Health Mental

Location:

Australia

Posted:

November 15, 2012

Contact this candidate

Resume:

Soc Psychiatry Psychiatr Epidemiol (****) **:*** *30 DOI 10.1007/s00127-007-0182-3

O RIGINAL PAPER

John R. Beard Kathy Heathcote Robert Brooks Arul Earnest Brian Kelly

Predictors of mental disorders and their outcome in a

community based cohort

Accepted: 3 January 2007 / Published online: 22 June 2007

j Abstract Background Only a limited number of up were disorder-free two years later. For participants

population-based studies have been able to prospec- with a disorder at both interviews, there was marked

tively follow the mental health of their participants. lability in diagnoses, with only a small minority

We aimed to describe diagnostic changes in a popu- having an unchanged diagnosis at both baseline and

lation based cohort over a two year period, and to follow-up. Factors predicting a poor outcome in

explore associations between a range of individual participants with a disorder included the number of

factors and recovery from, or onset of, disorders. baseline diagnoses, high neuroticism score and ad-

Methods: Two year, face-to-face follow-up of a com- verse life events. Conclusions: These ndings suggest

munity-based cohort drawn from random telephone that the diagnosis of common mental disorders is

screening using the CIDI as diagnostic instrument. complex and that diagnoses are relatively unstable.

Unlike most similar research we did not exclude The factors that in uence the emergence of mental

individuals with prior history from analysis. Results: disorders in individuals who may, or may not, have

1407 participants were administered face-to-face had a disorder in the past, are similar to those asso-

interviews and 968 were re-interviewed. In multivar- ciated with the development of new disorders in

iate analysis, recent adverse life events, poor physical subjects without a lifetime history.

health, and high neuroticism score were signi cant

predictors of developing a mental disorder in partic- j Key words depression anxiety mental disor-

ipants who were disorder free at baseline. Higher ders cohort study

baseline levels of physical activity were protective of

new disorders in univariate analysis. Most partici-

pants with a baseline disorder and not lost to follow-

Introduction

J.R. Beard K. Heathcote A. Earnest

Over the past decade a number of large cross sectional

School of Public Health

University of Sydney surveys have given us a better understanding of the

Sydney (NSW), Australia

prevalence of mental disorders in the general popu-

E-Mail: ******@***.****.***.**

lation [13, 15, 18, 21]. However, only a few commu-

J.R. Beard K. Heathcote A. Earnest nity based studies have been able to follow individuals

Faculty of Health and Applied Sciences

over time to determine the factors in uencing their

Southern Cross University

mental health and help seeking behaviours [6, 8, 10,

Lismore (NSW), Australia

14, 16, 25, 28]. This population based research is

J.R. Beard

particularly important, since the majority of individ-

Center for Urban Epidemiologic Studies

uals with a disorder may never seek clinical help [2],

New York Academy of Medicine

New York (NY), USA and studies of subjects drawn from clinical settings

may therefore not be representative of the broader

R. Brooks

Centre for Population Mental Health community.

The Liverpool Hospital

A number of longitudinal studies have used

Liverpool (NSW), Australia

structured instruments, such as the Composite

B. Kelly International Diagnostic Interview (CIDI) or Diag-

SPPE 182

NSW Centre for Rural and Remote Mental Health

nostic Interview Schedule, in face-to-face interviews

University of Newcastle

to estimate the incidence, and explore the determi-

Orange (NSW), Australia

624

nants, of new disorders in previously symptom free birth and invited to participate in a telephone interview, with 9,191

interviews being completed (approximately 9.2% of the adult

individuals [8, 10 12, 24, 28]. These studies have

population), a response rate of 75.8%. Some over-sampling of

generally excluded subjects with a known prior his- males was conducted toward the end of the screening period to

tory when determining the factors that may increase ensure they were not under-represented. No demographic infor-

the risk of incident disorders. However, from a pop- mation was available for refusals. The demographic characteristics

of telephone screening respondents were broadly similar to the

ulation perspective, the greatest burden of mental

underlying population. 5,201 (56.6%) were female, compared to an

disorders on the health of the community lies with underlying proportion of women among adults in the study area of

individuals having a past history, who are at greatest 50.96%. Subjects under 35 years of age were under-represented in

risk of new disorders. It is therefore important to also the sample population (5.7% of study population) compared to

other subjects (10.01% of study population). However, for subjects

identify ameliorable factors that in uence the likeli-

between 34 and 85 years of age, the sample was evenly distributed

hood of developing a repeat disorder in this large across all age groups (range 9.3% to 11.2% of study population).

