PEER-ADMINISTERED INTERVENTIONS FOR DEPRESSION:
A META-ANALYTIC REVIEW
by
Amanda E. B. Bryan
_____________________
Copyright © Amanda E. B. Bryan 2013
A Dissertation Submitted to the Faculty of the
Department of Psychology
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
2013
2
THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
As members of the Dissertation Committee, we certify that we have read the dissertation
prepared by Amanda Bryan
entitled Peer-Administered Interventions for Depression: A Meta-Analytic Review
and recommend that it be accepted as fulfilling the dissertation requirement for the
Degree of Doctor of Philosophy
____________________________________________________________Date: 3/8/2013
Hal Arkowitz
____________________________________________________________Date: 3/8/2013
John Allen
___________________________________________________________ Date: 3/8/2013
David Sbarra
___________________________________________________________ Date: 3/8/2013
Richard Bootzin
Final approval and acceptance of this dissertation is contingent upon the candidate's
submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and
recommend that it be accepted as fulfilling the dissertation requirement.
____________________________________________________________Date: 3/8/2013
Dissertation Director: Hal Arkowitz
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an
advanced degree at the University of Arizona and is deposited in the University Library
to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided
that accurate acknowledgment of source is made. Requests for permission for extended
quotation from or reproduction of this manuscript in whole or in part may be granted by
the copyright holder.
SIGNED: Amanda E. B. Bryan
4
ACKNOWLEDGEMENTS
Many thanks to my advisor, Hal Arkowitz, who has been a supportive and
inspiring guide and mentor. Thanks also to my committee members, John Allen, Dave
Sbarra, and Dick Bootzin, each of whom has been a wonderful and influential teacher to
me over the last several years. My sincerest appreciation to research assistants Karlee
Frye, Sean Gallagher, and Joseph Santa Cruz for their generous and capable work on
coding the studies included in this analysis. Finally, warmest thanks to my entire family
for their love, patience, and support as I’ve found my circuitous way through graduate
school.
5
DEDICATION
For Andrew, and the many years of discovery that await us.
6
TABLE OF CONTENTS
LIST OF FIGURES……………………………………………………………..8
LIST OF TABLES……………………………………………………………....9
ABSTRACT…………………………………………………………………….10
1. BACKGROUND AND RATIONALE……………………………………....12
Effectiveness of nonprofessional mental health services………………….......13
Prior meta-analyses of peer interventions for depression….............................15
Why peer interventions?.....................................................................................17
The present study…...........................................................................................19
2. METHOD…......................................................................................................20
Study selection…...............................................................................................20
Data extraction…..............................................................................................21
Effect size calculation…....................................................................................22
Weighted mean effect size calculation…...........................................................25
Moderation analysis…......................................................................................27
Alternate metrics of outcome…........................................................................28
Follow-up data….............................................................................................28
Publication bias…............................................................................................29
3. RESULTS…....................................................................................................32
Study characteristics…....................................................................................32
Depression measurement….............................................................................32
Observed effect sizes…....................................................................................33
7
TABLE OF CONTENTS – CONTINUED
Moderation models….....................................................................................35
Follow-up effect sizes…..................................................................................37
Outcomes using alternate metrics…...............................................................38
Impact of providing peer support…...............................................................39
Publication bias…..........................................................................................39
4. DISCUSSION…............................................................................................41
Summary and interpretation of findings….....................................................41
Clinical significance of findings….................................................................44
Methodological limitations….........................................................................45
Future directions…........................................................................................46
FIGURES…......................................................................................................49
TABLES…........................................................................................................61
APPENDIX A: CODING FORM.....................................................................73
APPENDIX B: METHODOLOGICAL QUALITY SCORE (MQS)
CODING SYSTEM…............................................................77
REFERENCES….............................................................................................78
8
LIST OF FIGURES
Figure 1:
Flow of studies.............................................................................49
Figure 2:
Forest plot of effect sizes comparing
PAI to active therapy conditions..................................................50
Figure 3:
Forest plot of effect sizes comparing
PAI to TAU/WL conditions.........................................................51
Figure 4:
Forest plot of effect sizes of PAI conditions
(change from pre- to post-treatment)...........................................52
Figure 5:
Moderators of standardized mean gain (SMG) for
PAI conditions (change from pre- to post-treatment)..................53
Figure 6:
Moderators of standardized mean difference (SMD)
comparing PAI to active therapy conditions...............................54
Figure 7:
Moderators of standardized mean difference (SMD)
comparing PAI to TAU/WL conditions......................................55
Figure 8:
Methodological quality is not significantly
associated with pre-post effect size.............................................56
Figure 9:
Methodological quality is not significantly
associated with between-groups effect size
(a) PAI compared with active therapy conditions......................57
(b) PAI compared with TAU/WL...............................................58
Figure 10:
Funnel plot of effect sizes comparing
PAI to active therapy conditions.................................................59
Figure 11:
Funnel plot of effect sizes comparing
PAI to TAU/WL conditions........................................................60
9
LIST OF TABLES
Table 1:
Descriptive characteristics of clinical trials
involving PAIs..........................................................................61
Table 2:
Characteristics of depression symptom
measurement instruments..........................................................67
Table 3:
Results of included studies.......................................................68
Table 4:
Mean SMD estimates (PAI compared with
active therapy conditions) adjusted for four
models of selection (publication) bias......................................71
Table 5:
Mean SMD estimates (PAI compared with
TAU/WL conditions) adjusted for four
models of selection (publication) bias......................................72
10
ABSTRACT
A variety of psychotherapies have been demonstrated to be efficacious and effective
treatments for depression. The cost of psychotherapy, however, and its low availability
in some contexts pose significant treatment barriers for many depressed individuals.
Based on the idea that peers (i.e., individuals who have successfully recovered from
similar problems) may be uniquely able to provide empathy and support to those
currently receiving treatment, some community mental health centers have implemented
peer treatment models that employ recovered former clients as cost-effective adjunct
providers. The effectiveness of these and other peer-administered interventions (PAIs)
has not been well-established. The current study is a meta-analysis of the existing
outcome research on PAIs for depression. Twenty-six studies were identified as eligible
for inclusion and yielded 30 between-groups effect sizes and 29 pre-post PAI effect sizes.
Study characteristics and methodological quality were coded and random-effects models
were used to calculate and compare mean effect sizes. PAIs produced significant pre-topost treatment reductions in depression symptoms that were comparable to those found in
well-established professionally-administered interventions (.4554). In direct
comparisons, PAIs performed as well as professionally-administered treatments (.0848).
but not significantly better than treatment-as-usual (e.g., periodic physician check-ins or
availability of community mental health services) and wait-list control conditions (.0978).
These findings did not change after adjusting for the moderate degree of publication bias
in the data. Moderation models revealed that professionally-co-administered PAIs
produced significantly worse outcomes than those that were purely peer-administered,
11
and that educational/skills-based PAIs (but not supportive PAIs) produced better
outcomes compared with professional treatments. Limitations of this analysis included
the heterogeneity of the included interventions and the lack of data on mediators and
moderators. Still, these findings suggest that PAIs have promise as effective depression
treatments and are worthy of further study.
12
1. BACKGROUND AND RATIONALE
“You never really understand a person until you consider things from his point of
view... Until you climb inside of his skin and walk around in it.”
― Harper Lee, To Kill a Mockingbird (1960)
Approximately 16.6% of Americans will meet the diagnostic criteria for Major
Depressive Disorder (MDD) during their lives (Kessler et al., 2005). The costs of MDD
both to individuals and to society are substantial, with suicide attempt rates 11 times
higher in individuals with lifetime diagnoses of depression than in individuals without
psychological diagnoses (Kessler, Borges, & Walters, 1999), and an estimated annual
economic burden of $83.1 billion incurred through direct, suicide-related, and lostproductivity costs of depression in the United States (Greenberg et al., 2003). Yet just
over half of individuals who have met DSM criteria for major depression within the last
year have received any treatment, and only about one-fifth have received adequate
treatment, defined as receiving pharmacotherapy for at least 30 days with at least four
physician visits, or at least eight psychotherapy sessions (Kessler et al., 2003).
Psychotherapeutic treatments for depression have demonstrated comparable
efficacy to antidepressant medication and possibly superior outcomes in preventing
relapse (Hollon et al., 2002), but the proportion of depressed individuals who receive
psychotherapy has declined in recent decades, as has the mean number of psychotherapy
sessions for depressed people who do receive it (Olfson et al., 2002). Contributing to this
13
discrepancy is the high cost of psychotherapy, which decreases the rate at which
psychotherapy is utilized in outpatient settings (e.g., Simon et al., 1996), especially
among lower-SES individuals (e.g., Steele, Dewa, & Lee, 2007). The expenses of
training and reimbursement for master’s- and doctoral-level providers contribute to the
costs of therapy, yet it is far from clear that level of training is correlated with treatment
quality or outcome (Stein & Lambert, 1984; Bickman, 1999); thus, exploring areas of
treatment need that could be fulfilled at a lower cost by providers with less training is
worthwhile. Along these lines, it has been suggested (e.g., Pfeiffer et al., 2011) that peer
support services could substantially reduce the barriers to psychosocial treatment of
depression by making services available at low costs, and by circumventing some of the
stigma associated with traditional psychotherapeutic and pharmacological treatments.
Effectiveness of nonprofessional mental health services
Similar to but not synonymous with peer support services, services provided by
paraprofessionals or “lay therapists” (i.e., individuals without professional credentials as
mental health providers, but not necessarily with personal experiences of the problems
they are treating) have shown promise in numerous research studies, suggesting that
individuals with little or no prior training as mental health providers may be able to
contribute effectively to treatments for depression. Paraprofessional therapists may
receive brief, role-specific preparatory training, but do not hold professional credentials
as health or mental health professionals, and generally have not received broad or
extensive training in mental health treatment. Den Boer, Wiersma, Russo, and van den
14
Bosch (2005) meta-analyzed controlled studies of paraprofessional treatment of anxiety
and depression, and found that the outcomes achieved by paraprofessionals delivering a
variety of cognitive-behavioral and supportive interventions were not statistically
different from those achieved by professionals, and were significantly better than waitlist
and placebo control conditions. In a recent review of cognitive-behavioral treatments for
anxiety and depression (Montgomery, Kunik, Wilson, Stanley, & Weiss, 2010), the
authors identified four studies that had compared professional and paraprofessional
outcomes and concluded that “paraprofessionals generally achieved comparable overall
outcomes” (p. 57). These and other summaries (e.g., Durlak, 1979; Hattie, Sharpley, &
Rogers, 1984; Berman & Norton, 1985; Christensen & Jacobson, 1994) have concluded it
is possible that high-quality treatment can be delivered by individuals who are not trained
as therapists.
Taking nonprofessional treatment services a step further, some community mental
health organizations have begun to integrate peer “consumer-provided” care – that is,
services delivered by current or former clients – into their treatment programs with
promising results. Like paraprofessionals, consumer providers typically have neither
professional mental health credentials nor a broader training background in a mental
health field; they, too, may receive brief, role-specific preparatory training, or they may
simply rely on their own treatment experiences for preparation. Chinman, Rosenheck,
Lam, and Davidson (2000) evaluated a case management program for seriously mentally
ill individuals in which some case managers were consumer providers and others were
not mental health care consumers; they found no significant interaction between time and
15
case manager type on clinical outcomes over 12 months. In a broader investigation of
consumer-provided services nationwide, Goldstrom and colleagues (2006) reported on a
survey conducted by the Substance Abuse and Mental Health Services Administration
(SAMHSA) of mutual support groups, self-help organizations “run by and for mental
health consumers,” and consumer-operated mental health services. They found that these
groups exceeded in number traditional mental health organizations, and that they served
upwards of half a million people per year. Regarding the effectiveness of these services,
Doughty and Tse (2011) reviewed 29 studies of consumer-led mental health services and
concluded that they produced outcomes similar to the outcomes of traditional mental
health services. Taken together, these findings suggest that peer-administered mental
health interventions are widely used and possibly effective in improving outcomes.
Prior meta-analyses of peer interventions for depression
In the more specific realm of depression, two recent meta-analyses have evaluated
the literature on interventions for depression that focus on the development of
relationships with peers (as opposed to with mental health professionals). Pfeiffer and
colleagues (2011) analyzed “peer support interventions” for depression, which were
defined as interventions that “placed individuals with current depression in regular
contact with at least one other person with either current or prior depression” (p. 30).
They included 10 controlled studies, and concluded that outcomes of peer support
interventions did not differ significantly from outcomes of group CBT, and that the
effects of peer support interventions compared with usual care were similar in size to the
16
effects of CBT compared with usual care in other published analyses. Mead and
colleagues (2010) evaluated the effect on depression symptoms of “befriending,” defined
as “an intervention that introduces the client to one or more individuals whose main aim
is to provide the client with additional social support through the development of an
affirming, emotion-focused relationship over time” (p. 96). They identified 24 studies to
include in their analyses, and found that befriending had a modest effect on depression
symptoms compared with usual care or no treatment, and was less effective than alternate
interventions, including cognitive-behavioral and family-based interventions.
