27-10-2016, 02:19 PM
1461725301-yoa1504410031011.pdf (Size: 154.91 KB / Downloads: 4)
Context: Patterns of comorbidity among mental disorders
are thought to reflect the natural organization of mental
illness. Factor analysis can be used to investigate this
structure and construct a quantitative classification system.
Prior studies identified 3 dimensions of psychopathology:
internalizing, externalizing, and thought disorder.
However, research has largely relied on common
disorders and community samples. Consequently, it is
unclear how well the identified organization applies to
patients and how other major disorders fit into it.
Objective: To analyze comorbidity among a wide range
of Axis I disorders and personality disorders (PDs) in the
general outpatient population.
Design: Clinical cohort study.
Setting: A general outpatient practice, the Rhode Island
Methods to Improve Diagnostic Assessment and Services
(MIDAS) project.
Participants: Outpatients (N=2900) seeking psychiatric
treatment.
Main Outcome Measures: The Structured Clinical Interview
for DSM-IV and the Structured Interview for
DSM-IV Personality
Results: We tested several alternative groupings of the
25 target disorders. The DSM-IV organization fit the data
poorly. The best-fitting model consisted of 5 factors: internalizing
(anxiety and eating disorders, major depressive
episode, and cluster C, borderline, and paranoid PDs),
externalizing (substance use disorders and antisocial PD),
thought disorder (psychosis, mania, and cluster A PDs),
somatoform (somatoform disorders), and antagonism
(cluster B and paranoid PDs).
Conclusions: We confirmed the validity of the 3 previously
found spectra in an outpatient population. We
also found novel somatoform and antagonism dimensions,
which this investigation was able to detect
because, to our knowledge, this is the first study to
include a variety of somatoform and personality disorders.
The findings suggest that many PDs can be
placed in Axis I with related clinical disorders. They
also suggest that unipolar depression may be better
placed with anxiety disorders than with bipolar disorders.
The emerging quantitative nosology promises to
provide a more useful guide to clinicians and
researchers.
COMORBIDITY AMONG MENtal
disorders in clinical and
community populations is
extensively documented.1-7
It complicates research design
and clinical decision making but provides
an opportunity to improve psychiatric
classification.6,8,9 Patterns of comorbidity
are thought to reflect the underlying structure
of psychopathology, and analyses of
these patterns may reveal the natural classification
of mental illness.8-11
This proposal inspired a significant
number of studies that seek to construct
a new, quantitative nosology with the aid
of factor analysis, a procedure designed to
elucidate the structure of the data based
on relations among variables (eg, comor bidity). Indeed, there is a long tradition of
factor-analytically derived classification
systems, especially in child psychiatry.12-14
This research consistently identified
2 fundamental dimensions of mental
illness: the internalizing and externalizing
spectra. Recent factor analyses11,15,16 of
community surveys extended the quantitative
approach to adult populations. They
focused on 11 common mental disorders
and replicated the 2 fundamental dimensions.6
The internalizing spectrum included
depressive and anxiety disorders.
The externalizing spectrum was composed
of substance use disorders (SUDs),
conduct disorder, and adult antisocial behavior.
These dimensions have been found
in many cultures.17,18 Some studies6,11,15 also identified 2 subgroups within the internalizing spectrum:
a distress cluster (consisting of major depressive
disorder, dysthymic disorder, generalized anxiety disorder,
and posttraumatic stress disorder) and a fear cluster
(panic disorder, obsessive-compulsive disorder, and
phobic disorders). However, these clusters sometimes are
so highly correlated that they do not emerge as separate
elements within the internalizing spectrum.17,18
This research produced valuable insights into the natural
organization of mental illness, but it has been limited
in 2 respects. First, most studies of adults have been
restricted to community samples. Findings of general
population surveys do not necessarily generalize to clinical
samples. Indeed, it is unclear how well the identified
organization applies to psychiatric patients. Factoranalytic
studies have begun examining specific patient
populations, namely, self-identified patients, treatmentseeking
veterans, and inpatients with psychosis.11,19,20 The
present investigation sought to extend this work by evaluating
a general outpatient sample.
