtain drugs or violent behaviour when intoxicated.
There are 10 personality disorders classified in DSMIV (American Psychiatric Association 1994) and nine in
ICD-10 (World Health Organization 1992); these show
considerable comorbidity and overlap and there are gains
in combining and simplifying them. This can be conducted in categories, combining paranoid and schizoid
personalities as an odd eccentric group (cluster A), the
emotionally unstable (borderline and impulsive), histrionic and dissocial (cluster B) and anxious, dependent and
anankastic personality disorders (cluster C). An alternative method of examining personality disorder is according to the severity of the disorder and this has been
investigated using a four-point hierarchy (Tyrer 1988;
Tyrer & Johnson 1996) as follows:
2004 Society for the Study of Addiction

1307

Level 0: no disorder;
Level 1: subthreshold personality disorder or personality difficulty;
Level 2: simple personality disorder (one or more personality diagnoses within one cluster); and
Level 3: diffuse or complex personality disorder
(personality disorders in more than one cluster).
It was decided to use these simplified methods in the
study of the prevalence of personality disorder in substance misuse services and by so doing we hoped to avoid
problems such as those found by Messina et al. (2001),
who found low levels of agreement between different
diagnostic measures of one personality disorder in substance misuse problems.
Study aims
In response to the conflicting prevalence evidence-base
and the difficulties of diagnosis and assessment we set out
to:
measure the prevalence of personality disorder in drug
and alcohol treatment populations;
compare the pattern of disorder in these two
populations;
examine the validity of the clinical diagnosis of personality disorder in these treatment populations; and
examine the association between personality disorder
(using both clusters and severity) and psychopathology, social morbidity and service characteristics.
METHOD
Design and setting
Data were collected between January 2001 and February
2002 in four urban UK centres, two in London (Hammersmith and Fulham; Brent) and two outside London (Sheffield; Nottingham). In total four drug services and three
alcohol services took part in the study. Participating drug
and alcohol teams offered separate structured, appointment-based services. All drug services had a strong
emphasis upon the management of opiate dependency.
Independent drug services were available in some areas
(including services for stimulant users), but not in others.
These latter agencies were not investigated. All patients
were under the clinical management of a consultant psychiatrist acting as responsible medical officer (RMO).
Patients were also allocated to a single keyworker from
the clinical teams, which comprised predominantly psychiatric nurses but also social workers and psychologists.
In most cases the keyworker was an experienced psychiatric nurse with responsibility for the daily management
of the patient.
The study formed part of a multi-centre, crosssectional survey designed to measure the prevalence of
Addiction, 99, 13061314

1308

Owen Bowden-Jones et al.

comorbid psychiatric disorder in substance misuse treatment populations. The methodology has been described
elsewhere (Weaver et al. 2003). In brief, we completed the
following two-stage investigation:
(i) We undertook a case-load census to identify the sampling frame. Demographic, diagnostic and service utilisation data were collected directly from keyworkers about
all patients in treatment who met the inclusion criteria
(see below).
(ii) A patient interview survey with case-note audit was
performed in a random subsample of the case-load census sampling frame. Patient interviews yielded data
about demographics and psychopathology (including
personality disorder) using standardized rating instruments (see below). The primary purpose of the case-note
audit was to obtain a valid assessment of psychotic
disorder.
Much of the prevalence evidence relates to treatment
populations in North America and our study represents
the largest survey to date of personality disorder in UK
substance-misuse treatment populations.

Subjects
Our intention was to measure the prevalence of personality disorder in the patient populations that drug and
alcohol services encounter in routine practice and to
ensure our findings were generalizable to the wider population of drug and alcohol service users. Hence, all current patients were included in the case-load census
population if they met the following inclusion criteria: (a)
the patient must have completed the assessment process
of their respective services; (b) been allocated to the caseloads of the RMO; (c) been assigned a personal keyworker
within the clinical team; and (d) lived in the catchment
area of the service. Patients temporarily in hospital,
prison, hostel or other temporary accommodation out of
the area but who retained residence within the catchment area were included.