In addition to their somewhat different results, each of these analyses has features
that complicate interpretation. First, in both analyses, treatments that were administered
or facilitated by a mental health professional were included as long as the treatment also
involved some degree of interaction with nonprofessional peers. This makes it difficult
to interpret the extent to which peer support actually had an effect on symptoms,
especially because neither analysis examined professional facilitation as a potential
moderator of treatment outcome. Research has shown (e.g., Toro et al., 1988) that the
presence of a mental health professional in a mutual help group alters the behaviors of
group members and their perceptions of the group.
Second, the inclusion of studies was not comprehensive in either analysis: some
studies were either missed or excluded for reasons that are unclear, and internet-based
interventions were not included in either analysis in spite of other evidence that
computer-based treatments for depression produce effect sizes similar to face-to-face
treatments (e.g., Andersson & Cuijpers, 2009). Additionally, each of these analyses
17
included only controlled studies, yet a substantial minority of the published studies in this
area have been uncontrolled; therefore, many relevant findings were excluded. Although
controlled studies produce effect sizes that are easier to interpret than pre-post effect sizes
(see p. 23), pre-post comparisons may also be informative when uncontrolled studies
comprise much of a body of literature, because there may be systematic differences
between controlled and uncontrolled studies that could affect outcome. Researcher
allegiance is a hypothetical example: community organizations with a strong allegiance
to peer support as a treatment approach may also be less likely than university researchers
to have the resources to conduct controlled studies.
Why peer interventions?
Why might one expect interventions administered by non-therapist peers to
effectively reduce depression symptoms? One reason is that peer-administered
interventions may provide a relationship characterized by an especially high degree of
empathy. Although there is a dearth of evidence regarding the effect of empathy on
depression symptoms in everyday interactions, it has been demonstrated several times
that therapist empathy is associated with better therapy outcome (see Greenberg, Elliot,
Watson, & Bohart, 2001, for a brief meta-analysis), and there is evidence that therapist
empathy causally affects outcome in cognitive therapy for depression (Burns & NolenHoeksema, 1992). This suggests that receiving empathy improves symptoms, at least in
the context of therapy. Yet it is not clear that therapists are better equipped than peers to
provide a high degree of empathy. Hassenstab and colleagues (2007) showed that
18
therapists scored higher than non-therapist control subjects (matched on age, gender,
intelligence, and education level) on self-report measures of cognitive empathy (i.e., the
ability to understand others’ emotions, thoughts, or intentions), but lower than controls on
emotional empathy (i.e., one’s own emotional responses to others’ affective states),
suggesting that therapists may actually sacrifice some empathy in favor of emotional
control.
On the other hand, peers who are identified specifically because of their similar
experiences may be uniquely equipped to provide empathy. Several studies have
demonstrated that individuals are more empathetic in response to others’ distress if they
themselves have experienced a similar distressing situation. For example, Batson and
colleagues (1996) presented subjects with vignettes that described an adolescent
experiencing distress about either severe acne or interpersonal rejection. They found that
women, but not men, reported experiencing more empathy if they had personally
experienced the same type of distress as the adolescent in the vignette they had read.
Others have reported the similar effects in women only across different targets of
empathy, including fear of darkness, fear of abandonment, and loss of a pet (Eklund,
Andersson-Straberg, & Hansen, 2009), and pregnancy (Hodges, Kiel, Kramer, Veach, &
Villanueva, 2010). Furthermore, it appears that therapists are no exception: they reported
experiencing greater empathy in response to fictional clients if they were able to generate
“reference points” from their own experiences that were in some way similar to the
clients’ (Hatcher et al., 2005).
19
The present study
The research summarized above suggests that (1) peers may be especially wellsuited to provide social support characterized by a high degree of empathy; (2) receiving
empathy may reduce depression symptoms; and (3) previous evaluations of research on
peer-support interventions for depression have not been comprehensive and have
included studies that were administered primarily by professionals. These findings
jointly support the need for the present analysis, which meta-analytically aggregates the
published effects of peer-administered interventions on depression symptoms using (a) a
highly specific definition of “peer-administered” that explicitly qualifies any involvement
by health or mental health professionals, and (b) a comprehensive search strategy to
identify studies that have been excluded from previous analyses. Additionally, the
analysis reported here includes examinations of several possible moderators of treatment
outcome, and evaluates the clinical significance of the findings. Finally, the effects (if
any) of administering treatment to peers on well-being will be examined if a sufficient
number of studies report data on this.
20
2. METHOD
Study selection
For this analysis, a peer-administered intervention (PAI) was defined as an
intervention administered by one or more lay individuals who have experienced problems
similar to those being treated, with minimal or no assistance from a health or mental
health professional. Interventions meeting this definition might include a support group
for bereaved spouses facilitated by someone who is widowed; an internet support forum
for depression in which depressed individuals provide support to each other; and a
program in which mothers with prior perinatal depression have regular contact with
mothers who are currently experiencing perinatal depression. Under this definition, it is
acceptable for peer facilitators to have received supervision or training from a
professional in preparation to administer the intervention, and/or ongoing supervision
from a professional during the course of administering the intervention. Interventions
that were primarily administered by a peer, but had a professional present in a supportive
role, were also included in this analysis, and the effect of having a professional present
was examined statistically (see Results section).
To be included in this analysis, studies were required to report outcomes in which:
(a) at least one condition involved a PAI as defined above; (b) depression symptoms were
measured as an outcome; (c) participants were adult outpatients with diagnosed
depression or elevated depression scores on a well-validated symptom measure; and (d)
the report was published in a peer-reviewed journal and was available in English.
Samples were not required to be defined by a diagnosis with depression; studies of
21
interventions for chronic illness populations or other stressed populations could be
included, but only if those samples met minimum clinical cutoffs for probable depression
on well-established measures. In order to allow for the most comprehensive review
possible, uncontrolled and quasi-experimental studies were included if they reported, at a
minimum, pre- and post-treatment outcomes for a PAI condition.
Figure 1 shows the progression of the study selection process. Searches of the
PsycINFO and PubMed online databases were conducted in November 2011 using the
search terms “(Peer OR mutual OR paraprofessional) AND depress*”. The abstract of
each search return was reviewed, and articles that were clearly ineligible (e.g., because
they did not report on an intervention study) were discarded, leaving a list of “potentiallyeligible” studies. Reference sections of these studies were then hand-searched to identify
any additional potentially-eligible studies that the database searches may have missed.
Finally, the complete text of each potentially-eligible study was reviewed and its status as
eligible or ineligible was determined. For articles that reported on an eligible study but
did not include sufficient statistical information to calculate effect size, the needed data
were requested from corresponding authors.
Data extraction
A sample coding form, which lists all coded variables, and the Methodological
Quality Scale, which raters also completed for all studies, are included as Appendices A
and B, respectively.
22
Each study was independently coded by two raters: the author (Bryan) and a
trained undergraduate research assistant. Interrater discrepancies were resolved by
discussion to decide values for the final dataset. Interrater reliability for the continuous
Methodological Quality Score (MQS) was calculated using Spearman’s rank correlation
coefficient (Spearman, 1904), calculated as
i
i
( xi
( xi
x)( yi
x) 2
i
( yi
y)
y) 2
where xi and yi are each rater’s raw ratings converted to ranks. The MQS showed good
interrater reliability (ρ = .82).
Interrater reliability for categorical variables was calculated using Cohen’s kappa
(Cohen, 1960), calculated as
Pr(a) Pr(e)
1 Pr(e)
where Pr(a) is the relative observed agreement among raters, and Pr(e) is the hypothetical
probability of chance agreement. Kappas for categorical variables were all at least
moderately high, ranging from .71-.89.
Effect size calculation
Standardized mean gains. Pre-post effect sizes (standardized mean gains, or
SMGs) were calculated for all PAI conditions in controlled and uncontrolled studies,
defined as
23
X T2
ES sg
X T1
sp
ES sg2
2(1 r )
n
SEsg
,
2n
,
1
,
SEsg2
wsg
where X T 1 is the mean at Time 1, X T 2 is the mean at Time 2, n is the common sample
size at Time 1 and Time 2, r is the correlation between Time 1 and Time 2 scores, and
s p is the pooled standard deviations of Time 1 and Time 2 means, calculated as
sp
s 2T 1
s 2T 2 / 2
As Westen and Morrison (2001) note, interpretation of the SMG is limited for many of
the same reasons as its components: effects of the treatment are likely to be correlated
with placebo effects, the effects of time’s passage, and the effects of common factors. At
the same time, it may be useful to compare SMGs of PAI conditions with pre-post effect
sizes from other published meta-analyses of empirically supported therapies for
depression – interpreting these comparisons judiciously, of course.
Standarized mean differences. For studies that included randomized control
groups, between-groups effect sizes were calculated as standardized mean differences,
defined as (Lipsey & Wilson, 2001)
ES sm
X G1
X G2
sp
,
24
where X G1 the posttest mean for Group 1 (comparison group in this analysis), X G 2 is
the posttest mean for Group 2 (PAI group in this analysis), and s p is the pooled posttest
standard deviations of the two group means, defined as
sp
2
(ntx 1) s tx2 (nctrl 1) s ctrl
.
(ntx 1) (nctrl 1)
Because standardized mean differences tend to be upwardly biased in samples of fewer
than 20 participants (Hedges, 1981), all standardized mean differences were corrected
using Hedges’ correction factor, and these corrected (unbiased) effect size estimates were
used in all analyses:
'
ES sm
SEsm
1
3
ES sm ,
4N 9
n G1 n G 2
n G1 n G 2
wsm
'
( ES sm
)2
,
2( n G 1 n G 2 )
1
2
SEsm
For this analysis, two sets of SMDs were calculated: those comparing a PAI
condition to another active intervention (e.g., a professionally-administered
psychotherapy group), and those comparing a PAI condition to a treatment-as-usual
(TAU) or wait-list (WL) condition. The decision to combine TAU and WL control
conditions into a single meta-analytic comparator was based on the observation that the
distinction between the two categories was often blurry or impossible to make in a
25
systematic way. For example, participants in chronic illness samples almost invariably
continued to receive standard medical treatment during the course of the study, even if
they were in a wait-list condition for the PAI or other psychosocial interventions; these
standard medical treatments may or may not have involved “therapeutic” interactions
with health care professionals. Similarly, across sample types, conditions described as
“wait-list control” often included periodic check-ins or “case management” visits that
may have had therapeutic value. Conversely, in conditions described as TAU or
“standard care,” often this simply meant that ongoing care of some kind was available to
participants, not necessarily that it was utilized during the study period. The difficulty of
distinguishing between TAU and WL conditions in this sample was corroborated by low
interrater reliability between the two (κ = .25). Thus, rather than trying to categorize
based on incomplete data or arbitrary differences, the two were combined as a single
category. This had the added benefit of increasing statistical power for comparisons
between PAIs and TAU/WL conditions.
Weighted mean effect size calculation
In order to summarize effect sizes across studies, weighted mean effect size
estimates were calculated separately for between-groups ESs (SMDs) and within-groups
ESs (SMGs) using the formula
ES
( wi ES i )
wi
26
where wi is the inverse variance weight for ESi. Upper and lower limits of the 95%
confidence intervals around mean effect sizes were calculated as:
ES
z (1
)
(SEES )
where SEES is the standard error of the mean effect size, calculated as
1
.
wi
SEES
To determine whether the effect sizes included in these means all estimate the same
population effect size (e.g., Hedges, 1982, Rosenthal & Rubin, 1982), the heterogeneity
statistic Q was computed for each mean effect size:
Q (
2
i
wi ES )
(
wi ES i ) 2
wi
Q falls in a chi-square distribution with k – 1 degrees of freedom, where k is the number
of effect sizes included in its calculation (Hedges & Olkin, 1985). If Q is significant for a
given mean effect size (i.e., the null hypothesis of homogeneity is rejected), this indicates
that the dispersion of effect sizes around that mean is greater than would be expected
from sampling error alone (e.g., that differences in particular study characteristics
account for some of the variance in effect size). In these cases, it may be appropriate to
use random effects models, which use an adjusted inverse variance weight based on the
assumption that in addition to subject-level sampling error, there is an additional
randomly-distributed study-level source of error.
27
Moderation analysis
First, fixed-effects (ordinary least squares) models were calculated for all
hypothesized moderators using David Wilson’s MetaReg and MetaF macros for SPSS
(Wilson, 2010). Fixed-effects models partition effect-size variance into two categories:
between-groups variance (i.e., variance explained by the moderator), and within-group
variance (i.e., residual unexplained variance, which is assumed to be subject-level
sampling error). If Qw, the heterogeneity statistic for the residual variance, is significant,
this indicates that modeling only systematic variance and subject-level sampling error (as
in a fixed-effects model) is insufficient (i.e., fails to meet the model’s assumptions that
these are the only sources of variability), and a mixed-effects model may be appropriate
(Lipsey & Wilson, 2001).
Because fixed-effects models produced significant Qw values for all of the mean
effect sizes (see Results section), we turned to mixed-effects models using method-ofmoments estimation procedures, which Wilson’s (2010) macros can also calculate.