Second, the existing literature focused on common diagnoses,
namely SUDs, anxiety and depressive disorders,
and antisocial personality disorder (PD). It is uncertain
whether the previously identified spectra will be
confirmed when a broader range of diagnoses is considered
and whether additional dimensions are needed to
capture less-common disorders. Several investigations
have sought to extend the 2-spectrum model. One17 reported
that symptoms of somatization and hypochondriasis
belong to the internalizing cluster, although they
are less central to it than anxiety and depression. Another
study21 found that eating disorders are part of the
internalizing dimension. A third investigation20 observed
that schizophrenia and schizotypal PDs form a distinct
thought disorder spectrum. Finally, borderline PD
was linked to both internalizing and externalizing dimensions.22,23
These findings require replication but suggest
hypotheses for the present study.
Other factor-analytic investigations have examined comorbidity
among PDs. O’Connor24 cumulated data from
33 studies and found support for 2 structures. The first
model consisted of dimensions that can be identified as
externalizing (composed of cluster B and paranoid PDs)
and internalizing (cluster C, cluster A, and borderline).
The second model included the same externalizing factor
but split the cluster A PDs—disorders linked to the
thought disorder dimension—from the other internalizing
conditions. Thus, factor analyses of PDs appear to
replicate the spectra found in studies centered on Axis I
disorders.
However, only joint analyses of Axis I and Axis II disorders
can link the 2 sets of findings. Few such investigations
have been undertaken. Beyond antisocial PD, there
are some initial data on borderline and schizotypal diagnoses,
but virtually nothing is known about placement
of other PDs in the overall quantitative classification.
The most comprehensive study25 to date analyzed
various Axis I and Axis II symptoms in a British community
sample and found 4 broad dimensions: internalizing,
externalizing, thought disorder (symptoms of psychosis
and cluster A PDs), and pathological introversion
(symptoms of avoidant and dependent PDs). It appears that the first 3 dimensions cut across Axis I and Axis II
symptomatology, whereas pathological introversion is specific
to the latter axis. It is uncertain, however, whether
the same dimensions would be found in analyses of the
corresponding disorders.
The aim of the present investigation was to broaden
the quantitative nosology by examining a wide range of
Axis I and Axis II conditions, many of which have not
been considered in this framework. In particular, we
sought to integrate personality pathology fully into this
system and to explicate the nature of the relations between
the axes. Moreover, we planned to evaluate the generalizability
of the previously identified spectra to the outpatient
population using a large, unselected sample
diagnosed with state-of-the-art procedures. We hypothesized
that the current DSM-IV organization of disorders
would fit the data poorly. We further predicted that
the internalizing, externalizing, and thought disorder spectra
would be confirmed in this sample. We also planned
to test whether the same spectra cut across Axis I and
Axis II. Finally, we sought to examine the distinction between
fear and distress disorders observed within the internalizing
cluster in several studies.6,11,15,16 Given that the
present analyses go well beyond previous research, we
made modifications to our a priori models when such
changes were clearly indicated by the data.
SAMPLE AND PROCEDURE
Data were obtained from the Rhode Island Methods to Improve
Diagnostic Assessment and Services (MIDAS) project, a
clinical program created to integrate research assessments into
routine care.26 Participants presenting at a community-based
outpatient psychiatric practice underwent a comprehensive diagnostic
assessment. The practice predominantly treats individuals
with medical insurance (including Medicare) on a feefor-service
basis. The main referral sources are primary care
physicians, psychotherapists, and family members or friends.
All individuals seeking treatment at this practice were asked
to participate in the MIDAS project. Exclusion criteria were age
younger than 18 years, inability to understand English, and severe
cognitive impairment. Nonparticipants were compared with
participants using self-administered symptom inventories, and
no significant differences were found, suggesting that this sample
is representative of the population served by the clinic with regard
to psychopathology.27,28 The Rhode Island Hospital’s institutional
review board approved the research protocol, and
all participants provided written informed consent.
The sample included the 2900 consecutive patients evaluated
in the MIDAS project since it began. Their mean (SD) age
was 38.5 (13.0) years; the majority were female and white
(Table 1). Of these patients, 2151 completed the PD assessment.