Data collection
Data collection procedures were agreed with Local
Research Ethics Committees in each centre. On a predetermined census date each substance misuse service
identified a case list of current patients. Patients fulfilling
the inclusion criteria described above were allocated a
unique code number used for all data collection to ensure
anonymity. Keyworkers were asked to complete census
questionnaires (one per patient) with respect to this population. On the census date a random subsample of one in
five cases was selected at the co-ordinating centre (Imperial College, London) using SPSS random case selection
procedures (SPSS 1999).
2004 Society for the Study of Addiction

By selecting the interview sample on the case-load

census date, we ensured this sampling was blind to the
keyworker assessments and the census response rates.
Keyworkers were asked to give all patients sampled for
interview an information sheet and to invite them to
meet a trained fieldworker. If the patients consented to
meet a fieldworker, further information was provided and
written consent for participation in the study was
obtained formally. The interviews were conducted by
trained fieldworkers with backgrounds in psychology,
psychiatric nursing and social research. Non-consenting
patients were regarded as non-respondents and were not
substituted.

Assessments
Recorded clinical diagnosis and clinical management
As part of the case-load census, keyworkers were asked
to report ICD-10 psychiatric diagnosis established or
confirmed by psychiatric assessment in the past year.
Keyworkers were asked specifically whether their
patients had a recorded clinical diagnosis of a personality disorder in the case-notes and if so of which type;
they were then asked to rate their patients on five scales,
including difficulty of clinical management, clinical
stability, aggressive behaviour, level of engagement
and adherence to treatment plan. We also obtained
demographic details and information about psychiatric
and substance misuse interventions provided during the
past month.

Assessment of personality disorder

Personality status was assessed using the Quick Personality Assessment Schedule (PAS-Q) (Tyrer 2000). The
PAS-Q was developed from the longer Personality
Assessment Schedule (PAS) (Tyrer et al. 1979; Tyrer
2000), which provides a severity rating for each ICD-10
personality subtype and cluster using a standard procedure (Tyrer & Johnson 1996). The shortened PAS-Q
takes 1015 minutes to complete and shows good
agreement with the rating using the longer PAS instrument. Both the PAS and the PAS-Q used in this study
are particularly useful in the context of comorbid substance misuse disorders as they are designed to measure premorbid personality. All interviewers received a
common 2-day training course from an experienced
assessor (SC) in the assessment of personality disorder
and specific use of these instruments. This concentrated on the essential training programme for the full
PAS, as this emphasizes the cut-off points for different
levels of personality disturbance that are also used in
the PAS-Q.
Addiction, 99, 13061314

Prevalence of personality disorder

Assessment of psychiatric disorder

Research psychiatrists in each study centre assessed the
presence of psychosis using the Operational criteria
(OPCRIT) diagnostic checklist (McGuffin et al. 1991). The
OPCRIT employs structured recording of information
about the severity and duration of psychiatric symptoms
based on case-note review and information from patient
interviews. These data are then used to determine
whether psychosis is present and to generate an ICD-10
diagnosis. Each research psychiatrist received common
training in using this instrument. We also used the Comprehensive Psychopathological Rating Scale (CPRS)
(sberg et al. 1978) to assess symptom severity, and its
associated subscales, the Montgomerysberg Depression Rating Scale (MADRS) (Montgomery & sberg
1979) and Brief Scale for Anxiety (BAS) (Tyrer et al.
1984) to assess severity of affective and anxiety disorder,
respectively. The case-note review also obtained information on service use, violent behaviour, offending, personal safety and episodes of deliberate self-harm. At
interview, detailed information was collected about
demographics (ethnicity, living situation, education, lifetime street homelessness) and psychiatric history.
Analysis
All analysis presented in this paper was undertaken with
the interview samples achieved in each treatment population. Simple estimates of prevalence for personality disorder were calculated from the PAS-Q with regard to (a)
severity measured on the above four-point scale, (b) by
cluster type and (c) simple categorical diagnosis (disorder
or no disorder). The response rates and the prevalence
rates of personality disorder were calculated for drug and
alcohol treatment populations separately. All prevalence
estimates are reported with exact binomial 95% confidence intervals (CI).
The relationship between personality disorder and
psychopathology was examined using aggregated data
from the drug and alcohol treatment populations. We
undertook two separate analyses using (a) the four-point
severity scale and (b) a classification of cases based on
personality disorder cluster type. This former analysis
was undertaken with the full study population (n = 278)
However, the intention of the analysis by cluster type was
to assess whether there were differences in psychopathology between subjects with different types of PD. Hence,
we excluded cases from this latter analysis if they did not
have a personality disorder. The number of cases with
Cluster A personality disorders were small (drug service,
eight cases, 3.7%; alcohol service, four cases, 6.5%) and
were therefore also excluded from the analysis. This left a
total of 108 cases with which we undertook this analysis
(78 drug cases and 30 alcohol cases).
2004 Society for the Study of Addiction