Mixed-effects models assume “that there is a remaining unmeasured (and possibly
unmeasurable) random effect in the effect size distribution in addition to sampling error.
That is, variability in the effect size distribution is attributed to systematic (modeled)
between-study differences, subject-level sampling error, and an additional random
component” (Lipsey & Wilson, 2001, p. 124). Mixed-effects models have lower
statistical power to detect moderation effects and result in wider confidence intervals than
fixed-effects models; they are, however, less prone to Type I errors, and their
assumptions are not violated by heterogeneity as the assumptions of fixed-effects models
28
are (Lipsey & Wilson, 2001). Thus, they were deemed more appropriate for testing
moderation in this set of data.
Alternate metrics of outcome
In addition to conventional effect size estimates, two alternate metrics of outcome
were examined to the extent that they were reported or available in order to determine
whether any treatment effects were clinically (not just statistically) significant. Clinical
significance has been defined as “the extent to which therapy moves someone outside the
range of the dysfunctional population or within the range of the functional population”
(Jacobson & Truax, 1991, p. 12), and is an important consideration when determining
whether detected effects represent meaningful changes in symptomology. First,
percentage of participants considered “improved” at treatment termination was coded
when reported. Second, posttreatment symptomology (i.e., mean scores on outcome
measures) was compared to published or widely-used guidelines for what constitutes
clinical elevation on the respective measures of depression symptoms.
Follow-up data
When available, follow-up data were examined to evaluate the durability of
treatment effects. SMGs and SMDs were calculated in the same manner as for pre-post
data (with SMGs indicating effect size between post-test and follow-up) When a study
29
reported on multiple follow-up time points (e.g., three- and six-month follow-ups), data
from the last available (e.g., the six-month follow-up) were used here.
For each set of follow-up effect sizes, two variables were tested as possible
moderators using the same procedures as described previously. First, the length of the
follow-up period was tested, as it was possible that shorter follow-ups would indicate
better maintenance simply because there was less time for symptoms to recur (and not
because interventions had a particularly enduring effect). Second, effect size at
posttreatment was tested as a moderator to determine whether PAIs that were more
effective in the short-term also produced more enduring benefits.
Publication bias
Several methods were considered for examining publication bias. The two most
commonly-reported indicators of publication bias, the funnel plot and the fail-safe N,
each have features that make reliable and valid interpretation difficult (Sutton, 2009),
rendering them less-than-ideal as standalone indicators. Funnel plots (Light & Pillemer,
1984) are scatterplots of effect size on the x-axis and standard error on the y-axis; in data
without publication bias, these plots are generally pyramid-shaped and symmetrical.
Interpretation of funnel plots typically relies on visual inspection for asymmetry.
Although funnel plots can be useful for developing hypotheses about publication bias,
interpreting them objectively and/or quantitatively presents challenges; furthermore,
when study characteristics moderate effect sizes, funnel plots can be misleading because
they effectively superimpose effect sizes drawn from multiple populations (Vevea &
30
Woods, 2005). The fail-safe N (Rosenthal, 1979; Orwin, 1983) is “the number of studies
with an effect size of zero needed to reduce the mean effect size to a specified or criterion
level” (Lipsey & Wilson, 2001, p. 166). This method has been criticized for its
assumption that all missing studies have effect sizes of zero, which is unlikely to be
reliably true (Vevea & Woods, 2005). Additionally, there is no theoretically-based
criterion for determining a threshold number of missing studies as indicative of
publication bias, and such a number may not be particularly informative anyway; as Field
and Gillett (2010) write, “…any fail-safe N method addresses the wrong question: it is
usually more interesting to know the bias in the data one has and to correct for it than to
know how many studies would be needed to reverse a conclusion” (p. 686). One
commonly-used method for correcting for publication bias, the iterative “trim-and-fill”
technique (Duval & Tweedie, 2000), produces a mean effect size corrected for the
number of studies estimated to be missing; this method, however, can lead to
overcorrection, particularly in datasets that contain heterogeneity, because it assumes that
all of the missing studies are those with the smallest effect sizes (Vevea & Woods, 2005).
Following the recommendation of Field and Gillett (2010), the analysis reported
here utilized Vevea and Woods’s (2005) weight-function sensitivity analysis to estimate
and correct for publication bias. Generally speaking, weight-function models estimate
the influence of different study characteristics on the likelihood of publication and make
corrections that are weighted according to these influences, thus avoiding the assumption
that missing studies are simply those with the smallest effect sizes. Typically, weightfunction models require very large samples sizes (k > 100; e.g., Vevea & Hedges, 1995),
31
but Vevea and Woods (2005) offer a modification for use with smaller-scale metaanalyses that “impose[s] a set of fixed weights determined a priori and chosen to
represent a specific form and severity of biased selection” (p. 432). This procedure
provides corrected effect size estimates for four different models of selection bias,
allowing the user to judge the most likely conditions of publication bias in his data (e.g.,
based on visual inspection of funnel plots) and select the model that most closely
represents those conditions. The models reported in this analysis were calculated using
the macros provided by Field and Gillett (2010).
32
3. RESULTS
Study characteristics
Search procedures identified 26 studies meeting inclusion criteria for which
sufficient data to calculate at least one effect size were either reported or obtainable from
authors. Included studies were published between 1988 and 2011 and reported on a total
of 29 PAI conditions. Twenty studies included at least one control group; in two of these
studies, however, participants were not randomly assigned to conditions, and only prepost data from the PAI conditions from those studies were included in this analysis. In
all, this yielded 30 effect sizes of PAI compared with another condition, and 29 pre-post
PAI effect sizes.
In all included studies, participants’ average pre-treatment depression scores met
minimum clinical cutoffs for likely diagnosable depression. In nine studies, elevation in
depression symptoms was the primary defining feature of the sample. Eleven studies’
samples were defined primarily as having a chronic illness, and the remaining six were
identified primarily by the experience of a severe psychosocial stressor. Additional study
characteristics are summarized in Table 1.
Depression measurement
Table 2 describes the measures of depression symptoms that were used in
included studies, including frequency of use and threshold scores for probable
depression.
33
Observed effect sizes
Results of included studies are summarized in Table 3. Effect sizes are shown in
Figures 2-4.
Standardized mean gains (pre-post effect sizes). It should be noted that, although
improvement in depression symptoms is indicated by a decreased score on all included
depression measures, for the sake of consistency all signs were reversed so that a positive
SMG indicates that symptoms improved from pre- to post-treatment.
The mean SMG for all PAI conditions using a fixed-effects model was .2194 (N =
29, p < .0001, 95% CI: .2003 - .2385), indicating that on average, depression scores
decreased between pre- and post-treatment in PAI conditions. However, the
heterogeneity statistic was highly significant (Q = 890.84, df = 28, p < .0001), suggesting
that the effect sizes comprising this mean do not all estimate the same population mean.
Using a random-effects model to account for the heterogeneity increased the estimated
mean SMG to .4554 (p < .0001, 95% CI: .3391 - .5717). As a standalone effect size
indicator, this estimate must be interpreted cautiously.
Standardized mean differences (between-groups effect sizes). It should be noted
that the PAI condition was considered “Group 2” for all SMD calculations; therefore, a
positive SMD indicates that PAI conditions produced fewer symptoms (i.e., better
outcomes) at post-treatment than control conditions.
Using a fixed-effects model, the mean SMD comparing PAI conditions to active
control conditions (i.e., other, non-PAI therapies) was .0306 (N = 13, p = .6664, 95% CI:
-.1085 - .1696), indicating that posttreatment depression scores were not significantly
34
different between PAI and active-control conditions. The heterogeneity statistic was
significant (Q = 30.15, df = 12, p = .0027), suggesting greater variability than would be
expected from sampling error if all effect sizes were estimates of the same population
mean; thus, a random-effects model may be more accurate. The mean PAI-vs-active
control SMD using a random-effects model was .0848 (p = .4706, 95% CI: -.1455 .3151), indicating that the effect size did not achieve statistical significance even after
accounting for the heterogeneity in the data.
Using a fixed-effects model, the mean SMD comparing PAI conditions to
TAU/WL conditions was .0913 (N = 17, p = .0235, 95% CI: .0123 - .1703), indicating
that posttreatment depression scores were significantly lower in PAI conditions than in
TAU/WL conditions. Again, the heterogeneity statistic was significant (Q = 72.88, df =
16, p < .0001), suggesting greater variability than would be expected from sampling error
if all effect sizes were estimates of the same population mean; thus, a random-effects
model may be more accurate. The mean PAI-vs-TAU/WL SMD using a random-effects
model was .0978 (p = .2966, 95% CI: -.0859 - .2815), indicating that the effect size did
not maintain statistical significance after accounting for heterogeneity. Given the
direction and small magnitude of the difference in effect size between the fixed- and
random-effects models, this discrepancy in statistical significance is more likely an
artifact of low statistical power than a meaningful indicator about the effect itself; that is,
although the heterogeneity in the data was statistically significant, its impact on effect
size was insufficient to support the statistical power demands of the random-effects
procedure.
35
In summary, these mean effect size estimates suggest that PAI interventions
produced significant reductions in depression symptoms from pre- to post-test, performed
similarly to the other active therapies to which they were compared, and did not
significantly outperform TAU/WL conditions. The significant heterogeneity in all three
estimates, however, suggests the presence of other significant sources of variability (e.g.,
moderators) that need to be considered in a valid interpretation.
Moderation models
Hypothesized moderators were PAI format (group or one-on-one), professional
coadministration of PAI (yes or no), sample type (depression, stressor, or chronic illness),
mutuality (mutual support or one person supporting another), training of peer
administrators (yes or no), professional supervision of peer administrators (yes or no),
PAI content (supportive or psychoeducational), PAI delivery mode (face-to-face,
telephone, or internet), PAI meeting frequency (weekly, more than weekly, or less than
weekly), blindness of assessments (yes or no), researcher allegiance (for or against PAI),
and total methodological quality score (MQS). Each of these constructs was tested as a
possible moderator of SMGs and SMDs, with two exceptions: researcher allegiance, for
which there was insufficient variability to test for moderation (i.e., all but two studies
expressed allegiance to PAI); and blindness of assessments, for which there were too
many missing data to test for moderation (only 56% of controlled studies reported on
blinding).
36
Results of mixed-effects moderation models are displayed in Figures 5-9. Prepost (SMG) models revealed a significant moderation effect of professional
coadministration, such that PAIs that were coadministered by health or mental health
professionals were less effective in reducing depression symptoms between pre- and posttest than PAIs that were entirely peer-administered. There was also a significant omnibus
test for supervision, but this is unlikely to be meaningful since post-hoc tests indicated
that the significant difference was between non-supervised groups and groups in which
supervision status was unspecified. Finally, there was an almost-significant (p = .0546)
omnibus test for sample type, such that pre-post improvements were greater for
depression samples than for stressor samples (with neither differing significantly from
chronic illness samples). There were no other variables that significantly moderated
SMG.
Mean SMD models comparing PAIs with active comparison conditions indicated
a significant moderation effect of PAI content, such that PAIs whose content was
primarily educational performed better than their active comparison groups, whereas
PAIs whose content was primarily supportive did not perform significantly differently
than their active comparison groups. There was also a moderating effect approaching
significance of meeting frequency (p = .0981), such that PAIs meeting weekly performed
better than their active comparison conditions, whereas PAIs with variable or unspecified
meeting frequencies performed no differently than their active comparison conditions.
There were no other significant moderators of SMDs between PAI and active control
conditions.
37
There were no significant moderators of SMDs between PAI and TAU/WL
comparison conditions.
Follow-up effect sizes
Eight studies reported sufficient follow-up data to calculate at least one effect
size. (A ninth study [Tudiver et al., 1992] reported follow-up depression scores, but no
standard deviations or other estimate of variability was available, precluding the
calculation of an effect size.) This resulted in eight follow-up SMGs comparing
outcomes at post-treatment to outcomes at follow-up, and nine follow-up SMDs (five
comparing PAI to active comparison, four comparing PAI to TAU/WL). Follow-up
periods ranged from 3-12 months (M = 6.89, SD = 3.62).
The mean SMG from post-treatment to follow-up was .0016 (p = .99), suggesting
that improvements in depression symptoms were maintained (but did not significantly
increase) after treatment ended. This effect was not moderated either by follow-up length
(p = .9214) or post-treatment SMG (p = .9501).
The mean SMD between PAI and active comparison conditions was .0477 (p =
.6322), indicating that PAI and active comparisons remained equally effective at followup. This effect was not significantly moderated either by follow-up length (p = .3922) or
by post-treatment SMD (p = .9777).
The mean SMD between PAI and TAU/WL conditions was .1226 (p = .0744),
which approached significance and, if significant, would indicate that PAI conditions had
maintained gains better (or deteriorated less) at follow-up compared with TAU/WL
38
conditions. This effect was not significantly moderated by follow-up length (p = .6505),
but it was significantly moderated by SMD at posttest (p = .0012), such that PAIs that
had performed better than TAU/WL comparisons at posttest also performed better at
follow-up.