section of the population. All 1,670 subjects identi ed as likely cases during screening

We used the Northern Rivers Mental Health Study and a random sample of 963 subjects identi ed as likely controls

(NoRMHS) [3, 5] to follow a cohort of community were invited to participate in the prospective face-to-face stages of

the study. 1407 invited subjects completed baseline face-to-face

based subjects over a two year period in order to

interviews [859 (51.4%) likely cases and 548 (56.9%) likely con-

identify the factors that were predictive of changes in trols]. Approximately two years after this baseline interview, we

their mental health status regardless of their past attempted to recontact all 1407 cohort members. 968 subjects were

history. reinterviewed and reliably matched to baseline (31% lost to fol-

lowup).

j Face-to-face interviews

Methods

Face-to-face interviews were generally conducted at the subject s

NoRMHS was designed to prospectively explore the incidence and place of residence. Interviews followed the same design and used

changing patterns of mental disorders in subjects who were drawn the same instruments as the National Survey of Mental Health and

from the general population and who were living in their usual Wellbeing (NSMHWB) undertaken in Australia in 1997 by the

community setting. The methods have been described in detail Australian Bureau of Statistics in a random national sample of

elsewhere [5]. The study comprised three phases: a telephone 10,600 community based subjects [13]. The NSMHWB used the

screening to recruit a cohort at risk of mental disorders, a baseline Composite International Diagnostic Interview (CIDI) as the core

face-to-face interview of the cohort, and a repeat face-to-face diagnostic instrument and also included a range of other measures

interview two years later. To ensure that the study did not in uence of disability, general health, psychological distress, social-demo-

the behaviours and outcomes of subjects, both subjects and inter- graphic variables, perceived health needs, help-seeking behaviour,

viewers were blind to diagnoses made during the study interviews. neuroticism, and health service utilisation. The CIDI comprises a

computer-aided diagnostic interview designed to be administered

by lay interviewers that is highly correlated with clinical assessment

j Screening and cohort selection [29]. All interviewers were trained by experienced clinicians fol-

lowing training guidelines from the World Health Organisation. We

We administered a telephone interview to subjects chosen by added further questions at the end of the interview on the number

random digit dialling from the Richmond Valley of New South and nature of adverse life events experienced by subjects in the

Wales, Australia. The total interview took approximately 20 min- previous 12 months, height and weight. Social connectedness was

utes to complete and comprised a screening instrument for mental investigated by a three item instrument asking the amount and

disorders, the MiniCIDI, and some basic questions on medical nature (mail, telephone or personal) of contact participants had

history and health service utilisation. with family and friends, while physical activity measurement was

The MiniCIDI is derived from the Composite International based on a widely used and validated instrument [7].

Diagnostic Interview Short Form (CIDI-SF) scales, a series of

diagnosis-speci c scales that were developed from item-level

j Analysis

analyses of the CIDI questions in the National Comorbidity Survey

[17, 22, 26]. The scales were designed to reproduce the full CIDI

Baseline and follow-up data were entered into a single dataset and

diagnoses as exactly as possible with only a small subset of the

analysed using SPSS for Windows Version 12.0.1 (SPSS Inc, Chi-

original questions. Comparison of the CIDI-SF classi cations of

cago, Ill, USA) and Stata V9.0 (Stata College, TX, USA). Mental

generalized anxiety disorder with the full CIDI classi cations in the

health diagnoses were drawn from the CIDI using similar programs

National Comorbidity Survey yielded a sensitivity of 96.6%, a

to those of the NSMHWB to de ne diagnoses by applying standard

speci city of 99.8%, and 99.6% overall agreement, while classi -

International Classi cation of Disease Version 10 (ICD10) criteria.

cations of major depression yielded a sensitivity of 89.6%, a spec-

These were grouped into the larger diagnostic categories shown in

i city of 93.9%, and an overall agreement of 93.2% [19, 20]. The

Table 1.

interview includes questions on physical health, impairment and

Univariate logistic regression was used to explore the in uence

demographics and has 8 stem questions for common mental dis-

of a range of predictor variables on new onsets of these outcomes.

orders that lead to more detailed questioning if required.