1309

The statistical significance of the observed differences

the above analyses were assessed using Pearsons c2.
Data provided by keyworkers assessing the degree of
clinical difficulty presented by clients were available in
270 cases across drug and alcohol populations, and were
also aggregated to increase the statistical power. The
combination of these data was justified on the basis that
both measures (keyworker assessment and PAS-Q assessment of personality disorder) were comparable across
both samples. This combined data set was also used to
compare keyworker recorded personality disorder with
PAS-Q defined personality disorder. The MannWhitney
test was used for all non-parametric data. All the above
analyses were undertaken using SPSS (SPSS 1999).

RESULTS
Number and characteristics of subjects
Keyworkers provided data about 1645 of 1674 patients
(98.3%) meeting census eligibility criteria. Response
rates were similarly high for drug services (98.0%) and
alcohol service (99.2%). Complete interview and casenote data were obtained in 278 of 353 cases (78.8%)
selected randomly from the case-load census population
(216 drug patients, 62 alcohol patients). For seven cases
(five drug, two alcohol) interviews were completed but
case-notes could not be obtained. Given the absence of a
full data set, these cases have been excluded from the
analysis.
The achieved interview samples for both drug and
alcohol populations were predominantly male (67.1%
and 62.9%, respectively) and overwhelmingly of white
UK/Irish or European ethnic origin (92.1% and 95.2%,
respectively). A large majority of drug patients of both
sexes were 40 years old or under (males: 73%; females:
81.2%), as were most female alcohol patients (79.6%).
However, a lower proportion of male alcohol patients
were aged 40 or less (46.2%).
The demographic profile of the achieved interview
sample matches closely that of the case-load census population from which it was drawn, although there was a
slight non-significant under-representation of younger
women within the alcohol patient sample.

Prevalence of personality disorder in drug and

alcohol populations
Drug service population
Table 1 shows that among drug patients 37% (95% CI:
30.643.9) rated positive for at least one personality disorder. Among those with a disorder, cluster B disorders
were most common (30.1%; 95% CI: 24.136.7)
Addiction, 99, 13061314

1310

Owen Bowden-Jones et al.

Table 1 Prevalence of personality disorder subtypes among drug and alcohol treatment population samples.
Drug treatment population
(n = 216)

Alcohol treatment population

(n = 62)

95% CI

95% CI

136
80

(63.0)
(37.0)

30.643.9

29
33

(46.8)
(53.2)

40.166.0

8
6
2

(3.7)
(2.7)
(0.9)

1.67.2
1.05.9
0.13.3

4
3
2

(6.5)
(4.8)
(3.2)

1.715.7
1.013.5
0.411.2

Cluster B personality disorders

Dissocial (inc antisocial, psychopathic and sociopathic)
Emotionally unstableimpulsive type
Emotionally unstableborderline type
Histrionic

65
22
36
17
8

(30.1)
(10.2)
(15.8)
(7.7)
(3.6)

24.136.7
6.515.0
12.022.3
4.712.3
1.67.2

15
7
2
6
2

(24.2)
(11.3)
(3.2)
(9.7)
(3.2)

14.236.7
4.721.9
0.411.2
3.619.9
0.411.2

Cluster C personality disorders

Anankastic (incl. obsessive and compulsive)
Anxious/avoidant
Dependent

28
2
11
18

(13.0)
(0.9)
(5.0)
(8.1)

8.818.2
0.13.3
2.68.9
5.012.9

22
2
17
10

(35.5)
(3.2)
(27.4)
(16.1)

23.748.7
0.411.2
16.940.2
8.027.7

No disorder
Disorder present
Cluster A personality disorders
Paranoid
Schizoid

Values for the three clusters do not equal the total number of cases with diagnosed personality disorder due to some patients having more than one recorded
personality disorder.

followed by cluster C (13%; 95% CI: 8.818.2). Seventeen

patients (7.8%) had more than one recorded personality
disorder (complex/diffuse PD), the most frequent combination being clusters B and C (n = 11, 5.0%).

reported homelessness (P = 0.002) and suicide attempts

(P < 0.001) were also highest in this group (Table 2).