Outcomes using alternate metrics
Only one of the 26 studies included in this meta-analysis (Bright et al., 1999)
reported an estimate of the percentage of participants who were considered improved
after treatment, precluding the aggregated use of this metric as an indicator of clinical
significance across studies.
Another indicator of clinical significance is whether post-treatment depression
scores fell below the thresholds for clinical elevation for the measures utilized. (Table 2
provides the elevation threshold for each measure included in this analysis.) In theory, at
least, falling below such a threshold would mean that participants improved enough that
their symptom levels at post-treatment were more similar to non-clinical than clinical
population means. Of the 29 PAI conditions for which they were available, posttreatment depression means fell below the elevation threshold for 10 conditions,
indicating that in almost two-thirds of the PAI conditions studied, participants were still
experiencing significant symptoms after receiving treatment. It should be noted that most
of these post-treatment means were well within one standard deviation of their thresholds
(and in many cases, fell mere fractions of points above or below); furthermore, this
indicator is only informative to the extent that variability is low (i.e., that a condition’s
39
mean outcome is a good representation of a typical participant’s outcome in that
condition). Therefore, it should be interpreted cautiously. To the extent that it is
accurate, it suggests that PAIs alone may not have been sufficient – at least in the doses
provided – to reduce symptoms to below a clinical level.
Impact of providing peer support
Only one of the studies included in this meta-analysis (Giese-Davis et al., 2006)
reported on any indicators of well-being for peers providing intervention. In that study,
peer “Navigators” (breast cancer survivors) who provided support to women newly
diagnosed with breast cancer did not show changes in depression symptoms or well-being
over the course of the study (though they did indicate increased dissatisfaction with their
medical teams). The lack of other outcome measures for peer supporters ruled out
aggregate analysis of the impact on peers of providing support.
Publication bias
Visual inspection of the funnel plot for PAI-vs-active therapy effect sizes (Figure
10) suggests a lack, but not a complete absence, of effect sizes on the left side of the
funnel (i.e., negative effect sizes, those that would indicate superiority of other therapy
over PAI). Therefore, the moderate one-tailed selection bias model (Table 4) was the
most likely to fit the data, and indicated that adjusting for publication bias would produce
a mean SMD of -.017 using a fixed-effects model (compared with an unadjusted SMD of
.031), or a mean SMD of -.014 using a random-effects model (compared with an
unadjusted SMD of .083). These adjustments suggest that the true mean SMD may be
40
closer to zero than the observed mean SMD. The observed mean SMD, however, was
not statistically different from zero; thus, the finding does not change when publication
bias is accounted for.
Visual inspection of the funnel plot for PAI vs TAU/WL conditions (Figure 11)
suggests a possible, but not dramatic, lack of effect sizes on the far left side of the funnel
(i.e., negative effect sizes, those that would indicate superiority of TAU/WL over PAI).
Therefore, here too the moderate one-tailed selection bias model (Table 5) was the most
likely to fit the data, and indicated that adjusting for publication bias would produce a
mean SMD of .061 using a fixed-effects model (compared with an unadjusted SMD of
.091), or a mean SMD of .006 using a random-effects model (compared with an
unadjusted SMD of .102). These adjustments suggest that the true mean SMD is closer
to zero than the observed mean SMD (possibly enough closer that it would not be
statistically different from zero).
In summary, it appears that there was a modest degree of publication bias in this
dataset, such that studies showing poorer outcomes for PAIs were less likely to be
published. Adjusting for this bias did not change the interpretation of comparisons
between PAIs and active therapy conditions; the outcomes remained statistically
equivalent. Comparisons between PAI and TAU/WL conditions also remained
equivalent after this adjustment. Therefore, the practical impact of this adjustment is
minimal (see Discussion, pp. 43-44).
41
4. DISCUSSION
Summary and interpretation of findings
The meta-analysis reported here examined the effects of peer-administered
interventions (PAIs) on symptoms of depression, defined as interventions administered
by one or more lay individuals who have experienced problems similar to those being
treated, with minimal or no assistance from a health or mental health professional. This
is the first analysis to comprehensively aggregate outcomes from studies of PAIs for
depression; prior meta-analyses of peer interventions for depression have suffered from
incomplete identification/inclusion of eligible studies and from overly-broad inclusion
criteria (e.g., including interventions that were primarily administered by professionals).
Effect size comparisons indicated that PAIs reduced depression symptoms comparably to
other active interventions (e.g., professionally-administered psychotherapies), but did not
significantly outperform TAU/WL conditions. Overall, these findings suggest that PAIs
may be effective, cost-efficient treatments for depression; their effectiveness, however, is
not unambiguous.
This is also the first meta-analysis to examine potential moderators of PAI
outcomes. Moderation models revealed that PAIs that were co-administered by health or
mental health professionals were less effective in reducing depression symptoms between
pre- and post-treatment than PAIs that were entirely peer-administered. Contrary to the
intuitive perspective that additional expertise should carry additional benefit, this finding
suggests that a “pure” approach to peer intervention may be preferable. It is possible that
the involvement of a professional undermined an important mechanism of peer change;
42
for example, perhaps peers were seen as less knowledgeable when a professional coadministrator was also present, reducing their impact as strong, capable role models of
successful recovery. Given that this moderation effect pertained to pre-post comparisons,
it is also possible that it was an artifact of a third variable unrelated to mechanisms of
therapeutic change, such as researcher allegiance. For example, researchers who
included professionals in their “peer” interventions may have started out with low
confidence in the abilities of peers to be independently therapeutic. Alternately,
researchers who knew that professionals would be involved may have seen less value in
dedicating time and resources toward preparing peers to intervene effectively.
There was also a moderating effect of intervention content, such that PAIs that
were psychoeducational or skills-based (e.g., teaching basic cognitive-behavioral skills)
performed better than other active therapies to which they were compared, whereas PAIs
that were primarily supportive did not perform differently than other active therapies.
There are several plausible interpretations of this finding, including: (1) that gaining
education and skills directly impacted depression symptoms more than receiving support;
(2) that peers administering educational/skills-based interventions were equally
supportive and empathic as those delivering primarily-supportive PAIs, resulting in
additive effects of skills and support; or (3) that peers delivering educational/skills-based
interventions were more likely to have undergone a brief preparatory training program to
prepare them for their roles, which might have resulted in a variety of advantages,
including higher confidence and better interpersonal skills.
43
It is also notable that PAIs performed similarly to other active therapies, yet did
not outperform TAU/WL conditions. Initially, this might seem to suggest that neither
PAIs nor active therapy comparisons were especially helpful. However, examination of
standardized mean gains (pre-post ESs) indicates otherwise: it appears that symptoms
tended to improve substantially between pre- and post-treatment not only for PAIs and
other active therapies, but also for participants in TAU/WL conditions (SMGPAI = .46;
SMGACTCTRL = .55; SMGTAU/WL = .32). This suggests that the failure of PAIs to perform
much better than TAU/WL conditions may be because participants improved regardless
of condition. One possible explanation for the high degree of improvement among
TAU/WL participants is that a substantial minority of included studies involved
depressed chronically ill populations, whose “usual” or “standard” care was likely to
involve regular contact with medical professionals. These contacts may, in some cases,
have been psychologically therapeutic, even if that was not their intended function. A
second explanation pertains to the fact that these analyses are based on study completers;
participants who remained in a study despite being assigned to a control condition may
have been highly motivated to change and/or less distressed to begin with, and thus likely
to improve whether or not they received treatment.
Publication bias appeared to be modestly present in the data, such that betweengroups comparisons that did not favor PAIs were somewhat less likely to appear in the
published literature. When effect sizes were adjusted for the likely degree of publication
bias, PAIs and other active therapies maintained their equivalent effectiveness, and the
difference in outcome between PAI and TAU/WL conditions remained non-significant.
44
Therefore, although it is notable that there was bias in the data, it posed little
complication to the interpretation of the findings.
Clinical significance of findings
The mean pre-post effect size (standardized mean gain) observed across PAI
conditions was statistically greater than zero, but statistical significance does not
necessarily imply clinical significance. What, if anything, does that number say about the
extent to which PAI participants experienced meaningful symptom reduction? There are
multiple ways to address this question.
First, we can calculate how an effect size translates to score reductions on the
symptom measures that were employed. Using the pooled standard deviation observed in
this sample of studies, the .4554 SMG is equivalent to a reduction of 3.54 points on the
BDI-II, or 4.41 points on the CES-D (by far the two most commonly-used measures in
this analysis). To put those numbers in context, we can consider them as minimal
proportions of the threshold for clinical elevation. For the BDI-II, scores ≥14 were
considered elevated; thus, a 3.54-point reduction would constitute a 25.3% minimum
reduction in symptoms. For the CES-D, scores >15 were considered elevated; thus, a
4.41-point reduction would constitute a minimum 29.4% reduction in symptoms. This
provides a conservative estimate of proportional score reductions for PAI participants
(since symptom means often exceeded the threshold considerably). Still, these average
raw score reductions are small in absolute terms – that is, they represent relatively small
changes in actual symptom presentation.
45
Second, we can use as benchmarks the pre-post effect sizes observed in recentlypublished meta-analyses of well-established, empirically-supported therapies for
depression. Large standardized mean gains have been observed in meta-analyses of a
variety of efficacious and possibly-efficacious (Chambless & Hollon, 1998) therapies for
adult unipolar depression, including but not limited to cognitive-behavioral therapy (d =
1.06; Hans & Hiller, 2013), short-term psychodynamic psychotherapy (d = 1.34;
Driessen, Cuijpers, Maat, Abbass, de Jonghe, & Dekker, 2010), and acceptance and
commitment therapy (Hedges’ g = .59; Hofmann, Sawyer, Witt, & Oh, 2010). These
previously-observed standardized mean gains are more than twice those observed in this
sample of PAIs. These comparisons suggest that although PAI participants experience
reduced symptoms, these symptom reductions are inferior to those seen in other, wellestablished psychotherapies for depression.
Methodological limitations
A significant limitation of this analysis was the heterogeneity of the interventions
falling under the PAI umbrella. All had the common feature of being peer-administered,
but they differed substantially in structure, duration, and content. Although every attempt
was made to code these differences and examine their moderating effects, it is possible
that some important moderators were missed or unavailable, or that significant
moderation effects did not achieve significance because of low statistical power.
Furthermore, the lack of clear distinction between TAU and WL conditions, and the
46
resulting need to combine them as a single comparison category are less than ideal, as
there was substantial variety in these conditions.
Another limitation was the lack of available data on potential moderators and
mediators of PAI outcomes. Although methodological features were coded and tested as
moderators, participant-level variables beyond basic demographics were largely
unavailable, let alone measured consistently enough across studies to test metaanalytically. For example, race and ethnicity were too inconsistently reported in these
studies to be tested as moderators, yet it is probable that racial and cultural factors impact
clients’ views on help-seeking generally, and their attitudes toward utilizing peers versus
professionals to manage distress (e.g., Diala, Muntaner, Walrath, Nickerson, LaVeist, &
Leaf, 2001; Sheu & Sedlacek, 2004; Townes, Chavez-Korell, & Cunningham, 2009;
Chen & Mak, 2008). Mechanisms of action underlying PAIs’ effectiveness were, for the
most part, neither tested nor speculated upon in this literature. One obvious example is
empathy: the capacity of peers to empathize with similar others’ distress is a
hypothesized process by which PAIs might be effective, yet too few studies measured
empathy or any other relationship or process-level variables to examine this hypothesis.
Thus, the currently-available literature largely leaves us to speculate about how, why, and
for whom PAIs are beneficial.
Future directions
Given the limitations details above, the literature on PAIs for depression would
benefit from future research examining for whom, under what circumstances, and via
47
what processes PAIs are helpful for depression. For example, in this analysis, there were
not statistically significant pre-post or between-groups differences in outcome for the
different sample types (depressed, stressed, or chronically ill), but there was a consistent
pattern of smaller effect sizes for stressor samples that would be worth investigating
further with a larger sample size. Furthermore, from a practical standpoint, it would be
useful to know whether PAIs function best as standalone interventions or as adjuncts to
traditional treatments (and, if as adjuncts, whether best delivered concomitantly or as
relapse prevention). Finally, in addition to empathy, other likely mediating processes
might include recovery modeling, the provision of hope, and specific techniques or
coping strategies.
Another distinction remains to be addressed. PAIs may be (1) treatments that are
already supported as professionally-administered interventions, but now administered by
peers; or (2) interventions that are specifically designed for peers to administer, without
prior empirical support as professionally-administered interventions. The former may be
more justifiable in terms of research funding; at the same time, interventions designed
specifically for peers may have therapeutic qualities that traditional interventions miss.
This distinction is worthy of attention in future research.