These included gender, age, number of life events in the 12 months

We used random digit dialling for the screening phase of the

preceding followup, baseline neuroticism measure of the Eysenck

study, calling a total of 12,138 residential telephone numbers in the

Personality Questionnaire (EPQ) [26], baseline physical activity

Richmond Valley of New South Wales, Australia. This area is

(measured as minutes of mild moderate and viorous exercise),

demographically heterogeneous, comprising coastal towns, large

baseline social connectedness, baseline physical dimension of

regional centres and smaller rural villages and districts. There is a

the SF12, and baseline psychological distress measured by the

higher than usual proportion of people of Indigenous background

Kessler10 [17].

in the population, and a smaller proportion of non English

We included a variable, caseness, to account for any in uence

speakers. At the time, approximately 97% of Australian households

of the screening status of subjects. We also used univariate analysis

had a landline telephone connection [23]. Adult members of

to explore the in uence of predictor variables on the outcome of

households called were randomly selected according to date of

625

Table 1 Re-categorisation of individual diagnoses into broader diagnostic categories

General category Sub-category Diagnostic (ICD-10) codes

Substance use Disorders due to alcohol 10.1 10.2

Disorders Disorders due to cannabis 12.1 12.2

Disorders due to other drugs 11.1 11.2 13.1 13.2 15.1 15.2

Depression Depression excluding mania and dysthymia 32.0 32.1 32.2 32.00 32.01 32.10 32.11

33.00 33.01 33.10 33.11 33.2

Anxiety disorders Phobias, panic, Obsessive Compulsive 40.0 40.00 40.01 40.1

Disorder 41.0 41.00 41.01

42.0 42.1 42.2

Generalised Anxiety Disorder 41.1

Post Traumatic Stress Disorder 43.1

comorbidity, and 8 (23.5%) of 34 subjects with sub-

subjects with mental disorders at baseline. Since there was con-

siderable lability between diagnoses over time, we used whether or stance use disorders without comorbidity still had the

not the subject met the criteria for any mental disorder at follow-up

same diagnosis at follow-up.

as our outcome measure. We included several additional variables

We also examined the help seeking behaviour of

including the number of comorbid diagnoses at baseline and, for

subjects with mental disorders (Table 3). Slightly less

subjects with depression, the number of ICD 10 depression criteria

met at baseline. We also examined the in uence of whether or not a than half the subjects meeting criteria for any mental

subject had seen a health professional for their mental disorder in

disorder at either interview had sought the help of a

the 12 months prior to baseline.

relevant health professional in the preceding 12

For multivariate analysis we calculated odds ratios and their

months. Low levels of help seeking were more marked

corresponding 95% con dence intervals by starting from the most

signi cant variable identi ed in the univariate analysis, and using for anxiety and substance use disorders, with subjects

the likelihood ratio test to see if inclusion of a covariate helped

with affective disorders and comorbid diagnoses more

improve the overall t of the multivariate model. For continuous

likely to have seen a mental health professional.

variables, we tested for linearity by including a quadratic term in

the model and, where the relationship between the covariate and

the outcome was not found to be linear, we categorised variables

using standard quartiles. j Predictors of new disorders in subjects without a

baseline disorder

Results The results of univariate analyses for the development

of any mental disorder at follow-up in subjects who

were disorder free at baseline, and for the develop-

j Patterns of disorders over time

ment of either anxiety or depressive disorders, spe-

ci cally, are shown in Table 4. The numbers of

A cross tabulation of subject diagnoses at baseline

subjects developing a new diagnosis of substance use

and follow-up is shown in Table 2. At baseline, 493

disorder were too small for satisfactory analysis. We

(35%) subjects were identi ed as having a mental

also used logistic regression to explore the relation-

disorder, with 318 having an anxiety disorder, 228

ship of Body Mass Index on disorder incidence and

having a depressive disorder and 128 having a sub-

found no signi cant association.

stance use disorder. 161 (33%) of the subjects with a

The results of multivariate logistic regression are

disorder had comorbidity with one or more further

shown in Table 5. The total number of adverse life

disorders.