Alcohol service population

Significant differences were observed in global psychopathology (CPRS) scores between cases with cluster B disorders (mean 22.9, SD 12.5), cluster C disorders (mean
28.6, SD 15.2) and multiple personality disorders (mean
24.5, SD 15.9) (KruskalWallis c2 2 df = 9.7, P = 0.008).
The proportion of cases with severe anxiety disorder in
cluster C (14/28, 50%) and among cases with multiple
personality disorders (12/24, 50%) were also double
that observed among those with cluster B disorders
(14/56, 25%; c2 2 df = 7.2, P = 0.027). In addition, the
proportion of cases with multiple personality disorders
that had recorded previous psychiatric admission
(16/24, 66.7%) was higher than that observed among
cases with cluster C disorders (14/28, 50%) and cluster
B disorders (14/56, 25%; c2 2 df = 13.4, P = 0.001).
High rates of psychopathology and social morbidity were
observed among all cluster types. However, there were
no significant differences between cluster types in the
proportions of cases with a diagnosis of psychotic disorder, affective disorder or in the life-time recorded suicide
attempts and homelessness.

Table 1 also shows that for the alcohol population 53.2%

(95% CI: 40.166.0) rated positive for at least one PD
type. Among those with disorders, cluster C disorders
were most common (35.5%; 95% CI: 23.748.7) followed closely by cluster B (24.2%; 95% CI: 14.236.7).
The proportion of cases diagnosed with more than one
PD (n = 7, 11.3%) was slightly higher than found in the
drug population (n = 17, 7.8%) but the difference was not
statistically significant. All of these cases had a combination of cluster B and C personality types.

Association between severity of PD and psychopathology

Of the total study population of 278 patients, 34 (12.2%)
had no personality disorder (level 0), 131 (47.1%) had
personality difficulty (level 1) and simple (level 2) and
complex (level 3) personality disorders were identified in
89 (32%) and 24 (8.6%) patients, respectively (Table 2).
There were statistically significant differences between
the categories for measures of psychosis (P < 0.001),
affective disorder (P < 0.001) and anxiety disorder
(P < 0.001) with the highest prevalence of psychopathology observed within the level 3 group. The proportion of
patients ever being admitted to hospital (P < 0.001),
2004 Society for the Study of Addiction

Association between personality disorder cluster

and psychopathology

Keyworker assessment of clinical management

Keyworkers found patients with PAS-Q-defined personality disorder globally more difficult to manage, more
Addiction, 99, 13061314

Prevalence of personality disorder

1311

Table 2 Drug and alcohol treatment population: severity of PD related to social and psychiatric morbidity.

Psychiatric disorder

No disorder
(level 0)
n = 34 (%)

Personality
difficulty
(level 1)
n = 131 (%)

Single PD
(level 2)
n = 89 (%)

Multiple
clusters
(level 3)
n = 24 (%)

c2, P

All cases
n = 278 (%)

Psychotic disorder (OPCRIT)

No disorder
Psychosis present

33 (97.1)
1 (2.9)

125 (95.4)
6 (4.6)

74 (83.1)
15 (16.9)

16 (66.7)
8 (33.3)

c2 3 df = 23.5, P < 0.001

248 (89.2)
30 (10.8)

Affective(depressive) disorder (MADRS)

No disorder
Minor depressive disorder
Major depressive disorder

17 (50.0)
14 (41.2)
3 (8.8)

50 (38.2)
56 (42.7)
25 (19.1)

15 (16.9)
32 (36.0)
42 (47.2)

1 (4.2)
6 (25.0)
17 (70.8)

c2 6 df = 50.8, P < 0.001

83 (29.9)
108 (38.8)
87 (31.3)

