Context Despite the importance of mental illness in Africa, few controlled intervention
trials related to this problem have been published.
Objectives To test the efficacy of group interpersonal psychotherapy in alleviating
depression and dysfunction and to evaluate the feasibility of conducting controlled
trials in Africa.
Design, Setting, and Participants For this cluster randomized, controlled clinical trial (February-June
2002), 30 villages in the Masaka and Rakai districts of rural Uganda were
selected using a random procedure; 15 were then randomly assigned for studying
men and 15 for women. In each village, adult men or women believed by themselves
and other villagers to have depressionlike illness were interviewed using
a locally adapted Hopkins Symptom Checklist and an instrument assessing function.
Based on these interviews, lists were created for each village totaling 341
men and women who met Diagnostic and Statistical Manual
of Mental Disorders, Fourth Edition (DSM-IV)
criteria for major depression or subsyndromal depression. Interviewers revisited
them in order of decreasing symptom severity until they had 8 to 12 persons
per village, totaling 284. Of these, 248 agreed to be in the trial and 9 refused;
the remainder died or relocated. A total of 108 men and 116 women completed
the study and were reinterviewed.
Intervention Eight of the 15 male villages and 7 of the 15 female villages were randomly
assigned to the intervention arm and the remainder to the control arm. The
intervention villages received group interpersonal psychotherapy for depression
as weekly 90-minute sessions for 16 weeks.
Main Outcome Measures Depression and dysfunction severity scores on scales adapted and validated
for local use; proportion of persons meeting DSM-IV major
depression diagnostic criteria.
Results Mean reduction in depression severity was 17.47 points for intervention
groups and 3.55 points for controls (P<.001).
Mean reduction in dysfunction was 8.08 and 3.76 points, respectively (P<.001). After intervention, 6.5% and 54.7% of the intervention
and control groups, respectively, met the criteria for major depression (P<.001) compared with 86% and 94%, respectively, prior
to intervention (P = .04). The odds of postintervention
depression among controls was 17.31 (95% confidence interval, 7.63-39.27)
compared with the odds among intervention groups. Results from intention-to-treat
analyses remained statistically significant.
Conclusions Group interpersonal psychotherapy was highly efficacious in reducing
depression and dysfunction. A clinical trial proved feasible in the local
setting. Both findings should encourage similar trials in similar settings
in Africa and beyond.
Depression is a leading cause of disability in both developed and developing
regions of the world, including Africa.1,2 In
2000, we conducted a community-based survey in an impoverished part of southwest
Uganda that has been severely affected by the human immunodeficiency virus
(HIV) epidemic. Using Diagnostic and Statistical Manual
of Mental Disorders, Fourth Edition (DSM-IV)
major depression criteria, we found a current depression prevalence rate of
21% (P.B., unpublished data, 2000), consistent with previous research implicating
socioeconomic disadvantage and bereavement in depressive symptoms.
World Vision International, a nongovernmental humanitarian organization,
was interested in addressing this substantial mental health burden in Uganda.
Both antidepressants and psychotherapy have been shown to be efficacious in
numerous controlled trials in developed countries, including evidence of equivalence
in reducing the symptoms of acute depression.3 However,
use of antidepressants is not feasible in this region because of high cost
and limited supply infrastructure. Psychotherapy was therefore the preferred
option, although its use raised other issues. While there is substantial evidence
for the efficacy of "talking therapies,"4 these
have been developed in industrialized nations in the Western Hemisphere. The
extent to which the concepts and therapeutic strategies they use are appropriate
among other populations is unknown. In sub-Saharan Africa, conditions are
very different from those in which psychotherapy was developed, in ways that
could reduce effectiveness. For example, many populations are reluctant to
communicate directly about sensitive issues; others live in conditions of
extreme chronic deprivation that are rare in developed countries.
The need to test the local effectiveness of psychotherapy raised an
additional problem. Such testing has been hampered in Africa by a lack of
field methods for cross-cultural adaptation and validation of assessment instruments.
The lack of these methods, as well as perceived logistic and ethical difficulties,
have led some to believe that clinical trials of psychotherapy are not feasible
in Africa. We therefore began by developing a field method that has since
been successfully tested in 2 sites—Rwanda and the same villages in
Uganda as in the current study.5,6 In
both settings, we created or modified and then validated measures of depressive
symptoms and social functioning. These instruments were then used in community-based
prevalence surveys. The instruments developed in Uganda form the basis of
the current study.
