Non-uniform effectiveness of structured patient-clinician communication in community mental healthcare: an international comparison.

Abstract

The effectiveness of psychosocial interventions in community mental healthcare has been shown to depend on the setting in which they are implemented. Recently structured patient-clinician communication was found to be effective in a multi-centre trial in six European countries, the DIALOG trial. In the overall study, differences between centres were controlled for, not studied. Here, we test whether the effectiveness of structured patient-clinician communication varies between services in different countries, and explore setting characteristics associated with outcome.
The study is part of the DIALOG trial, which included 507 patients with schizophrenia or related disorder, treated by 134 keyworkers. The keyworkers were allocated to intervention or treatment as usual.
Positive effects were found on quality of life (effect size 0.20: 95% CI 0.01-0.39) and treatment satisfaction (0.27: 0.06-0.47) in all centres, but reductions in unmet needs for care were only seen in two centres (-0.83 and -0.60), and in positive, negative and general symptoms in one (-0.87, -0.78, -0.87). The intervention was most effective in settings with patient populations with many unmet needs for care and high symptom levels.
Psychosocial interventions in community mental healthcare may not be assumed to have uniform effectiveness across settings. Differences in patient population served and mental healthcare provided, should be studied for their influence on the effectiveness of the intervention. Structured patient-clinician communication has a uniform effect on quality of life and treatment satisfaction, but on unmet needs for care and symptom levels its effect differs between mental healthcare settings.

Full text

Page 1

Non-uniform effectiveness
of structured patient-clinician communication
in community mental healthcare:
An international comparison

Rob van den Brink · Durk Wiersma · Kerstin Wolters · Jens Bullenkamp · Lars
Hansson · Christoph Lauber · Rafael Martinez-Leal · Rosemarie McCabe · Wulf
Rössler · Hans Salize · Bengt Svensson · Francisco Torres-Gonzales · Stefan Priebe


Dr. R. van den Brink (corresponding author) · D Wiersma · K. Wolters
Dept. of Psychiatry
University Medical Center Groningen
University of Groningen
P.O. Box 30.001, CC73
9700 RB Groningen, The Netherlands
E-mail: r.h.s.van.den.brink@med.umcg.nl
(tel: +31 50 3612089; fax: +31 50 3619722)

J. Bullenkamp · H. Salize
Central Institute for Mental Health
Mannheim, Germany

Page 2

2
L. Hansson · B. Svensson
Dept. of Health Sciences
University of Lund
Lund, Sweden

C. Lauber · W. Rössler
Dept. for Social and Clinical Psychiatry
Psychiatric University Hospital
Zürich, Switzerland

R. Martinez-Leal · F. Torres-Gonzales
Dept. of Psychiatry
University of Granada
Granada, Spain

R. McCabe · S. Priebe
Unit for Social and Community Psychiatry
Queen Mary University of London
London, UK

Page 3

3
Abstract
Background The effectiveness of psychosocial interventions in community mental
healthcare has been shown to depend on the setting in which they are implemented.
Recently structured patient-clinician communication was found to be effective in a
multi-centre trial in six European countries, the DIALOG trial. In the overall study
differences between centres were controlled for, not studied. Here we test whether the
effectiveness of structured patient-clinician communication varies between services in
different countries, and explore setting characteristics associated with outcome.
Methods The study is part of the DIALOG trial, which included 507 patients with
schizophrenia or related disorder, treated by 134 keyworkers. The keyworkers were
allocated to intervention or treatment as usual.
Results Positive effects were found on quality of life (Effect Size 0.20: 95%CI 0.01–
0.39) and treatment satisfaction (0.27: 0.06–0.47) in all centres, but reductions in
unmet needs for care were only seen in two centres (-0.83 and -0.60), and in positive,
negative and general symptoms in one (-0.87; -0.78; -0.87). The intervention was
most effective in settings with patient populations with many unmet needs for care
and high symptom levels.
Conclusions Psychosocial interventions in community mental healthcare may not be
assumed to have uniform effectiveness across settings. Differences in patient
population served and mental healthcare provided, should be studied for their
influence on the effectiveness of the intervention. Structured patient-clinician
communication has a uniform effect on quality of life and treatment satisfaction, but
on unmet needs for care and symptom levels its effect differs between mental
healthcare settings.