Anxiety disorder (BAS)

No disorder
Minor anxiety disorder
Severe anxiety disorder

29 (85.3)
4 (11.8)
1 (2.9)

69 (52.7)
46 (35.1)
16 (12.2)

30 (33.7)
27 (30.3)
32 (36.0)

3 (12.5)
9 (37.5)
12 (50.0)

c2 6 df = 53.8, P < 0.001

131 (47.1)
86 (30.9)
61 (21.9)

Psychiatric hospital admission (ever)

Homeless (ever)
Suicide attempt (ever)

3 (8.8)
14 (41.2)
5 (14.7)

20 (15.3)
54 (41.2)
19 (14.5)

31 (34.8)
53 (59.6)
38 (42.7)

16 (66.7)
18 (75.0)
13 (54.2)

c2 3 df = 38.0, P < 0.001

c2 3 df = 14.3, P = 0.002
c2 3 df = 33.1, P < 0.001

70 (25.2)
139 (50.0)
75 (27.0)

Table 3 Comparison of keyworker perceptions of the degree of difficulty of clinical management between PAS-Q-defined PD and non-PD
patients.

Keyworker assessments of clinical management

N of valid cases:
Global difficulty of management
(low score = greater difficulty)
Stablechaotic
(high score = more chaotic)
Aggressivedocile
(low score = more aggressive)
Goodpoor engagement
(high score = poor engagement)
Adherence to care plan
(high score = less compliant)

No disorder
(n = 161)
Median (range)

PD present
(n = 109)
Median (range)

MannWhitney
U

U, P
P-value

All cases
(n = 270)
Median (range)

8 (110)

5 (110)

6741.0

< 0.01

7 (110)

4 (110)

6 (110)

6278.0

< 0.01

5 (110)

8 (210)

7 (110)

6724.0

< 0.01

8 (110)

4 (110)

5 (110)

6935.0

< 0.01

4 (110)

4 (110)

5 (110)

7022.0

< 0.01

4 (110)

chaotic, more aggressive and less engaged in treatment

with poorer compliance with care plans (P < 0.01 for all
comparisons) (Table 3).
Validity of service defined personality disorder diagnosis
Despite keyworkers identifying patients with PAS-Qdefined personality disorder as clinically more difficult to
manage, these patients were poorly diagnosed by services. The sensitivity of clinical diagnosis when compared
to PAS-Q was only 20.4% and the negative predictive
value 64.7% (Table 4).

DISCUSSION
These findings represent the most comprehensive
2004 Society for the Study of Addiction

multi-centre, UK assessment of personality pathology

among clients attending substance misuse services to
date and have important implications for service development. The overall prevalence of personality disorder we
observed was 37% in the drug misuse and 53% in the
alcohol misuse service populations. These prevalence
estimates are somewhat lower than observed in other
comparable studies (DeJong et al. 1993; Tyrer et al. 1997)
but our study assessed all in contact with drug and alcohol services, not just high-risk groups such as in-patients.
The estimates are still considerably higher than the prevalence of personality disorder in the general population
(around 10%) (de Girolamo & Reich 1993) and those
attending general practice (2934%) (Casey et al. 1984;
Moran et al. 2000). Another large-scale study using the
PAS-Q showed a prevalence of 42% in a population of 480
Addiction, 99, 13061314

1312

Owen Bowden-Jones et al.

Table 4 Drug and alcohol service patients: matched case comparison of key worker reported personality disorder (case-load census) and
PAS-Q assessment (interview survey).
PAS-Q (reference measure)

Keyworker reported diagnosis

Personality disorder
indicated (+)

No personality disorder
indicated ()

Total

Personality disorder reported (+)

No personality disorder reported ()
Total

23 (8.3%)
90 (32.3%)
113 (40.6%)

0
165 (59.4%)
165 (59.4%)

23 (8.3%)
255 (91.7%)
278 (100%)

Observed proportion of cases in which assessments differed:

32.3%.
Validity of keyworker assessments:
Sensitivity (% of reference measure + cases correctly reported)
20.4%.
Specificity (% of reference measure cases correctly reported)
100%.
Positive predictive value: (probability of reported + case being correct)
100%.
Negative predictive value: (probability of reported case being correct)
64.7%.
NB: In the PAS-Q a score of 2 or 3 on any personality diagnosis indicates personality disorder.