The intervention we studied is a group-based interpersonal psychotherapy
(IPT) for depression. Extensive evidence for its efficacy and effectiveness
comes from randomized controlled clinical trials in which treatment was time-limited
and specified in a procedural manual.4 "Time-limited"
means that treatment is not open-ended but that the number, frequency, and
duration of sessions are specified at treatment outset. Selection of this
intervention allowed us to more accurately budget the intervention and also
made it cost-effective compared with open-ended therapies. The IPT manual
(available by e-mail from the authors at mmw3@columbia.edu or kfclougherty@aol.com)7 was essential
for accurate provision of IPT to this population. Prior experience also suggested
that the focus of IPT on interpersonal relationships was compatible with Ugandan
culture. The full rationale behind the development of IPT, its adaptation
for use in Uganda, and the training of local care providers is described elsewhere.4,7
This article reports the results of a controlled clinical trial of group
IPT. The study was conducted in the same Ugandan villages surveyed in 2000.
Screening and baseline assessments were conducted in February 2002. The IPT
took place from March through June 2002 (all groups began and finished within
a week of each other), and the follow-up assessment was conducted within 2
weeks of IPT completion.
Our purposes were (1) to test the efficacy of group IPT for Uganda (IPT-G-U)
in relieving depressive symptoms and improving functioning and (2) to evaluate
the feasibility of such studies in sub-Saharan Africa. To our knowledge, this
is the first published controlled clinical trial of a psychological intervention
in resource-poor sub-Saharan Africa.
The study area included all of Rakai province and the contiguous half
of Masaka province in southwest Uganda, the area of operations for World Vision
International. There are 154 villages in this area, each with several hundred
to several thousand adults and separated from other villages by surrounding
agricultural fields. Because of prior agreements with local leaders, we studied
the same 30 villages assessed in the 2000 prevalence survey. These were originally
chosen by weighted random sampling based on government census data. Hence,
sample size was based on these agreements rather than on sample size calculations.
However, calculations were later performed using a formula that accounts for
individual clustering within villages: m = 2(z∀+ z∃)2 Φ2 (1 + (n − 1)Δ)/nd2 where m is the number of clusters (villages) in each study arm; n
is the number of individuals per group; Φ2 is the variance
of the change score for each individual; Δ is the correlation of change
scores between any 2 individuals in the same village; and d2 is
the square of the expected difference between baseline and follow-up for controls
minus the expected difference for the intervention groups. These calculations
suggested that 30 is an appropriate sample size, assuming a 50% difference
in symptom severity scores between intervention and control groups after the
intervention, using a 2-tailed test with α = .05 and 80% power.
Villages formed the unit of randomization, each with 1 IPT or control
group. Because sex-segregated groups were more culturally acceptable, villages
were randomly sorted into blocks of 15 villages for male participants and
15 villages for female participants. Within each block, each village was randomly
allocated to the intervention or control study arm to ensure a sex-balanced
design. Study participants in the control villages did not receive IPT; however,
participants in both control and intervention villages were free to seek whatever
other interventions they wished throughout the study. Prior to randomization,
all potential participants were informed that if the intervention proved effective,
it would later be offered to controls (currently being implemented by World
Vision International).
The outcome measures were originally created for the 2000 study. They
include the depression section of the Hopkins Symptom Checklist,8 adapted
and validated for local use by a combination of ethnographic and quantitative
methods,6 and a sex-specific 9-item questionnaire
to assess functional impairment. This latter measure was also developed and
validated locally and is based on tasks that participants in the ethnographic
study reported were important to local people. For each task, respondents
were asked to state whether they were experiencing "no more" difficulty (scored
as 0), "a little more" difficulty (score, 1), "a moderate amount more" difficulty
(score, 2), "a lot more" difficulty (score, 3), or were "frequently unable
to do the task" (score, 4) compared with most others of their age and sex.
Overall severity of dysfunction was calculated for each individual by summing
the scores for all tasks. A similar process was used to calculate overall
depression severity using the responses to the Hopkins Symptom Checklist (details
of the development and validation of the depression and function questionnaires
have been published previously5,6).
Questions on demographics and duration of symptoms and 12 questions on HIV-related
knowledge, behavior, and attitude were added to form the final study instrument.
The latter were included to investigate the link between depression and HIV,
a topic outside the scope of this article.
To assess presence or absence of major depression, we used a DSM-IV9–based algorithm originally
developed by Mollica et al10 and expanded to
include the DSM-IV A, C, and E diagnostic criteria.
This algorithm includes the number and duration of symptoms and dysfunction.
The DSM-IV B and D criteria were not included because
they require exclusion of medical conditions and drug effects, which was beyond
our resources. The diagnostic algorithm is described in detail elsewhere.11
Study Eligibility and Screening
Eligible persons were identified using a 3-stage screening process performed
by 10 locally based World Vision International staff who had a minimum education
status of high school graduate. Nine had been interviewers for the 2000 study;
the other had previous survey experience.