Page 4

4
Key words
physician-patient relations – computer-assisted decision making – community mental
health services – population characteristics – health services research

Page 5

5
Introduction

Established psychosocial interventions for people with severe mental illness, such as
assertive community treatment [1, 2] and supported employment [3, 4], have been
found to have non-uniform effectiveness across mental healthcare settings. This may
be because settings differ in crucial aspects, such as patient selection or the ‘treatment
as usual’ offered in the comparison condition of a trial, i.e. characteristics that are
often left unspecified in research reports [5]. Alternatively, interventions may be
perceived and appreciated differently in different cultures [6]. Burns and Catty [4]
conclude that such differences in effectiveness of an intervention between settings
may be as informative – e.g. with respect to understanding the mechanisms behind the
effects – as their similarities, and that they should be studied rather than being treated
as a complexity to be overcome.
Recently the effectiveness of a novel intervention – structured patient-clinician
communication – was studied across six European community mental healthcare
settings, in the DIALOG trial [7]. This showed that two-monthly, computer-assisted
discussions between the patient and clinician of the patient’s satisfaction with
different domains of life, current treatment, and needs for additional or different help,
had a positive influence on the patient’s quality of life, unmet needs for care and
treatment satisfaction. In the report by Priebe et al. [7] the six sites in which the trial
was conducted, were treated as a random sample of all possible community mental
healthcare settings that treat the target group of patients. Thus, study site was
considered a random variable, that was treated as a ‘nuisance’ variable to be
controlled for in the analysis, rather than being studied for its influence on the
effectiveness of the intervention.

Page 6

6
The DIALOG intervention was intended to foster a ‘partnership model of care’
between patients and their clinicians, and to encourage patients to take an active role
in care planning [8]. However, preference to participate in medical decision making is
not universal across cultures [9, 10], and has been found to depend on age, gender,
education, and social class [11, 12]. Therefore, universal effectiveness of the
DIALOG intervention may not be assumed, but should be tested.

■ Aims of the study
We aim to test whether the effectiveness of the DIALOG intervention varies across
mental healthcare settings in different countries, and to explore whether the
effectiveness is associated with characteristics of the patient population served and the
mental healthcare provided. The objective is to identify favourable and less
favourable settings for the implementation of structured patient-clinician
communication in community mental healthcare.


Method

■ Study design
The present study is part of the DIALOG trial, which consisted of a cluster
randomized controlled trial in community mental health services in six European
countries. Eligible were services that provided comprehensive, outpatient care for
people with schizophrenia or related psychotic disorders, and that operated a
keyworker system in which every patient has a designated clinician. Keyworkers were
randomly assigned to either the intervention of structured patient-clinician

Page 7

7
communication or treatment as usual. The effect of the intervention was evaluated
over a 1-year period, by pre and post interviews with the patients. Details of the
randomization procedure and the eligibility criteria for keyworkers and patients have
been provided by Priebe et al. [7]. The study was approved by relevant ethics
committees in the six countries, and written informed consent was obtained from all
keyworkers and patients.

■ Intervention
Keyworkers in the control group continued with standard treatment with their
participating patients. In the intervention group keyworkers added the experimental
intervention to standard treatment. The intervention consisted of a computer mediated
procedure that the keyworker administered every two months in routine meetings with
participating patients. The procedure specified that keyworkers asked their patients to
rate their satisfaction with 11 domains of life or treatment, followed by the question
whether patients wanted any additional or different help in the given domain.
Patients’ answers were entered directly onto a (hand-held) computer. This enabled
keyworkers and patients to immediately display and evaluate a response on a domain
in the context of the responses on all other domains and in comparison with ratings at
previous meetings. The intervention was designed to alter patient-keyworker
interactions, so that the patient’s views of their life and treatment and their needs for
care would become a central point for the dialogue between patient and clinician and
inform all treatment discussions. Recently the intervention has been incorporated in a
routine evaluation method in first episode psychosis services in London [13].