attending accident and emergency departments with

recurrent self-harm (Tyrer et al. 2003), a population
showing overlapping characteristics with substance misusers. The PAS-Q, although a brief measure, is conservative in its diagnosis of personality disorder and the level of
personality difficulty recorded in the PAS systems is likely
to be equivalent to the ICD-10 threshold of personality
disorder (Tyrer et al. 1994). The attempt of all PAS instruments to assess pre-misuse personality characteristics
and not just current personality pathology probably
reduced the prevalence figures to more realistic levels.
Nevertheless, a short instrument such as the PAS-Q is
really only a screening device and too much should not
be expected of any instrument that takes only a short
time to complete.
The prevalence of personality disorder recorded by the
keyworkers (around 8%) showed very low sensitivity
(20.4%) but high specificity (100%) when compared to
PAS-Q diagnoses. The low sensitivity may be explained by
reluctance of service professionals to label patients with a
pejorative diagnosis, by the consequence of unfamiliarity
with the diagnostic criteria for personality disorder, or by
raising of the threshold for diagnosis in a service dealing
with high levels of substance-related chaotic behaviour.
Small numbers in the cluster A group complicated
interpretation of personality disorder by cluster and no
consistent cluster specific associations were found. In
both populations clusters B and C were associated with
increased rates of psychotic, affective and anxiety disorders compared with substance-misusing patients without personality disorder. The study supports the use of a
dimensional classification of personality disorder as measured by PAS-Q showing a clear association between personality disorder severity and psychopathology, both
psychotic and affective. Social morbidity and service use
(as measured by admission to psychiatric hospital) were
also significantly more common with increasing disorder.
2004 Society for the Study of Addiction

The concept of a patients prognosis worsening with

increasing severity of personality disorder has face validity, although only a few other studies have had sufficient
numbers to demonstrate clearly such a relationship
(Gandhi et al. 2001; Moran et al. 2003). Hernandez-Avila
et al. (2000) suggested an association between the number of personality disorder diagnoses and the number of
crimes against property; however, other results have been
contradictory.
Certain study limitations should be acknowledged:
First, given that all centres were located in urban areas,
this may reduce the generalizability of our findings.
Patients in rural areas tend to receive more of their
mental health treatment for all conditions in primary
care (Seivewright et al. 1991).
Secondly, the study population consisted exclusively of
those who had voluntarily attended substance misuse
services. While this is representative of the substance
misuse treatment populations we intended to describe,
our findings will not be generalizable to the same
diagnostic or substance using groups who are not in
treatment.
Thirdly, some proportions lack precision due to the
small sample sizes and the 95% confidence intervals
around our proportions are sometimes wide.
The study results indicate that patients with personality disorder suffer higher rates of psychiatric and social
morbidity when compared to those without the diagnosis
and are perceived by their keyworkers as more difficult to
engage and less compliant with treatment. Despite this,
personality disorder is poorly identified by substance misuse services precluding early specialist intervention as
recommended by recent government guidance (NIMHE
2003). The PAS-Q could be a useful screening tool in
identifying and managing patients with personality disorder at an early stage of treatment, particularly if better
ways of treating this group can be identified. This study
Addiction, 99, 13061314

Prevalence of personality disorder

did not set out to examine the level of clinical understanding regarding personality disorder within substance misuse services. However, our findings suggest this is an area
of clinical practice in which further training is indicated.
As a multi-centre study, the COSMIC study represents
an advance on previous single centre studies, but further
investigation in more UK centres would be desirable to
strengthen the prevalence evidence-base further.

Acknowledgements
This work was undertaken by the authors who received
funding under the Department of Healths Drug Misuse
Research Initiative (grant no: 1217194). The views
expressed in this publication are those of the authors and
not necessarily those of the sponsors. The authors would
like to express their gratitude to the funders for supporting this work. CRDHB is core funded by the London
Regional Office of the NHS Executive.