events occurring in the 12 months prior to follow-up,

The prognosis over the study period for subjects

increasing EPQ neuroticism score and poor score on

with a disorder was good, with 167 (51%) of the 330

the physical components of the SF12 were predictive

subjects with a baseline diagnosis who were not lost to

of developing a disorder in our nal model.

follow-up being disorder free at follow-up. The

When speci c diagnostic outcomes were explored,

prognosis was best for subjects with a single disorder

smaller numbers of outcomes resulted in the study

without comorbid mental disorders. However, the

having less power. However, adverse life events were a

prognosis for subjects with baseline comorbidity was

signi cant predictor for the development of all out-

poorer with only 32 (31%) being diagnosis free at

comes. For subjects developing depression, EPQ

follow-up.

neuroticism score and physical components of the

For those subjects with disorders identi ed at both

SF12 remained signi cant predictors. For anxiety

baseline and follow-up, there was considerable lability

disorders as a group, gender replaced physical health

in the nature of their diagnosis. Only 5 (6.8%) of 73

in the model. For generalised anxiety disorder, only

subjects with a diagnosis of depression without a

neuroticism was a signi cant additional predictor,

comorbid mental disorder at baseline met the criteria

while for speci c anxiety disorders such as phobias

for depression without comorbidity at follow-up. 33

and obsessive compulsive disorder screening caseness

(28.3%) of 121 subjects with anxiety disorders without

626

was signi cant. No variable predicted the onset of

Total

1407

914

173

Post Traumatic Stress Disorder.

(37.5%)

(30%)

(26%)

(43%)

(32%)

(44%)

(45%)

(31%)

j Predictors of outcome in subjects with baseline

follow-up

lost to

disorders

276

439

Univariate analyses for outcome at follow-up in sub-

jects with any baseline disorder, anxiety disorder and

anxiety and SUD

affective disorder are shown in Table 6. The numbers

depression,

for subjects with substance use disorders were too

(10%)

(2%)

(6%)

small for satisfactory analysis. For each baseline

diagnosis we then built a multivariate model (Ta-

ble 7). For all disorders, the signi cant variables in

(12.5%)

the multivariate model included the total number of

(10%)

and SUD

(1%)

(3%)

(9%)

(1%)

anxiety

diagnoses at baseline. When this measure was intro-

duced into the model, most other measures of severity

became non signi cant. For any disorder at baseline,

the model also included the total number of life events

depression

SUD and

between interviews and high neuroticism score. High

1 (1%)

neuroticism score also predicted adverse outcome in

subjects with affective disorders and adverse life

events were signi cantly associated with a worse

anxiety and

depression

(16%)

(10%)

prognosis in subjects with anxiety disorders grouped

(1%)

(5%)

(4%)

(6%)

(3%)

together. Living with a partner at baseline gave

subjects with generalised anxiety disorder a better

prognosis. Living with a partner generally had a

pure SUD

(13%)

mild positive effect in all univariate analysis, but

(1%)

(3%)

(1%)

(9%)

(5%)

(2%)

only approached signi cance for generalised anxiety

disorder.

Diagnosis at follow-up (% of baseline diagnosis)

depression

(10%)

(2%)

(4%)

(5%)

(6%)

(3%)

pure

Discussion

While our study contains some methodological

pure anxiety

(19%)

(14%)

(23%)

(16%)

weaknesses (see below), our ndings reinforce the

(6%)

(2%)

(5%)

(9%)

disorder

complexity and interconnectedness of common

126

mental disorders. The majority of subjects with a

disorder at baseline and not lost to follow-up were

No diagnosis

disorder free at follow-up, and the outlook was even

(37.5%)

(12.5%)

(59%)

(40%)

(43%)

(37%)

(23%)

(50%)

(5%)

better for those with a speci c baseline diagnosis

without comorbid mental disorders. Most of the

540

707

participants with a diagnosis at both baseline and

follow-up had changed the nature of their diagnosis

Anxiety, depression and SUD

during this two year period.

Table 2 ICD 10 diagnoses at baseline and follow-up

While there is strong evidence that clinical inter-

Anxiety and depression

pure anxiety disorder

ventions are of bene t for individuals with mental

SUD and affective

Contact this candidate