Stage 1 screening involved contacting local leaders, healers, or other
knowledgeable persons in each village. They were asked for a list of at least
20 men or women (according to the village's sex assignment) aged at least
18 years who were believed by other villagers to have depressionlike illness.
In the local Luganda language, there is no single term to describe depression.
Instead the interviewers asked for persons with Yo'kwekyawa or Okwekubazida, 2 locally described depressionlike
syndromes identified in the 2000 study (P.B., unpublished data, 2000) These
syndromes are frequently comorbid and together include all the DSM-IV major depression symptom criteria.
For stage 2 screening, interviewers visited each person on the village
list. Informed consent was obtained by reading the consent form to the participant
because illiteracy is common in this area. If the respondent agreed to participate,
the interviewer signed the form as a witness. They then asked the respondent
if they thought that they had Yo'kwekyawa and/or Okwekubazida. If the person denied having either syndrome,
they were not interviewed further. The stage 1 and 2 screenings were performed
because the likelihood of depression was approximately 60% when both the individual
and others believed they had either of the local syndromes, according to the
2000 study data (P.B., unpublished data, 2000). If the person thought that
he/she currently had Yo'kwekyawa and/or Okwekubazida, the participant was interviewed using the study instrument,
which constituted the third stage of screening. Because each local syndrome
is only an approximation of depressive illness, self-report and outside reports
of their presence were used in screening only and not as outcome measures.
The purpose of this process was to develop lists of approximately 12
eligible persons per village—those who believed themselves and were
believed by others to have depressionlike illness and who at interview met
the DSM-IV criteria according to the algorithm. However,
unanticipated problems arose. First, in only 5 villages did at least 12 persons
meet the algorithm diagnostic criteria. Second, there was a very wide age
range among eligible persons in some villages; this was a problem because
experience suggests that IPT-G-U is more effective among participants of similar
age. Third, some interviewees stated that they were currently bothered "quite
a bit" or "extremely" by thoughts of suicide and, hence, could not be ethically
enrolled in a trial of an (as yet) unproven intervention.
The eligibility criteria were therefore revised: (1) diagnostic criteria
were expanded to include persons with subthreshold depression, meaning that
they fell short of a major depression diagnosis by a single DSM-IV symptom criterion; (2) in villages where 1 eligible person was
of a very different age than the rest, that person was excluded; and (3) persons
who appeared to be currently suicidal were revisited by the interviewer. If
the person confirmed a current danger of suicide, he/she was not enrolled
but referred to a clinical psychologist (L.N.) for further attention. In villages
where more than 12 interviewees met these criteria, we chose the 12 with the
highest depression scores.
Randomization and Participation
After creation of the lists of eligible participants, 8 of the 15 male
villages and 7 of the 15 female villages were randomly assigned to the intervention
arm and the remainder to the control arm. Random assignment was performed
by enumerating the villages and using a random-number table to determine study
allocation. Each list began with those who met the original diagnostic criteria,
followed by those who fell short by a single criterion, in order of decreasing
depression score. Interviewers visited each person in the order they appeared
on the list. The interviewer reread the consent form, advised the person about
the study group to which their village had been allocated, and asked them
to confirm their willingness to continue in the study. Interviewers continued
down the list until they had at least 8 participants (at which point they
did not contact the remainder of the list) or until they reached the end of
the list. The target group size was between 8 and 10, based on the clinical
judgment of the 2 senior IPT therapists (H.V. and K.F.C.).
Timing of the Follow-up Assessment and Interviewer Blinding
Both the intervention and control groups were reinterviewed with the
study instrument within 2 weeks of the intervention groups' completion of
IPT. Because interviewers likely knew which villages had received the intervention,
all interviewees were taken to the same town and interviewed there. They were
instructed not to mention their village during the interview or whether they
had received IPT. Each was interviewed by the same interviewer as at baseline
to reduce interviewer effects. The study was approved by the Johns Hopkins
University Institutional Review Board and by local government authorities
in Rakai and Masaka districts.
Interpersonal psychotherapy was originally developed for individual
treatment of major depression. Since then, it has been found effective for
a variety of mood and nonmood disorders in numerous clinical trials.4 A group adaptation of IPT was devised for treatment
of binge eating disorder, social phobias, adolescent depression, and posttraumatic
stress disorder. Interpersonal psychotherapy makes no assumption about etiology
but uses as a critical point for intervention the connection between symptom
onset and current interpersonal problems. The therapist begins with a careful
diagnostic assessment, then explains the diagnosis and works with the patient
to identify the problem areas associated with the onset of current symptoms.