References
American Psychiatric Association (APA) (1994) Diagnostic and
Statistical Manual of Mental Disorders, 4th edn. DSM-IV. Washington, DC: APA.
sberg, M., Montgomery, S., Perris, C., Schelling, D. & Sedvall,
G. (1978) A comprehensive psychopathological rating scale.
Acta Psychiatrica Scandinavica, 271, 527.
Brooner, R. K., Kidorf, M., King, V. L. & Stoller, K. (1998) Preliminary evidence of good treatment response in antisocial
drug abusers. Drug and Alcohol Dependence, 49, 249260.
Casey, P. R., Dillon, S. & Tyrer, P. J. (1984) The diagnostic status
of patients with conspicuous psychiatric morbidity in primary
care. Psychological Medicine, 14, 673681.
Clopton, J. R., Weddige, R. L., Contreras, S. A., Fliszar, G. M. &
Arredondo, R. (1993) Treatment outcome for substance misuse patients with personality disorder. International Journal of
Addiction, 28, 11471153.
Darke, S., Hall, W. & Swift, W. (1994) Prevalence, symptoms
and correlates of antisocial personality disorder among methadone maintenance clients. Drug and Alcohol Dependence, 34,
253257.
DeJong, C. A. J., van den Brink, W., Harteveld, F. M. & van der
Wielen, E. G. M. (1993) Personality disorders in alcoholics
and drug addicts. Comprehensive Psychiatry, 34, 8794.
Gandhi, N., Tyrer, P., Evans, K., McGee, A., Lamont, A. &
Harrison-Read, P. (2001) A randomised controlled trial of
community-oriented and hospital oriented care for discharged psychiatric patients: influence of personality disorder
on police contacts. Journal of Personality Disorders, 15, 94
102.
George, T. & Krystal, J. (2000) Comorbidity of psychiatric and
substance abuse disorders. Current Opinion in Psychiatry, 13,
327331.
de Girolamo, G. & Reich, J. H. (1993) Personality Disorders.
Geneva: World Health Organization.
Griggs, S. M. L. B. & Tyrer, P. J. (1981) Personality disorder,
social adjustment and treatment outcome in alcoholics. Journal of Studies on Alcohol, 42, 802805.
2004 Society for the Study of Addiction

1313

Hassiotis, A., Tyrer, P. & Cicchetti, D. (1997) Detection of personality disorders by community mental health teams: a
study of diagnostic accuracy. Irish Journal of Psychological
Medicine, 14, 8891.
Hernandez-Avila, C., Burleson, J., Poling, J., Tennen, H., Rounsaville, B. & Kranzler, H. (2000) Personality and substance
use disorders as predictors of criminality. Comprehensive Psychiatry, 41, 276283.
Links, P. S., Heslegrave, R. J., Mitton, J. E., van Reekum, R. &
Patrick, J. (1995) Borderline personality disorder and substance abuse: consequences of comorbidity. Canadian Journal
of Psychiatry, 40, 914.
McGuffin, P., Farmer, A. & Harvey, I. (1991) A polydiagnostic
application of operational criteria in studies of psychiatric illness. Archives of General Psychiatry, 48, 764770.
Messina, N., Wish, E., Hoffman, J. & Nemes, S. (2001) Diagnosing antisocial personality disorder among substance abusers:
the SCID versus the MCMI-II. American Journal of Drug and
Alcohol Abuse, 27, 699717.
Messina, N., Wish, E., Hoffman, J. & Nemes, S. (2002) Antisocial
personality disorder and therapeutic community treatment
outcomes. American Journal of Drug and Alcohol Abuse, 28,
197212.
Montgomery, S. A. & sberg, M. (1979) A new depression scale
designed to be sensitive to change. British Journal of Psychiatry, 134, 382389.
Moran, P., Jenkins, R., Tylee, A., Blizard, R. & Mann, A. (2000)
The prevalence of personality disorder among primary care
attenders. Acta Psychiatrica Scandinavica, 102, 5257.
Moran, P., Walsh, E., Tyrer, P. & Fahy, T. (2003) The impact of
co-morbid personality disorder on violence in psychosis
report from the UK700 trial. British Journal of Psychiatry, 182,
129134.
Nadeau, L., Landry, M. & Racine, S. (1999) Prevalence of personality disorder among clients in treatment for addiction.
Canadian Journal of Psychiatry, 44, 592596.
National Institute of Mental Health in England (2003) Personality Disorder: No Longer a Diagnosis of Exclusion. London:
Department of Health.
Reich, J. H. & Vasile, R. G. (1993) Effect of personality disorder
on the treatment outcome of Axis I conditions: an update.
Journal of Nervous and Mental Diseases, 181, 475483.
Rousar, E., Brooner, R. K., Reiger, M. W. & Bigelow, G. E. (1994)
Psychiatric distress in antisocial drug abusers: relation to
other personality disorders. Drug and Alcohol Dependence, 34,
149154.
Seivewright, N. & Daly, C. (1997) Personality disorder and drug
use: a review. Drug and Alcohol Review, 16, 235250.
Seivewright, H., Tyrer, P., Casey, P. & Seivewright, N. (1991) A
three-year follow-up of psychiatric morbidity in urban and
rural primary care. Psychological Medicine, 21, 495503.
SPSS, Inc. (1999) SPSS for Windows, Release 10.0.5. Chicago,
Illinois: SPSS, Inc.
Tyrer, P. (1988) Whats wrong with DSM-III personality disorders? Journal of Personality Disorders, 2, 281291.
Tyrer, P. (2000) Quick Personality Assessment Schedule: PASQ. In: Tyrer, P., ed. Personality Disorders: Diagnosis, Management and Course, 2nd edn, pp. 181190. London: Arnold.
Tyrer, P., Alexander, M., Cicchetti, D., Cohen, M. & Remington,
M. (1979) Reliability of a schedule for rating personality disorders. British Journal of Psychiatry, 135, 168174.
Tyrer, P., Gunderson, J., Lyons, M. & Tohen, M. (1997) Extent of
comorbidity between mental state and personality disorders.
Journal of Personality Disorders, 11, 242259.
Addiction, 99, 13061314