Another area of interest for future research is the effect on peers of delivering a
PAI. Ideally, helping others with problems one has personally overcome would reinforce
recovery or at least provide a subjectively positive experience. There is some evidence
that peer counselors have experienced increased well-being, increased confidence, and
decreases in mild depression symptoms (e.g., Garcia, Metha, Perfect, & McWhirter,
48
1997; de Vries & Petty, 1992; Byrd, 1984), but this evidence is meager and
methodologically limited. Moreover, we know little about how individual differences
among peer supporters might influence these outcomes. For example, for peers who are
currently depressed or at higher risk for relapse, spending time with other depressed
individuals and exerting energy in a helping capacity could be emotionally draining,
increase the probability of relapse, or prolong symptoms. On the other hand, for those
with solid, durable recovery experiences, supporting peers may provide gratification,
connectedness, a sense of purpose, or other positive emotional states that could buffer
against relapse and improve quality of life.
49
FIGURES
Figure 1: Flow of studies
Search terms:
“(Peer OR mutual OR paraprofessional) AND depress*”
(not unique)
PsycINFO:
3344 results
Medline:
1491 results
112 potentiallyeligible studies for
review from
database searches
4723 results excluded
because non-eligibility
was clear from
abstract (e.g., not an
empirical study or not
an intervention study)
+
33 potentiallyeligible studies for
review from
hand-search of
reference sections
145 potentiallyeligible studies
26 studies met
inclusion criteria
20*
controlled
111 studies excluded
because participants
or intervention did
not meet inclusion
criteria
8 studies excluded
because insufficient
data were available
6
uncontrolled
*Includes two non-randomized controlled studies, which
were not used to calculate between-groups differences.
50
Figure 2: Forest plot of effect sizes comparing PAI to active therapy conditions
51
Figure 3: Forest plot of effect sizes comparing PAI to TAU/WL conditions
52
Figure 4: Forest plot of effect sizes of PAI conditions (change from pre- to post-treatment)
53
Figure 5: Moderators of standardized mean gain (SMG) for PAI conditions (change from pre- to post-treatment)
Darker bars denote effect sizes that are significantly different from zero.
** p<.01
* p<.05
† p<.1
54
Figure 6: Moderators of standardized mean difference (SMD) comparing PAI to active therapy conditions
Darker bars denote effect sizes that are significantly different from zero.
** p<.01
* p<.05
† p<.1
55
Figure 7: Moderators of standardized mean difference (SMD) comparing PAI to TAU/WL conditions
Darker bars denote effect sizes that are significantly different from zero.
** p<.01
* p<.05
† p<.1
56
Figure 8: Methodological quality is not significantly associated with pre-post effect size
57
Figure 9a: Methodological quality is not significantly associated with between-groups
effect size (PAI compared with active therapy conditions)
58
Figure 9b: Methodological quality is not significantly associated with between-groups
effect size (PAI compared with TAU/WL)
59
Figure 10: Funnel plot of effect sizes comparing PAI to active therapy conditions
60
Figure 11: Funnel plot of effect sizes comparing PAI to TAU/WL conditions
61
TABLES
Table 1: Descriptive characteristics of clinical trials involving PAIs
1st Author,
Year
Setting/ format
Inclusion criteria
Depression
measure*
PAI(s)
Control
condition(s)
Duration of
intervention
Andersson,
2005
Internet/ group
BDI
Moderated online
discussion group;
therapist-monitored
Online self-help
CBT program +
moderated online
discussion group
M = 10
weeks
Bright,
1999
Outpatient
psychology
dept clinic/
group
facilitated by
prior
participants in a
communitybased self-help
org
Community
mutual aid
organization/
group
Swedish adults
with likely mildto-moderate
unipolar
depression (per
CIDI, and
MADRS-S score
15-30)
Memphis area
adults scoring ≥10
on HRSD and
meeting SCID-NP
criteria for MDE,
dysthymia, or dep
NOS
HRSD
(1) Mutual support
group led by
paraprofessional
(1) CBT group led
by professional
Hong Kong adults
who were still
symptomatic after
a 12-week group
CBT program
CES-D:
Chinese
version
Mutual-aid groups
of about 8
members each
Immediately
postpartum
Ontario women
scoring >9 on
EPDS
EPDS
Usual postpartum
care plus peer
telephone support
Cheung,
2000
Dennis,
2009
Ontario public
health
departments
(hospitals)/ 1:1
telephone with
[BDI]
(2) CBT group led
by paraprofessional
[SCID]
Follow up?
(months
post tx end)
6 months
MQS
10 weekly
90-minute
sessions
None
reported
5
Monthly
meetings over
12 months
(mean #
attended not
reported)
Minimum of
4 contacts; M
= 8.8 contacts
over 12
weeks
None
reported
2
3 months
(note that
29% had
continued
contact
10
8
(2) Mutual support
group led by
paraprofessional
Usual postpartum
care only
62
GieseDavis,
2006
peer w/selfreported hx of
and recovery
from PPD
Not specified/
1:1 telephone
(primarily) with
peer
“Navigator”
survivor of
several years
Not specified/
group
between
posttx and
FU)
California adult
women with
recent b.c. dx (M
= 2.23 mo postdx)
CES-D
Supportive and
educational peer
counseling
San Francisco
area adults
receiving care for
symptomatic
HIV/AIDS
CES-D
National c.b.
organization/
1:1 telephone
with peer w/ hx
of b.c.
recurrence
Not specified/
1:1 telephone
with another
participant
Adults
experiencing a
first recurrence of
b.c. (dx’ed in last
56 days)
CES-D
Group (10-15
participants)
educational
intervention taught
by teams of pair
leaders (1 with
HIV/AIDS)
Peer-delivered
telephone support
Indiana area, lowsupport,
community living,
elderly women
CES-D
Hill,
2006
Internet/ group
forum
Western US adult
rural-living
women with
chronic illness
CES-D
Letourneau,
2011
Home-based
and university/
1:1 in-person
Canadian adult
postpartum
women caring for
EPDS
Gifford,
1998
Gotay,
2007
Heller,
1991
3-6 months of
contact 14x/week (M
= 4 months
None
reported
3
Usual medical care
7 weekly
sessions / 16
total hours
None
reported
7
Usual medical care
3 months/ 4-8
sessions
None
reported
7
(1) Peer contact
(dyad initiator)
(1) Supportive staff
contact
20-30 weeks/
1-2 contacts
per week
None
reported
5
(2) Peer contact
(dyad receiver)
Online,
asynchronous,
peer-led support
group and
educational units
Peer support and
maternal infant
interaction program
(2) Assessment
only
No treatment
22 weeks
None
reported
4
Waiting list
(received 2 weeks
of peer support
12 weeks/ M
= 9 visits
and/or phone
None
reported
7
63
Lieberman,
2005
and telephone
with peer
w/self-reported
hx of and
recovery from
PPD
Internet/ group
Lieberman,
2006
Internet/ group
Lorig,
2009
Community
settings/ group
led by peers
with type-2
diabetes
Outpatient
medical clinic/
group led by
peers with
depression
experience
Ludman,
2007
Marmar,
1988
Not specified/
group led by
peer women
who had lost
husbands
several years
an infant <9
months old,
scoring >12 on
EPDS
after a 12-week
waiting period)
New members to
online selfdirected breast
cancer support
forums
New members to
online selfdirected breast
cancer support
forums
San Francisco
area adults with
type-2 diabetes
CES-D
Washington state
HMO adult
patients with
persistent
depression (SCL90 >.75) after ≥6
months antidep
medication
SCL-90:
depression
subscale
Adult women
whose husbands
died 4-36 months
previously
BDI short
form
CES-D
PHQ-9
Participation in
active peer-support
breast cancer
“bulletin board”
forum
Participation in
active peer-support
breast cancer
“bulletin board”
forum
Educational peerled diabetes selfmanagement
program
Telephone care
management +
peer-led group
(1) Telephone care
management
(2) Usual medical
care
[SCID]
[SCL-90]
Usual medical care
(waitlist for
program)
Supportive mutualhelp group
treatment
(3) Telephone care
management +
professionally-led
group
Professionallyadministered brief
dynamic
psychotherapy
calls
6 months/
variable (selfdirected)
participation
frequency
6 months/
variable (selfdirected)
participation
frequency
6 weekly 2.5hour sessions
None
reported
3
None
reported
3
6 months
7
6 weekly
workshops
followed by
ongoing bimonthly
groups
9 months
8
12 weekly
1.5-hour
sessions
12 months
9
64
McKay,
2002
Nagel,
1988
Onrust,
2010
Preyde,
2003
RobinsonWhelan,
2007
before
Internet
Nursing home/
1:1 in person
(half with
seniors and half
with
adolescents)
Home-based;
1:1 in person
with peer
supporter who
were widowed
themselves for
several years
Hospital NICU/
1:1 in person
with peers who
had previously
experienced
having a
preterm infant
Centers for
independent
living/ group
led by peer
women with
disabilities
Adults diagnosed
with type-2
diabetes
CES-D
(1) Educational
material + online
peer support forum
(2) Educational
material + online
peer support forum
+ professional selfmanagement coach
(1) Empathyfocused peer
counseling
(1) Educational
material only
10 months/
variable (selfdirected)
participation
frequency
None
reported
6
No treatment
5 weeks/ two
1-hour
meetings per
week
None
reported
6
Brief informational
brochure on
depression
symptoms
M = 8.3
home visits
6-9 months
9
(2) Professional
self-management
coach only
Elderly nursing
home residents
with mild to
moderate
depression
symptoms (Zung
score 40-61)
Netherlands
adults age ≥55
whose spouses
died during the
past year
Zung SDS
Adult women
immediately
postpartum with
very preterm
infants less than
10 days old
BDI short
form
“Parent buddy”
peer support
program
Standard medical
care
16 weeks/ M
= 9 contacts
None
reported
7
Adult rural-living
women with a
physical disability
and mild to
moderate
depression
BDI-II
Peer-led depression
self-management
program
Usual medical care
8 weekly 2.5hour sessions
3 months
8
CES-D
[CES-D]
(2) Educationfocused peer
counseling
Peer support home
visiting service
65
Roman,
1995
Regional
perinatal center/
1:1 in person
and phone with
peers who had a
preterm infant
≥1 year before
Silver,
1997
Urban medical
center/ 1:1 in
person and
phone with peer
women who
had raised
children with
health problems
HIV primary
care outpatient
clinic/ group in
person and 1:1
phone calls with
peer supporters
(HIV+ and on
HAART)
Outpatient
mental health
clinics/ 1:1
phone with
another
participant
Community
facility/ group
led by peer
(widowed)
facilitator
Simoni,
2007
Travis,
2010
Tudiver,
1992
symptoms (BDI-II
14-29)
Adult women
immediately
postpartum, no
prior NICU
experience, with
preterm infants
less than 6 days
old
Adult mothers of
5-8 y.o. children
with ongoing
health conditions
POMS:
depression
subscale
Veteran parent peer
support program
No treatment
12 months/
variable
contact (M =
34 hrs)
None
reported
6
PSI:
depression
subscale
Supportive peer
contact program
No treatment
12 months/ 6
face-to-face
meetings and
at least
biweekly
phone contact
None
reported
8
New York area
adult HIVpositive patients
on HAART
regimen
CES-D
Peer support
intervention
targeting
medication
adherence and
depression
symptomology
Standard medical
care
3 months/ 6
bimonthly
group
meetings, and
phone calls 13x/week
3 months