Page 8

8
■ Outcomes
Primary outcome was the patient’s subjective quality of life at the 12 months follow-
up assessment, controlling for the score at baseline. Quality of life was assessed with
the Manchester Short Assessment of Quality of Life (MANSA) [14], in which
patients rate their satisfaction with life in general and different life domains on 7-point
scales ranging form ‘couldn’t be worse’ to ‘couldn’t be better’. The mean score on all
satisfaction ratings is taken as the indicator for subjective quality of life.
Secondary outcomes were number of unmet needs for care and patient satisfaction
with treatment. Needs for care were measured with the Camberwell Assessment of
Need Short Appraisal Schedule, patient version (CANSAS) [15, 16], which assesses
health and social needs across 22 domains, as perceived by the patient. For each
domain it distinguishes between ‘no need’, ‘met need’ and ‘unmet need’. Patient’s
satisfaction with treatment was assessed on the Client Satisfaction Questionnaire
(SCQ-8) [17], which consists of eight items rated from 1 to 4, with higher scores
indicating greater treatment satisfaction. The total number of unmet needs for care
and the sum score of satisfaction ratings at 12 months were studied as secondary
outcomes, controlling for baseline scores.
The DIALOG intervention was hypothesized to increase subjective quality of life
and satisfaction with treatment, and to reduce the number of unmet needs for care [7].
In addition, the influence of the intervention on psychiatric symptomatology was
explored, but no effect was anticipated [7]. Symptomatology was assessed at baseline
and 12 months follow-up with the Positive and Negative Syndrome Scale interview
(PANSS) [18], which assesses positive, negative and general symptoms of
schizophrenia on 7-point scales, with higher scores indicating more severe symptoms.

Page 9

9
The total scores on the three subscales at 12 months were taken as separate outcome
measures of psychiatric symptomatology, controlling again for baseline scores.

■ Mental healthcare settings
The DIALOG trial was conducted in community mental healthcare services in
Granada (Spain), Groningen (the Netherlands), London (United Kingdom), Lund
(Sweden), Mannheim (Germany), and Zürich (Switzerland), covering urban and
mixed urban-rural areas. The number of participating teams per centre varied between
two (Lund) and six (London).
In the exploratory analysis of correlates of difference in effectiveness of the
DIALOG intervention between centres, centre characteristics with respect to patient
group served and mental healthcare provided, will be studied. The patient group
characteristics include (1) mean age, (2) gender distribution, (3) main diagnostic
categories, (4) number of years in mental healthcare, (5) number of admissions to
mental health clinics, and (6) level of functioning of the patient group. The mental
healthcare characteristics consist of (7) mean caseload of keyworkers, (8) number of
years in current job of keyworker, (9) number of face to face contacts per month per
patient, (10) duration of these contacts, (11) duration of all care provided to a patient
per month, and (12) number of meetings during the follow-up period in which the
DIALOG intervention was administered in the intervention group.
The demographic characteristics of the patients, the number of years since first
contact with mental healthcare and the number of hospital admissions (the above
centre characteristic 1, 2, 4 and 5) were taken from the baseline patient interview of
the DIALOG trial. Psychiatric diagnosis (centre characteristic 3) was obtained
through a standardized and computer-based method using operationalized criteria

Page 10

10
(OPCRIT) [19], which were checked in the patient’s medical record. Eligible for the
DIALOG trial were patients with a primary diagnosis of schizophrenia or related
psychotic disorder (ICD-10 F20–F29) [7], who were grouped for the present study
into the categories schizophrenia (F20.0–F20.5), schizoaffective disorders (F25), and
other psychotic disorders (rest). The mean level of functioning of the patient group
(centre characteristic 6) was assessed by the mean baseline scores on quality of life,
number of unmet needs for care, satisfaction with treatment, and symptomatology of
the patients participating in the DIALOG trial. These variables were not combined,
but were analyzed as separate indicators of the level of functioning of the patient
group served.
Characteristics of the mental healthcare provided in the centres were obtained from
interviews with the keyworkers who participated in the DIALOG trial. At baseline the
keyworkers were asked about their total case load size and number of years in the
current job (centre characteristics 7 and 8). The number and duration of face to face
contacts per patient in the previous two months and the total duration of care per
patient (centre characteristics 9 to 11) were assessed for the DIALOG patients, in an
interview with their keyworker, 8 months after the patient’s baseline assessment.
Finally, the (hand-held) computers used in the DIALOG intervention registered the
number of times the intervention was administered during the follow up period in the
intervention group (characteristic 12).
It should be noted that the above centre characteristics (with the exception of
characteristics 7 and 8) only refer to the patients included in the DIALOG trial. This
enables comparison between the study centres. But on the other hand, it does not
show that all participating community mental health centres served a broader