1314

Owen Bowden-Jones et al.

Tyrer, P. & Johnson, T. (1996) Establishing the severity of personality disorder. American Journal of Psychiatry, 153, 1593
1597.
Tyrer, P., Jones, V., Thompson, S., Catalan, J., Schmidt, U.,
Davidson, K., Knapp, M. & Ukoumunne. O. C. (2003) Service
variation in baseline variables and prediction of risk in a randomised controlled trial of psychological treatment in
repeated parasuicide: the POPMACT study. International Journal of Social Psychiatry, 49, 5869.
Tyrer, P., Merson, S., Onyett, S. & Johnson, T. (1994) The effect
of personality disorder on clinical outcome, social networks
and adjustment: a controlled clinical trial of psychiatric emergencies. Psychological Medicine, 24, 731740.
Tyrer, P., Owen, R. T. & Cicchetti, D. (1984) The Brief Scale for
Anxiety: a subdivision of the Comprehensive Psychopathological Rating Scale. Journal of Neurology, Neurosurgery and Psychiatry, 47, 970975.
Verheul, R. (2001) Co-morbidity of personality disorders in individuals with substance use disorders. European Psychiatry, 16,
274282.

2004 Society for the Study of Addiction

Verheul, R., van den Brink, W. & Hartgers, C. (1995) Prevalence

of personality disorder among alcoholics and drug addicts: an
overview. European Addiction Research, 1, 166177.
Weaver, T., Madden, P., Charles, V., Stimson, G., Renton,
A., Tyrer, P., Barnes, T., Bench, C., Middleton, H.,
Wright, N., Paterson, S., Shanahan, W., Seivewright, N. &
Ford, C. (2003) Co-morbidity of substance misuse and
mental illness collaborative (COSMIC) study: findings and
service development implications of a multi-centre prevalence study in NHS drug, alcohol and adult community
psychiatric services. British Journal of Psychiatry, 183,
304313.
Westen, D. (1997) Divergences between clinical and research
methods for assessing personality disorders: implications for
research and the evolution of Axis II. American Journal of Psychiatry, 154, 895903.
World Health Organization (1992) The ICD-10 Classification of Mental and Behavioural Disorders. Clinical descriptions and diagnostic guidelines. Geneva: World Health
Organization.

Addiction, 99, 13061314