8
Michigan adults
with depression
symptoms (BDI-II
>12) and hx of at
least 2 antidep
med trials
Adult men
recently widowed
(3-12 months
prior)
BDI-II
Telephone-based
mutual peer
support program
None
reported
2
BDI short
form
Peer-facilitated
mutual help groups
12 weeks/ at
least 1
weekly call
(M = 10.3
completed
calls)
9 weekly 1.5hour sessions
8 months
(not
included in
analyses b/c
of
7
[GHQ]
Waiting list
(offered tx after 8
months)
66
Uccelli,
2004
Not specified/
group led by
peer (MS
patient
nominated by
neurologist)
Italian adults
diagnosed with
MS
BDI: MS
adaptation
Peer support group
intervention
8 weekly
sessions
insufficient
data
reported)
None
reported
4
67
Table 2: Characteristics of depression symptom measurement instruments
Instrument
Beck Depression
Inventory (BDI)
Beck Depression
Inventory-II (BDI-II)
Hamilton Rating Scale for
Depression (HRSD)
Center for
Epidemiological StudiesDepression (CES-D)
Edinburgh Postnatal
Depression Scale (EPDS)
Patient Health
Questionnaire-9 (PHQ-9)
Symptom Check List-90
(SCL-90)2
Zung Self-Rating
Depression Scale
Beck Depression
Inventory-Short Form
Profile of Mood States
(POMS)2
Psychiatric Symptom
Index (PSI)2
1
Author
Beck et al., 1961
Beck et al., 1996
Hamilton, 1960
Radloff, 1977
Cox et al., 1987
Kroenke et al., 2001
Derogatis et al., 1977
Zung, 1965
Beck et al., 1993
McNair et al., 1981
Ilfeld, 1976
N1
Description
21-item self-report; scores
range 0-63
21-item self-report; scores
range 0-63
Interviewer-administered;
scores range 0-52
20-item self-report; scores
range 0-60
2
Score threshold
for elevation
≥ 10
2
≥ 14
1
≥8
11
> 15
10-item self-report; scores
range 0-30
9-item self-report; scores
range 0-27
20-item self-report; scores
range 0-60
20-item self-report; scores
range 20-80
13-item self-report; scores
range 0-39
15-item self-report; scores
range 0-60
10-item self-report; scores
range 0-100 (standardized)
2
≥9
1
≥5
1
≥ 253
1
≥ 40
3
≥5
1
≥7
1
≥ 20
Number of studies included in meta-analysis for which this was considered the primary depression measure
Depression subscale score
3
Represents summed item scores; sometimes reported as mean score per item (as in Ludman et al., 2007)
2
68
Table 3: Results of included studies
Study
Andersson,
2005
Defining
condition of
sample
Depression
%
female
Mean
age
74
36.1
Bright,
1999
Depression
Cheung,
2000
Dennis,
2009
Depression
83
38
Depression
(postnatal)
100
Giese-Davis,
2006
Chronic
illness
(breast
cancer)
Chronic
illness
(HIV/AIDS)
Chronic
illness
(breast
cancer)
Depression
100
78%
ages
20-34
51.31
Gifford,
1998
Gotay,
2007
Heller,
1991
Hill,
2006
Chronic
illness
71
0
100
100
100
Group
Enrolled
N
Completed
Follow-up
Depression mean (SD)
Pre-treatment Post-treatment Follow-up
PAI
CG
60
57
49
36
35
36
20.9 (8.5)
20.5 (6.7)
19.5 (8.1)
12.2 (6.8)
PAI(1)
PAI(2)
CG(1)
CG(2)
22
21
27
28
14
13
18
22
18.01 (4.7)
17.86 (5.18)
17.11 (3.59)
17.67 (4.70)
6.07 (2.65)
6.85 (3.71)
8.17 (6.41)
8.50 (6.39)
PAI
83
65
21.54 (12.28)
23.57 (13.16)
PAI
CG
349
352
297
316
12.5 (2.8)
12.62 (2.76)
7.93 (4.68)
8.89 (5.24)
PAI
29
19
21.75 (12.51)
Not given
PAI
CG
34
37
25
33
17.3 (8.8)
19.6 (11.3)
18.7 (10.5)
20.0 (11.6)
PAI
CG
152
152
128
124
17.72 (10.71)
15.62 (10.12)
15.39 (9.69)
14.14 (9.81)
PAI(1)
PAI(2)
CG(1)
CG(2)
49
49
44
53
49
46
44
53
31.1 (8.4)
31.6 (6.8)
32.0 (9.2)
32.8 (8.3)
31.3 (10.1)
30.5 (7.6)
31.2 (9.3)
30.4 (7.1)
PAI
61
43
18.52 (11.64)
15.50 (11.90)
13.1 (7.6)
13.1 (9.1)
45.8
289
311
45.26
54
74
92%
older
7.00 (4.66)
7.61 (4.59)
69
Letourneau,
2011
Lieberman,
2005
Lieberman,
2006
Lorig,
2009
Ludman,
2007
(mixed, e.g.,
diabetes,
MS, heart
disease,
cancer)
Depression
(postnatal)
Chronic
illness
(breast
cancer)
Chronic
illness
(breast
cancer)
Chronic
illness (type2 diabetes)
Depression
than
42
100
100
100
66
71
Major
ity
26-35
46.2
18.13 (10.34)
16.91 (11.62)
PAI
CG
27
33
23
28
18.20 (4.76)
16.1 (4.67)
11.80 (4.68)
8.68 (5.44)
PAI
114
91
20.0 (.85)
15.3 (1.2)
PAI
74
52
19.90 (10.01)
14.60 (10.08)
PAI
CG
186
159
161
133
102
n/a
6.35 (5.85)
6.46 (5.60)
4.84 (4.33)
7.00 (4.49)
3.42 (5.03)
n/a
PAI(1)
CG(1)
CG(2)
CG(3)
26
26
26
26
22
23
23
26
21
20
23
25
1.63 (.68)
1.61 (.50)
1.66 (.54)
1.72 (.56)
1.22 (.54)
1.65 (.68)
1.21 (.58)
1.44 (.66)
1.24 (.95)
1.19 (.68)
1.19 (.65)
1.14 (.70)
PAI
CG
30
31
26
29
30
31
8.80 (4.66)
8.65 (4.69)
8.77 (5.72)
6.77 (5.82)
7.40 (6.59)
5.35 (4.88)
PAI(1)
PAI(2)
CG(1)
CG(2)
40
40
40
40
30
33
33
37
15.83 (12.79)
16.61 (11.67)
15.27 (12.25)
18.11 (11.79)
13.86 (12.62)
14.87 (9.69)
15.31 (9.15)
16.42 (12.98)
PAI(1)
PAI(2)
CG
20
20
20
20
20
20
45.60 (4.77)
46.05 (3.65)
49.60 (6.97)
39.80 (6.62)
42.40 (4.26)
48.50 (6.20)
50.2
Nagel,
1988
Depression
Not
available
74
Onrust,
Stressor
64
68.9
47
57
66.7
100
McKay,
2002
59
45.5
Stressor
(spousal
bereavemt)
Chronic
illness (type2 diabetes)
Marmar,
1988
CG
58
59
70
2010
Preyde,
2003
RobinsonWhelan,
2007
Roman,
1995
Silver,
1997
Simoni,
2007
Travis,
2010
Tudiver,
1992
Uccelli,
2004
(spousal
bereavemt)
Stressor
(preterm
infant)
Depression
+ Chronic
Illness
Stressor
(preterm
infant)
Stressor
(child’s
ongoing
health
problem)
Chronic
illness
(HIV)
Depression
Stressor
(spousal
bereavemt)
Chronic
illness
(multiple
sclerosis)
100
100
100
100
44
110
106
93
95
PAI
CG
32
28
24
25
PAI
CG
69
65
42
54
PAI
CG
27
31
PAI
CG
91
94
17.1 (9.4)
16.2 (9.1)
15.1 (8.1)
14.7 (8.7)
4.53 (3.81)
5.57 (3.28)
2.20 (4.00)
4.88 (3.77)
25.2 (11.3)
29.07 (11.30)
19.93 (11.3)
27.50 (11.30)
22
25
11.7 (11.4)
14.25 (10.35)
9.00 (11.25)
7.78 (8.25)
183
182
174
169
23.8 (17.5)
20.8 (16.6)
22.1 (17.5)
19.6 (16.9)
PAI
CG
71
65
54
51
19.9 (12.4)
19.6 (11.2)
17.6 (11.6)
16.4 (11.5)
PAI
54
32
26.0 (10.1)
22.1 (10.7)
PAI
CG
61
52
57
37
7.3
8.5
6.2
5.4
PAI
44
42
27.19 (7.83)
26.24 (6.80)
13.1 (7.3)
13.5 (7.9)
30
51.59
42
54
17.58 (11.30)
24.88 (11.30)
26
34.4
42.6
37
52.4
0
63
60
PAI
CG
>18,
no
mean
given
59
57
57
37
21.3 (14.5)
17.9 (11.2)
5.6
3.9
71
Table 4: Mean SMD estimates (PAI compared with active therapy conditions) adjusted
for four models of selection (publication) bias
FIXED EFFECTS
Parameter
Standard
estimate
error
Unadjusted ES
(SMD)
0.03141657
Moderate onetailed selection
Severe one-tailed
selection
Moderate twotailed selection
Severe two-tailed
selection
-0.01738422
RANDOM EFFECTS
Parameter
V (estimate of)
estimate
population effectsize variance
0.08347356
0.09470620
(SE =
(SE =
0.11564934)
0.06408005)
-0.01431520
0.09691550
-0.09302688
-0.83928611
0.97821510
0.02677201
0.07177643
0.09400445
0.02142850
0.05942731
0.09353144
0.07084147
72
Table 5: Mean SMD estimates (PAI compared with TAU/WL conditions) adjusted for
four models of selection (publication) bias
FIXED EFFECTS
Parameter
Standard error
estimate
Unadjusted ES
(SMD)
0.09130349
Moderate onetailed selection
Severe one-tailed
selection
Moderate twotailed selection
Severe two-tailed
selection
0.06115963
RANDOM EFFECTS
Parameter
V (estimate of)
estimate
population
effect-size
variance
0.1026520 (SE 0.14413175
= 0.1060054)
(SE =
0.07496142)
0.006533058
0.1516806
0.02258148
-1.018147072
1.5018998
0.07947753
0.095101156
0.1438070
0.06487532
0.086774433
0.1436267
0.04030200
73
APPENDIX A: CODING FORM
First Author:
Title:
Year of Publication:
1. Control group? YES NO
If yes:
a. Randomized? YES NO NOT SPECIFIED
b. Assessments blinded? YES NO NOT SPECIFIED
c. Number of control groups:
d. Type of CG1: OTHER THERAPY
TAU
MED
WAITLIST
i. If other therapy, specify type:
e. Type of CG2: OTHER THERAPY
TAU
MED
WAITLIST
i. If other therapy, specify type:
f. Researcher allegiance: PAI was presented as:
i. _____ An active treatment (i.e., a treatment of interest)
ii. _____ A control condition
2. Sample type:
OTHER
DEPRESSION
STRESSOR
CHRONIC ILLNESS
3. Intended duration of PAI treatment: ______ weeks and/or ______ contacts
4. Observed average duration of PAI treatment: ______ weeks and/or ____
contacts
5. Frequency of PAI treatment: WEEKLY MORE THAN WEEKLY
LESS THAN WEEKLY
VARIABLE/UNCLEAR
6. Format of PAI treatment:
GROUP
7. Mode of delivery of PAI treatment:
INTERNET
8. Support is:
ANOTHER
MUTUAL
ONE-ON-ONE
FACE-TO-FACE
PHONE
DELIVERED FROM ONE PERSON TO
74
a. If one person to another, any outcomes for supporters reported?
YES NO
9. Peer supporters were selected by: THEIR OWN RESPONSE TO AN AD
RECRUITED AS FORMER PATIENTS
OTHER
NOT SPECIFIED
10. Pre-intervention training of peer supporters:
SPECIFIED
SOME
NONE
NOT
11. Did peer supporters have ongoing contact with a professional supervisor over
the course of the intervention?
YES
NO NOT SPECIFIED
12. Was the PAI treatment co-administered by a health or mental health
professional, or entirely peer-administered?
PROFESSIONAL COADMINISTERED
ENTIRELY PEER-ADMINISTERED
13. Content of PAI treatment is described as primarily:
SUPPORTIVE
EDUCATIONAL
NOT SPECIFIED
14. Total N screened for study: _____
15. Total N enrolled in study (providing pretest data): _____
a. N enrolled in PAI: _____
b. N enrolled in CG 1:______; CG 2: ______
16. N completed study (providing posttest data): _____ (specify whether N or %)
a. N completed PAI: _____
b. N completed CG1:_____; N completed CG2: ______
17. Gender breakdown of total sample: _____% female, _____% male
a. Gender breakdown of PAI group: _____% female, _____% male
b. Gender breakdown of CG1: _____% female, _____% male
c. Gender breakdown of CG2: _____% female, _____% male
18. Mean age of total sample: _____ years
a. Mean age of PAI group: _____ years
b. Mean age of CG1: _____ years
75
c. Mean age of CG2: _____ years
19. Instrument(s) used to measure depression (check all that were used)*:
a. _____ HRSD (a.k.a., HAM-D)
b. _____ BDI or BDI-II
c. _____ CES-D
d. _____ Other: ______________________________________
*If more than one measure was used, write “primary” or “secondary” next to each one checked. HRSD is
always primary if it was used, followed in order of priority by BDI, CES-D, EPDS, HADS, and finally any
other measure.
20. PRETEST depression score of PAI group:
a. Primary measure name: ______ Mean: ______
SD:______
b. Secondary measure name: ______ Mean: ______
SD:______
21. POSTTEST depression score of PAI group:
a. Primary measure name: ______ Mean: ______
SD:______
b. Secondary measure name: ______ Mean: ______
SD:______
22. PRETEST depression score of CG1:
a. Primary measure name: ______ Mean: ______
SD:______
b. Secondary measure name: ______ Mean: ______
SD:______
23. POSTTEST depression score of CG1:
a. Primary measure name: ______ Mean: ______
SD:______
b. Secondary measure name: ______ Mean: ______
SD:______
24. PRETEST depression score of CG2:
a. Primary measure name: ______ Mean: ______
SD:______
b. Secondary measure name: ______ Mean: ______
SD:______
25. POSTTEST depression score of CG2:
a. Primary measure name: ______ Mean: ______
SD:______
b. Secondary measure name: ______ Mean: ______
SD:______
26. If reported, % considered “improved” at posttest:
PAI: ______
CG1: ______
CG2: ______
76
27. Are any follow-up data reported (including in a separate manuscript)?
YES NO
If yes:
a. Number of months post-treatment: ______
b. Was help-seeking behavior during the follow-up period reported?