This component was not introduced until the study was under
way and the procedures for incorporating research interviews
into clinical practice had been well established. As a result, 749
participants were missing PD data. There were no significant differences
between participants with and those without PD assessment
on any demographic characteristics or Axis I diagnoses except
that the latter were more likely to have a psychotic disorder
(12.3% vs 6.7%,P.001) andlesslikely to have generalized anxiety
disorder (18.8% vs 30.6%,P.001). Thus,missing datalikely
had little systematic effect on the results. We addressed missing data with the Full Information Maximum Likelihood method,29
which uses all availableinformation without deleting any records
and is recommended for such missing data patterns.
MEASURES
Lifetime Axis I diagnoses were made using the Structured Clinical
Interview for DSM-IV (SCID),30 which was modified to relax
certain hierarchical exclusion rules and thus allow some
nonhierarchical diagnoses. Lifetime rather than current diagnoses
were chosen for consistency with the PD assessment. Axis
II conditions were measured with the Structured Interview for
DSM-IV Personality (SIDP).31 Each DSM-IV PD criterion was
rated on a 0 (not present) to 3 (strongly present) scale, with a
score of 2 (present) or higher considered positive. The SIDP
questions are grouped thematically to reduce halo effects (ie,
ratings for a criterion are influenced by how other criteria of
that diagnosis are rated).
Both assessments were administered by highly trained interviewers
(including C.J.R.) who were monitored throughout the
study to minimize rater drift. Interviewers typically were PhDlevel
psychologists. Every diagnostician underwent intense training
lasting 3 to 4 months.26 The raters were required to demonstrate
exact agreement with a senior diagnostician on 5 consecutive
evaluations. Ongoing supervision by one of theinvestigators (M.Z.)
included weekly case conferences and review of written reports
and item ratings of every case. Fourteen raters performed joint
interviews to assess the diagnostic reliability of the SCID (based
on 65 participants) and SIDP (based on 47 participants). The SCID
reliability estimates () ranged from 0.64 to 1.00 (median, 0.88).
Reliability of any PD on the SIDP was 0.90. Individual disorders
were too rare to compute coefficients, but intraclass correlation
coefficientsfor criterion counts rangedfrom 0.82 to 0.97 (median,
0.94).
The SCID covers 7 DSM-IV sections: SUDs and mood, psychotic,
anxiety, somatoform, adjustment, and eating disorders.
In selecting variables for the analyses, we considered both
frequency and hierarchical exclusion rules. Disorders with low
frequency (defined as 20 cases) were excluded because their
associations with other variables cannot be estimated reliably.
Diagnoses affected by hierarchical rules could not be analyzed
because those rules prohibit certain combinations of diagnoses
and therefore would dictate the structure, leading to spurious
findings.
Consequently, we examined mood episodes (major depressive
and manic) rather than mood disorders, as these diagnoses
contain exclusion rules. We used a nonhierarchical generalized
anxiety disorder diagnosis. Psychosis—defined as the
presence of definite psychotic symptoms, including psychosis
during mood episodes—was analyzed as a single category and
could not be subdivided because individual psychotic disorders
incorporate complex hierarchical rules. For the same reason,
we examined a broad eating-disorder group that consisted
of anorexia nervosa, bulimia nervosa, and binge eating
disorder. In addition, the undifferentiated somatoform disorder
group included cases with somatization disorder, which represents
an extreme form of this condition. Body dysmorphic
disorder was too infrequent to be analyzed. Adjustment disorders
were not considered because all involved hierarchical rules
that could not be relaxed. Not otherwise specified diagnoses
were not counted in any of the categories. Overall, 15 Axis I
conditions were selected (Table 1).
The SIDP assesses all 10 PDs, but several diagnoses had low
frequency. To ensure comprehensive coverage of personality
pathology, we expanded PD categories to include subthreshold
cases. Specifically, we required 1 criterion less than DSM-IV
thresholds and thus were able to analyze all 10 resulting PD
traits. Similar to prior studies,6,20 we treated adult antisocial traits
and childhood conduct problems as separate variables instead
of combining them into antisocial PD, which allowed us to test
rather than assume this link. We also found that avoidant PD
was highly overlapping with social phobia (tetrachoric r=0.81).
This is consistent with reports arguing that avoidant PD is an
extreme form of social phobia.32-34 Given this problematic redundancy,
avoidant PD was excluded from the analysis