Page 11

11
population of severely mental ill patients, not just the patients with schizophrenia or
related psychotic disorders who were eligible for the DIALOG trial.

■ Analysis
Differences between centres in the effectiveness of the DIALOG intervention are
tested per outcome measure by linear mixed effects analysis, with baseline score for
that measure, length of follow-up, treatment allocation, study centre and interaction
between treatment allocation and study centre as fixed effects, and keyworker as
random effect. Length of follow-up was also taken into account by Priebe et al. [7],
because it is a potentially confounding covariate, that might have introduced post-
randomization variance. Keyworker is included in the analysis as a random effect, to
adjust for the effect of clustering of patients within keyworkers. Effects are tested at a
significance level of .05. Standardized effect sizes and their 95% confidence intervals
will be presented for statistically significant effects, in the form of adjusted mean
differences between factor levels, standardized by the within subjects standard
deviation.
Differences between centres in patient group and mental healthcare characteristics
will be tested by oneway analysis of variance, at a .05 significance level, and
subsequently by pairwise comparisons between centres, using the Bonferroni
correction for multiple testing. Significant differences will be discussed and explored
for their associations with any main centre effect or interaction between centre and
treatment allocation.

Page 12

12
Results

■ Participants
Participants in the DIALOG trial consisted of 507 patients of 134 keyworkers. At the
12 months follow-up assessment, 451 patients were reinterviewed (89% follow-up). A
detailed description of participant flow from recruitment to data analysis has been
presented by Priebe et al. [7].

■ Centre differences
Table 1 here Table 1 shows the characteristics of the participating centres in the DIALOG trial,
with respect to patient group served and mental healthcare provided. The centres
differ on all characteristics studied, except on the age and gender distributions of their
patient populations. In Groningen and Zürich fewer patients are diagnosed as having
schizophrenia than in the other centres, and more as having a psychotic disorder other
than schizophrenia or a schizoaffective disorder. The patients in Granada and London
have a relatively brief history in mental healthcare, with fewer hospital admissions,
while the patients in Mannheim and Lund had a larger number of hospital admissions.
The comparison of patient functioning at the baseline assessment of the DIALOG
trial shows that patients in Mannheim and Zürich experience a relatively high quality
of life, and that in Mannheim the number of unmet needs for care is low. In contrast,
patients in London and Granada have comparatively more unmet needs for care.
Nevertheless, treatment satisfaction is highest in Granada, and lowest in Groningen.
Patients in London and Groningen experience the most positive, negative and general
symptoms, and the patients in Lund and Granada the least.