YES NO
i. If Yes, % who sought help: PAI: ____ CG1: ____ CG2:
____
c. Was % remaining improved at follow-up reported? YES NO
i. If Yes, % remaining improved: PAI: ____ CG1: ____ CG2:
____
d. FOLLOW-UP depression score of PAI group:
i. Primary measure mean: ______
SD:______
ii. Secondary measure mean: ______
SD: ______
e. FOLLOW-UP depression score of CG1:
i. Primary measure mean: ______
SD:______
ii. Secondary measure mean: ______
SD: ______
f. FOLLOW-UP depression score of CG2:
i. Primary measure mean: ______
SD:______
ii. Secondary measure mean: ______
SD: ______
77
APPENDIX B: METHODOLOGICAL QUALITY SCORE (MQS) CODING SYSTEM
(adapted from Miller et al., 1995)
Methodological
feature
Group allocation
Quality control
Immediate
posttreatment
assessment rate
Additional
posttreatment followups
Dropouts
Blinding
Analyses
Multisite
Points available
2 = randomized comparison group
1 = nonrandomized comparison group
0 = no comparison group
1 = treatment standardized by manual,
specific training, etc.
0 = no standardization specified
2 = 85-100% of post assessments completed
1 = 70-84.9% of post assessments completed
0 = <70% of post assessments completed, or
not specified
1 = at least one additional post-treatment
follow-up completed
0 = no additional post-treatment follow-up,
or not specified
1 = treatment dropouts included in at least
some outcome data (e.g., intent-to-treat
analysis or compared on dependent variable)
0 = treatment dropouts not discussed or
accounted for
1 = posttreatment assessment done by
blinded interviewer
0 = posttreatment assessment nonblinded, or
unspecified
1 = acceptable statistical analyses
0 = no statistical analyses, inappropriate
analyses, or unspecified
1 = parallel replications at two or more sites
with separate research teams
0 = single site or comparison of sites
offering different treatments
SUM QUALITY SCORE:
Points
awarded
78
REFERENCES
Asterisked entries denote studies that were included in the meta-analysis.
*Andersson, G., Bergström, J., Holländere, F., Carlbring, P., Kaldo, V., & Ekselius, L.
(2005). Internet-based self-help for depression: Randomised controlled trial.
The British Journal of Psychiatry, 187, 456-461. doi:10.1192/bjp.187.5.456
Andersson, G., & Cuijpers, P. (2009). Internet-based and other computerized
psychological treatments for adult depression: A meta-analysis. Cognitive
Behaviour Therapy, 38, 196-205. doi:10.1080/16506070903318960
Batson, C.D., Sympson, S.C., Hindman, J.L., Decruz, P., Todd, R.M., Weeks, J.L.,
Jennings, G., & Burns, C.T. (1996). “I’ve been there, too”: Effect on empathy of
prior experience with a need. Personality and Social Psychology Bulletin, 22,
474-482. doi: 10.1177/0146167296225005
Beck, A.T., & Steer, R.A. (1993). Beck Depression Inventory Manual. San Antonio,
TX: Psychological Corporation.
Beck, A.T., Steer, R.A., & Brown, G.K. (1996). Manual for the Beck Depression
Inventory-II. San Antonio, TX: Psychological Corporation.
Beck, A.T., Ward, C.H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory
for measuring depression. Archives of General Psychiatry, 4, 561-571.
doi:10.1001/archpsyc.1961.01710120031004
Berman, J.S., & Norton, N.C. (1985). Does professional training make a therapist more
effective? Psychological Bulletin, 98, 401-407. doi:10.1037/0033-2909.98.2.401
Bickman, L. (1999). Practice makes perfect and other myths about mental health
services. American Psychologist, 54, 965-978. doi:10.1037/h0088206
*Bright, J.I., Baker, K.D., & Neimeyer, R.A. (1999). Professional and paraprofessional
group treatments for depression: A comparison of cognitive-behavioral and
mutual support interventions. Journal of Consulting and Clinical Psychology, 67,
491-501. doi:10.1037/0022-006X.67.4.491
Burns, D.D., & Nolen-Hoeksema, S. (1992). Therapeutic empathy and recovery from
depression in cognitive-behavioral therapy: A structural equation model. Journal
of Consulting and Clinical Psychology, 60, 441-449. doi: 10.1037/0022006X.60.3.441
79
Byrd, M. (1984). Personal growth aspects of peer counselor training for older adults.
Educational Gerontology, 10, 369-385. doi:10.1080/0380127840100503
Chambless, D.L., & Hollon, S.D. (1998). Defining empirically supported therapies.
Journal of Consulting and Clinical Psychology, 66, 7-18. doi:10.1037/0022006X.66.1.7
Chen, S.X., & Mak, W.W.S. (2008). Seeking professional help: Etiology beliefs about
mental illness across cultures. Journal of Counseling Psychology, 55, 442-450.
doi: 10.1037/a0012898
*Cheung, S., & Sun, S.Y.K. (2000). Effects of self-efficacy and social support on the
mental health conditions of mutual-aid organization members. Social Behavior
and Personality, 28, 413-422. doi:10.2224/sbp.2000.28.5.413
Chinman, M.J., Rosenheck, R., Lam, J.A., & Davidson, L. (2000). Comparing consumer
and nonconsumer provided case management services for homeless persons with
serious mental illness. The Journal of Nervous and Mental Disease, 188, 446453. doi:10.1097/00005053-200007000-00009
Christensen, A., & Jacobson, N.S. (1994). Who (or what) can do psychotherapy: The
status and challenge of nonprofessional therapies. Psychological Science, 5, 814. doi:10.1111/j.1467-9280.1994.tb00606.x
Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and
Psychological Measurement, 20, 37-46. doi:10.1177/001316446002000104
Cox J.L., Holden J.M., & Sagovsky R. (1987) Detection of postnatal depression:
Development of the 10-item Edinburgh Postnatal Depression Scale. British
Journal of Psychiatry, 150, 782-786. doi:10.1192/bjp.150.6.782
De Vries, B., & Petty, B.J. (1992). Peer-counseling training: Analysis of personal growth
for older adults. Educational Gerontology, 18, 381-393.
doi:10.1080/0360127920180406
Den Boer, P.C.A.M., Wiersma, D., Russo, S., & van den Bosch, R.J. (2005).
Paraprofessionals for anxiety and depressive disorders. Cochrane Database of
Systematic Reviews, 2, CD004688. doi:10.1002/14651858.CD004688
*Dennis, C., Hodnett, E., Kenton, L., Weston, J., Zupancic, J., Stewart, D.E., & Kiss, A.
(2009). Effect of peer support on prevention of postnatal depression among high
risk women: Multisite randomized controlled trial. BMJ, 338, a3064.
doi:10.1136/bmj.a3064
80
Derogatis, L.R., & Cleary, P.A. (1977). Confirmation of the dimensional structure of the
SCL-90: A study in construct validation. Journal of Clinical Psychology, 33,
981-989. doi:10.1002/1097-4679(197710)33:4<981::AIDJCLP2270330412>3.0.CO;2-0
Diala, C.C., Muntaner, C., Walrath, C., Nickerson, K., LaVeist, T., & Leaf, P. (2001).
Racial/ethnic differences in attitudes toward seeking professional mental health
services. American Journal of Public Health, 91, 805-807.
doi:10.2105/AJPH.91.5.805
Doughty, C., & Tse, S. (2011). Can consumer-led mental health services be equally
effective? An integrative review of CLMH services in high-income countries.
Community Mental Health Journal, 47, 252-266. doi:10.1007/s10597-010-93215
Driessen, E., Cuijpers, P., de Maat, S.C.M., Abbass, A.A., de Jonghe, F., & Dekker,
J.J.M. (2010). The efficacy of short-term psychodynamic psychotherapy for
depression: A meta-analysis. Clinical Psychology Review, 30, 25-36.
doi:10.1016/j.cpr.2009.08.010
Durlak, J.A. (1979). Comparative effectiveness of paraprofessional and professional
helpers. Psychological Bulletin, 86, 80-92. doi:10.1037/0033-2909.86.1.80
Duval, S., & Tweedie, R. (2000). Trim and fill: A simple funnel plot-based method of
testing and adjusting for publication bias in meta-analysis. Biometrics, 56, 455463. doi:10.1111/j.0006-341X.2000.00455.x
Eklund, J., Andersson-Straburg, T., & Hansen, E.M. (2009). “I’ve also experienced loss
and fear”: Effects of prior similar experience on empathy. Scandinavian Journal
of Psychology, 50, 65-69. doi:10.1111/j.1467-9450.2008.00673.x
Field, A.P., & Gillett, R. (2010). How to do a meta-analysis. British Journal of
Mathematical & Statistical Psychology, 63, 665-694. Files accessed 12/1/12 at
http://www.statisticshell.com/meta_analysis/how_to_do_a_meta_analysis.html.
Garcia, Y.E., Metha, A., Perfect, M.C., & McWhirter, J.J. (1997). A senior peer
counseling program: Evaluation of training and benefits to counselors.
Educational Gerontology, 23, 329-344. doi:10.1080/0360127970230403
*Giese-Davis, J., Bliss-Isberg, C., Carson, K., Star, P., Donaghy, J., Cordova, M.J. … &
Spiegel, D. (2006). The effect of peer counseling on quality of life following
diagnosis of breast cancer: An observational study. Psycho-Oncology, 15, 10141022. doi:10.1002/pon.1037
81
*Gifford, A.L., Laurent, D.D., Gonzales, V.M., Chesney, M.A., & Lorig, K.R. (1998).
Pilot randomized trial of education to improve self-management skills of men
with symptomatic HIV/AIDS. Journal of Acquired Immune Deficiency
Syndromes and Human Retrovirology, 18, 136-144.
Goldstrom, I.D., Campbell, J., Rogers, J.A., Lambert, D.B., Blacklow, B., Henderson,
M.J., & Manderschied, R.W. (2006). National estimates for mental health mutual
support groups, self-help organizations, and consumer-operated services.
Administration and Policy in Mental Health and Mental Health Services
Research, 33, 92-103. doi: 10.1007/s10488-005-0019-x
* Gotay, C.C., Moinpour, C.M., Unger, J.M., Jiang, C.S., Coleman, D., Martino, S.,
Parker, B.J., Bearden, J.D., Dakhil, S., Gross, H.M., Lippman, S., & Albain, K.S.
(2007). Impact of a peer-delivered telephone intervention for women
experiencing a breast cancer recurrence. Journal of Clinical Oncology, 25, 20932099. doi:10.1200/JCO.2006.07.4674
Greenberg, L.S., Elliott, R., Watson, J.C., & Bohart, A.C. (2001). Empathy.
Psychotherapy, 38, 380-384. doi:10.1037/a0022187
Greenberg, P.E., Kessler, R.C., Birnbaum, H.G., Leong, S.A., Lowe, S.W., Berglund,
P.A., & Corey-Lisle, P.K. (2003). The economic burden of depression in the
United States: How did it change between 1990 and 2000? Journal of Clinical
Psychiatry, 64, 1465-1475. doi:10.4088/JCP.v64n1211
Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery
and Psychiatry. 23, 56-62. doi:10.1136/jnnp.23.1.56
Hans, E., & Hiller, W. (2013). Effectiveness of and dropout from outpatient cognitive
behavioral therapy for adult unipolar depression: A meta-analysis of
nonrandomized effectiveness studies. Journal of Consulting and Clinical
Psychology, 81, 75-88. doi:10.1037/a0031080
Hassenstab, J., Dziobek, I., Rogers, K., Wolf, O.T., & Convit, A. (2007). Knowing what
others know, feeling what others feel: A controlled study of empathy in
psychotherapists. The Journal of Nervous and Mental Disease, 195, 277-281.
doi:10.1097/01.nmd.0000253794.74540.2d
Hatcher, S.L., Favorite, T.K., Hardy, E.A., Goode, R.L., Deshetler, L.A., & Thomas,
R.M. (2005). An analogue study of therapist empathic process: Working with
difference. Psychotherapy: Theory, Research, Practice, Training, 42, 198-210.
doi:10.1037/0033-3204.42.2.198
82
Hattie, J.A., Sharpley, C.F., & Rogers, H.J. (1984). Comparative effectiveness of
professional and paraprofessional helpers. Psychological Bulletin, 95, 534-541.
doi:10.1037/0033-2909.95.3.534
Hedges, L.V. (1981). Distribution theory for Glass’s estimator of effect size and related
estimators. Journal of Educational Statistics, 6, 107-128.
doi:10.3102/10769986006002107
Hedges, L.V. (1982). Estimation of effect size from a series of independent experiments.
Psychological Bulletin, 92, 490-499. doi:10.1037/0033-2909.92.2.490
Hedges, L.V., & Olkin, I. (1985). Statistical methods for meta-analysis. Orlando, FL:
Academic Press.
*Heller, K., Thompson, M.G., Trueba, P.E., Hogg, J.R., & Vlachos-Weber, I. (1991).
Peer support telephone dyads for elderly women: Was this the wrong
intervention? American Journal of Community Psychology, 19, 53-74.
doi:10.1007/BF00942254
*Hill, W., Weinert, C., & Cudney, S. (2006). Influence of a computer intervention on the
psychological status of chronically ill rural women: Preliminary results. Nursing
Research, 55, 34-42. doi:10.1097/00006199-200601000-00005
Hodges, S.D., Kiel, K.J., Kramer, A.D.I., Veach, D., & Villanueva, B.R. (2010). Giving
birth to empathy: The effects of similar experience on empathic accuracy,
empathic concern, and perceived empathy. Personality and Social Psychology
Bulletin, 36, 398-409. doi:10.1177/0146167209350326
Hofmann, S.G., Sawyer, A.T., Witt, A.A., & Oh, D. (2010). The effect of mindfulnessbased therapy on anxiety and depression: A meta-analytic review. Journal of
Consulting and Clinical Psychology, 78, 169-183. doi:10.1037/a0018555
Hollon, S.D., Thase, M.E., & Markowitz, J.C. (2002). Treatment and prevention of
depression. Psychological Science in the Public Interest, 3, 39-77.
doi:10.1111/1529-1006.00008
Ilfeld, F.W. (1976). Further validation of a psychiatric symptom index in a normal
population. Psychological Reports, 39, 1215-1228.
doi:10.2466/pr0.1976.39.3f.1215
Jacobson, N.S., & Truax, P. (1981). Clinical significance: A statistical approach to
defining meaningful change in psychotherapy research. Journal of Consulting
and Clinical Psychology, 59, 12-19. doi:10.1037/0022-006X.59.1.12
83
Kessler, R.C., Berglund, P., Demler, O., Jin, R., & Walters, E.E. (2005). Lifetime
prevalence and age-of-onset distributions of DSM-IV disorders in the National
Comorbidity Study Replication. Archives of General Psychiatry, 62, 593-602.
doi:10.1001/archpsyc.62.6.593
Kessler, R.C., Borges, G., & Walters, E.E. (1999). Prevalence of and risk factors for
lifetime suicide attempts in the National Comorbidity Survey. Archives of
General Psychiatry, 56, 617-626. doi:10.1001/archpsyc.56.7.617
Kessler, R.C., Berglund, P., Demler, O., Jin, R., Koretz, D., Merikangas, K.R., Rush,
A.J., Walters, E.E., & Wang, P.S. (2003). The epidemiology of major depressive
disorder: Results from the National Comorbidity Study Replication (NCS-R).
JAMA: Journal of the American Medical Association, 289, 3095-3105.
doi:10.1001/jama.289.23.3095
Kroenke, K., Spitzer, R.L., & Williams, J.B. (2001). The PHQ-9: Validity of a brief
depression severity measure. Journal of General Internal Medicine, 16, 606-613.
doi:10.1046/j.1525-1497.2001.016009606.x
Lee, H. (1960). To Kill A Mockingbird. New York, NY: J.B. Lippincott Company.
*Letourneau, N., Stewart, M., Dennis, C., Hegadoren, K., Duffett-Leger, L., & Watson,
B. (2011). Effect of home-based peer support on maternal–infant interactions
among women with postpartum depression: A randomized, controlled trial.
International Jouranl of Mental Health Nursing, 20, 345-357.
doi:10.1111/j.1447-0349.2010.00736.x
*Lieberman, M.A., & Goldstein, B.A. (2005). Self-help on-line: An outcome evaluation
of breast cancer bulletin boards. Journal of Health Psychology, 10, 855-862.
doi:10.1177/1359105305057319
*Lieberman, M.A., & Goldstein, B.A. (2006). Not all negative emotions are equal: The
role of emotional expression in online support groups for women with breast
cancer. Psycho-Oncology, 15, 160-168. doi:10.1002/pon.932
Light, R.J., & Pillemer, D.B. (1984). Summing up: The science of reviewing research.
Cambridge, MA: Harvard University Press.
Lipsey, M.W., & Wilson, D.B. (2001). Practical Meta-Analysis. Thousand Oaks, CA:
Sage Publications, Inc.
*Lorig, K., Ritter, P.L., Villa, F.J., & Armas, J. (2009). Community-based peer-led
diabetes self-management: A randomized trial. The Diabetes Educator, 35, 641651. doi:10.1177/0145721709335006
84
*Ludman, E.J., Simon, G.E., Grothaus, L.C., Luce, C., Markley, D.K., & Schaefer, J.
(2007). A pilot study of telephone care management and structured disease selfmanagement groups for chronic depression. Psychiatric Services, 58, 10651072. doi:10.1176/appi.ps.58.8.1065
*Marmar, C.R., Horowitz, M.J., Weiss, D.S., Wilner, N.R., & Kaltreider, N.B. (1988). A
controlled trial of brief psychotherapy and mutual-help group treatment of
conjugal bereavement. American Journal of Psychiatry, 145, 203-209.
*McKay, H.G., Glasgow, R.E., Feil, E.G., Boles, S.M., Barrera, M. Jr. (2002). Internetbased diabetes self-management and support: Initial outcomes from the Diabetes
Network Project. Rehabilitation Psychology, 47, 31-48. doi:10.1037/00905550.47.1.31
McNair, D.M., Loor, M., & Droppleman, L.F. (1981). Profile of Mood States. San Diego,
CA: Educational and Industrial Testing Service.
Mead, N., Lester, H., Chew-Graham, C., Gask, L., & Bower, P. (2010). Effects of
befriending on depressive symptoms and distress: Systematic review and metaanalysis. The British Journal of Psychiatry, 196, 96-101.
doi:10.1192/bjp.bp.109.064089
Miller, W.R., Brown, J.M., Simpson, T.L., Handmaker, N.S., Bien, T.H., Luckie, L.F., et
al. (1995). What works? A methodological analysis of the alcohol treatment
outcome literature. In R.K. Hester and W.R. Miller (Eds.), Handbook of
Alcoholism Treatment Approaches: Effective Alternatives (2nd ed., pp. 12-44).
Boston: Allyn & Bacon.
Montgomery, E.C., Kunik, M.E., Wilson, N., Stanley, M.A., & Weiss, B. (2010). Can
paraprofessionals deliver cognitive-behavioral therapy to treat anxiety and
depressive symptoms? Bulletin of the Menninger Clinic, 74, 45-62.
doi:10.1521/bumc.2010.74.1.45
*Nagel, J., Cimbolic, P., & Newlin, M. (1988). Efficacy of elderly and adolescent
volunteer counselors in a nursing home setting. Journal of Counseling
Psychology, 35, 81-86. doi:10.1037/0022-0167.35.1.81
Olfson, M., Marcus, S.C., Druss, B., Elinson, L., Tanielian, T., & Pincus, H.A. (2002).
National trends in the outpatient treatment of depression. JAMA: Journal of the
American Medical Association, 287, 203-209. doi:10.1001/jama.287.2.203
85
*Onrust, S., Willemse, G., Van Den Bout, J., & Cuijpers, P. (2010). Effects of a visiting
service for older widowed individuals: A randomized clinical trial. Death
Studies, 34, 777-803. doi:10.1080/07481181003761252
Orwin, R.G. (1983). A fail-safe N for effect size in meta-analysis. Journal of
Educational Statistics, 8, 157-159. doi:10.3102/10769986008002157
Pfeiffer, P.N., Heisler, M., Piette, J.D., Rogers, M.A.M., & Valenstein, M. (2011).
Efficacy of peer support interventions for depression: A meta-analysis. General
Hospital Psychiatry, 33, 29-36. doi:10.1016/j.genhosppsych.2010.10.002
*Preyde, M., & Ardal, F. (2003). Effectiveness of a parent “buddy” program for mothers
of very preterm infants in a neonatal intensive care unit. CMAJ, 168, 969-973.
Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the
general population. Applied Psychological Measurement, 1, 385-401.
doi:10.1177/014662167700100306
*Robinson-Whelan, S., Hughes, R.B., Taylor, H.B., Hall, J.W., & Rehm, L.P. (2007).
Depression self-management program for rural women with physical disabilities.
Rehabilitation Psychology, 52, 254-262. doi:10.1037/0090-5550.52.3.254
*Roman, L.A., Lindsay, J.K., Boger, R.P., DeWys, M., Beaumont, E.J., Jones, A.S., &
Haas, B., (1995). Parent-to-parent support initiated in the neonatal intensive care
unit. Research in Nursing & Health, 18, 385-394. doi:10.1002/nur.4770180504
Rosenthal, R. (1979). The “file drawer problem” and tolerance for null results.
Psychological Bulletin, 86, 638-641. doi:10.1037/0033-2909.86.3.638
Rosenthal, R., & Rubin, D.B. (1982). Comparing effect sizes of independent studies.
Psychological Bulletin, 92, 500-504. doi:10.1037/0033-2909.92.2.500
Sheu, H., & Sedlacek, W.E. (2004). An exploratory study of help-seeking attitudes and
coping strategies among college students by race and gender. Measurement and
Evaluation in Counseling and Development, 37, 130-143.
*Silver, E.J., Ireys, H.T., Bauman, L.J., & Stein, R.E.K. (1997). Psychological outcomes
of a support intervention in mothers of children with ongoing health conditions:
The parent-to-parent network. Journal of Community Psychology, 25, 249-264.
doi:10.1002/(SICI)1520-6629(199705)25:3<249::AID-JCOP3>3.0.CO;2-R
Simon, G.E., Grothaus, L., Durham, M.L., VonKorff, M., & Pabiniak, C. (1996). Impact
of visit copayments on outpatient mental health utilization by members of a
health maintenance organization. American Journal of Psychiatry, 153, 331-338.
86
*Simoni, J.M., Pantalone, D.W., Plummer, M.D., & Huang, B. (2007). A randomized
controlled trial of a peer support intervention targeting antiretroviral medication
adherence and depressive symptomatology in HIV-positive men and women.
Health Psychology, 26, 488-495. doi:10.1037/0278-6133.26.4.488
Spearman, C. (1904). The proof and measurement of association between two things.
The American Journal of Psychology, 15, 72-101. doi:10.2307/1412159
Steele, L., Dewa, C., & Lee, K. (2007). Socioeconomic status and self-reported barriers
to mental health service use. The Canadian Journal of Psychiatry, 52, 201-206.
Stein, D.M., & Lambert, M.J. (1984). On the relationship between therapist experience
and psychotherapy outcome. Clinical Psychology Review, 4, 127-142.
doi:10.1016/0272-7358(84)90025-4
Sutton, A.J. (2009). Publication bias. In The Handbook of Research Synthesis and MetaAnalysis (2nd Ed), pp. 435-452, H. Cooper, L.V. Hedges, & J.C. Valentine (Eds.).
New York, NY: Russell Sage Foundation.
Toro, P.A., Reischl, T.M., Zimmerman, M.A., Rappaport, J., Seidman, E., Luke, D.A., &
Roberts, L.J. (1988). Professionals in mutual help groups: Impact on social
climate and members' behavior. Journal of Consulting and Clinical Psychology,
56, 631-632. doi:10.1037/0022-006X.56.4.631
Townes, D.L., Chavez-Korell, S., & Cunningham, N.J. (2009). Reexamining the
relationships between racial identity, cultural mistrust, help-seeking attitudes, and
preference for a Black counselor. Journal of Counseling Psychology, 56, 330336. doi:10.1037/a0015449
*Travis, J., Roeder, K., Walters, H., Piette, J., Heisler, M., Ganoczy, D., Valenstein, M.,
& Pfeiffer, P. (2010). Telephone-based mutual peer support for depression: A
pilot study. Chronic Illness, 6, 183-191. doi:10.1177/1742395310369570
*Tudiver, F., Hilditch, J., Permaul, J.A., & McKendree, D.J. (1992). Does mutual help
facilitate newly bereaved widowers? Report of a randomized controlled trial.
Evaluation & the Health Professions, 15, 147-162.
doi:10.1177/016327879201500202
*Uccelli, M.M., Mohr, L.M., Battaglia, M.A., Zagami, P., & Mohr, D.C. (2004). Peer
support groups in multiple sclerosis: Current effectiveness and future directions.
Multiple Sclerosis, 10, 80-84. doi:10.1191/1352458504ms973oa
87
Vevea, J.L., & Hedges, L.V. (1995). A general linear model for estimating effect size in
the presence of publication bias. Psychometrika, 60, 419-435.
doi:10.1007/BF02294384
Vevea, J.L., & Woods, C.M. (2005). Publication bias in research synthesis: Sensitivity
analysis using a priori weight functions. Psychological Methods, 10, 428-443.
doi:10.1037/1082-989X.10.4.428
Westen, D., & Morrison, K. (2001). A multidimensional meta-analysis of treatments for
depression, panic, and generalized anxiety disorder: An empirical examination of
the status of empirically-supported therapies. Journal of Consulting and Clinical
Psychology, 69, 875-899. doi:10.1037/0022-006X.69.6.875
Wilson, D. B. (2010). Meta-analysis macros for SAS, SPSS, and Stata. Retrieved
September 8, 2012, from http://mason.gmu.edu/~dwilsonb/ma.html.
Zung, W.W. (1965). A self-rating depression scale. Archives of General Psychiatry, 12,
63-70. doi:10.1001/archpsyc.1965.01730060026004