Page 13

14
■ Effect sizes
Table 3 here Table 3 presents the standardized sizes of the effects found. The uniform effects of the
DIALOG intervention on quality of life (0.20; 95% CI: 0.01 – 0.39) and treatment
satisfaction (0.27; 0.06 – 0.47) are small, which is consistent with the original study
report by Priebe et al. [7]. However, the present study shows that the DIALOG
intervention does not have a uniform effect on unmet needs for care, but that this
effect depends on the centre in which the intervention is implemented. A large
reduction in unmet needs is seen in London (-0.83; -1.29 – -0.36), and Granada (-
0.60; - 1.19 – -0.02), while in the other centres no clear effect on unmet needs is
found. In London there is also a large reduction in general symptoms (-0.87; -1.39 – -
0.34), which – based on marginally significant interactions – is also seen for positive
and negative symptoms in London (-0.87; -1.36 – -0.39, and -0.78; -1.31 – -0.26,
respectively), but not in the other centres. These other centres show no effect of the
DIALOG intervention on symptoms, as reported by Priebe et al. [7] for the study as a
whole.
In addition to the intervention effects, centre effects are found on treatment
satisfaction and positive and negative symptoms. On top of the uniform intervention
effect on treatment satisfaction, patients in Granada – both in the experimental and
control group – show a more favourable development over the follow-up period than
the patients in the other centres (p<.01 on all pairwise comparisons). And when the
marginally significant interactions between intervention and centre on positive and
negative symptoms are disregarded: (1) patients in Granada, Groningen and Lund are
found to experience a greater decrease in positive symptoms than patients in London,
as is true for patients in Lund compared to patients in Mannheim and Zürich, and (2)
patients in Granada and Lund experience a greater decrease in negative symptoms

Page 14

15
than patients in London, Mannheim and Zürich, as is true for patients in Groningen
compared to patients in Zürich (p<.05 on all pairwise comparisons), irrespective of
the study group they were in.

■ Association between effectiveness and centre characteristics
Apart from the uniform effectiveness of the DIALOG intervention on quality of life
and treatment satisfaction, the intervention proved to be effective on symptomatology
in London and on number of unmet needs for care in London and Granada. Table 1
shows that of all centres, the patients in London experienced the most positive,
negative, and general symptoms at baseline, as well as the most unmet needs for care,
with the patients in Granada being second on unmet needs for care. This suggests that
the effectiveness of the DIALOG intervention on symptomatology and unmet needs
for care may be associated with the baseline level of functioning on these aspects, in
such a way that the intervention is more effective in patient groups with more
symptoms, or more unmet needs for care. On the patient group level, the correlation
between the mean level of functioning at baseline in a centre (as reported in table 1)
and the effect size of the intervention in that centre on that aspect of functioning (as
reported in table 3) is -0.46 for positive symptoms, -0.32 for negative symptoms, -
0.61 for general symptoms, and -0.92 for number of unmet needs for care (n=6). All
these correlations indicate a stronger effectiveness of the intervention (i.e. a greater
reduction in symptoms or unmet needs for care) in centres with patient groups that
function more poorly at baseline (i.e. that have more symptoms or more unmet needs
for care). On the individual patient level – in linear mixed effects models that do not
take the centre into account – significant interactions between the baseline level of
functioning and the intervention effect on level of functioning at follow-up are found

Page 15

16
for positive symptoms (F=4.69; p=.03), general symptoms (F=4.89; p=.03), and
unmet needs for care (F=4.53; p=.03), but not for negative symptoms (F=0.18; p=.67).
Again, these interactions are all in the direction of a greater intervention effect for
patients with a poorer level of baseline functioning.
Besides their relatively high number of psychiatric symptoms at baseline and unmet
needs for care, the patient groups in London and Granada also stand out for their
somewhat shorter history in mental healthcare and fewer hospital admissions, than the
patients in the other centres. In addition, the care the patients in London and Granada
receive from their keyworkers, is comparatively limited in duration. As for the
baseline level of functioning, these latter patient and care characteristics are also
correlated on a patient group level with the outcomes that showed non-uniform
effectiveness of the intervention between centres (range .26–.88; n=6). All of these
correlations indicate more effectiveness of the intervention (i.e. a greater reduction in
symptoms or unmet needs for care) in centres with patients who – on average – have a
shorter history in mental healthcare and fewer hospital admissions, or in which the
keyworkers have less time per patient for care. But on the individual patient level, a
significant influence of the patient or care characteristic on intervention effectiveness
was found in only on one of the twelve interactions tested: a greater reduction in
unmet needs for care is seen in patients with a shorter history in mental healthcare
(F=3.24; p=.07). The present study, therefore, provides more extensive empirical
support for the assertion that the effectiveness of the DIALOG intervention on unmet
needs for care and symptomatology is related to the baseline level of functioning of
the patients on these outcomes, than for the idea that it would be related to the number
of years the patients are in mental healthcare, the number of times they have been