Difficulties in 4 interpersonal areas are considered triggers of depressive
episodes, and 1 or more of these form the treatment focus: grief (due to death
of a loved one), interpersonal disputes (disagreements with important people
in one's life), role transitions (changes in life circumstances), and deficits
(persistent problems in initiating or sustaining relationships).
In Uganda, each IPT group met for 90 minutes weekly for 16 weeks. Groups
were led by a local person of the same sex as the group who had received 2
weeks of intensive training in IPT-G-U by 2 of the authors (H.V. and K.F.C.).
During each session, the group leader reviewed each participant's depressive
symptoms. The participant was then encouraged to describe the past week's
events and to link those events to his/her mood. The group leader then facilitated
support and suggestions for change from other group members. Attendance was
high; the dropout rate was 7.8%, and 54% of participants attended at least
14 sessions. Details of IPT and its adaptation for Uganda have been described
elsewhere.4,7
Baseline characteristics of intervention recipients and controls were
compared using χ2 tests for categorical data and t tests for continuous data. The main indicators of effect derive from
a comparison of the mean changes in the depression and function scores from
baseline to postintervention assessment in the intervention groups with those
in the control groups. Change in depression prevalence in both study arms
was also assessed using a dichotomous variable (1 = depression present and
0 = depression absent according to the algorithm).
Three sets of analyses were performed using different categories of
participants. The first analysis included only persons interviewed before
and after the intervention; we excluded those on the village lists whom we
tried but failed to recontact for inclusion in the
trial, those whom we found but who refused, and those whom we found and who
agreed but were lost to follow-up. The second analysis included all of these
persons. Therefore, this analysis consisted of all of those included in the
lists of eligible persons who were then sought by interviewers for inclusion
in the trial, whether or not they were found. The third analysis included
all of those on the lists of eligible participants, whether or not we tried
to recontact them. Hence, the second and third analyses
were on an intention-to-treat basis to allow for possible selection bias by
participants or interviewers. For participants whom we failed to recontact
and/or reinterview, we used an end-point analysis, a method widely used in
clinical trials of major depression disorder. The initial interview was carried
forward and imputed as the end-of-treatment score, thereby assuming no change
over the course of intervention for these individuals.
Since the unit of analysis is the participant and the study design is
a cluster randomized trial, we used a 2-level model to allow for within-village
(cluster) correlation and between-village variability.12 It
is also necessary to adjust for potential nonindependence of outcomes of the
individuals within each village. For continuous outcome measures such as change
in depression scores, we used mixed-effects models to determine the intervention
effect. In these models, treatment status was regarded as a fixed effect and
the effect of villages was treated as random. Mixed models were also used
to determine whether variation among group leaders was associated with the
change in depression scores. For dichotomous outcome variables such as prevalence
of depression, logistic regression models based on the generalized estimating
equations with robust variance estimates13 were
used to adjust for potential correlation of outcomes within a village. Odds
ratios and their corresponding 95% confidence intervals (CIs) were computed
for categorical outcomes and regression coefficients and corresponding P values were computed for continuous outcomes. The level
of statistical significance was set at P<.05.
We used Stata version 7.0 (Stata Corp, College Station, Tex) for the descriptive
analyses and SAS version 8.0 (SAS Institute Inc, Cary, NC) for the cluster
analyses.
In the 30 study villages, 631 people were identified by themselves and
others as having a depressionlike illness and consented to be interviewed.
Of these, 341 met the revised eligibility criteria, of whom 271 were diagnosed
as having major depression using the DSM-IV algorithm.
A total of 6 men and 7 women expressed suicidal ideation at the initial interview.
However, on further assessment, none were determined to be actively suicidal
and all were included in the study.
As described in the "Methods" section, in each intervention village
interviewers visited eligible persons in the order they were listed, inviting
them to join the village IPT group. When at least 8 persons had been recruited,
those remaining on the list were not approached (in some villages, interviewers
reached the end of the list before recruiting 8). In this way, interviewers
attempted to contact 139 of the 163 eligible persons on the intervention village
lists. Of these, 9 declined, 12 could not be found, and 2 had died since the
screening interview, leaving 116 enlisted. Nine of these 116 were lost to
follow-up for various reasons during the intervention, leaving 107 (53 men
and 54 women) who completed the intervention and were reinterviewed at follow-up. Figure 1 presents a flow diagram of the study
participants.
Similarly, we attempted to contact 145 of the 178 eligible persons on
the control village lists. Two had died since screening and 11 were unavailable,
leaving 132 who agreed to continue in the study (in some control villages,
>8 persons were recruited to compensate for recruiting <8 persons in other
control villages). Of these 132, 1 died during the intervention and 14 could
not be found, leaving 117 controls (55 men and 62 women) who were reinterviewed
(Figure 1).
Intervention and control groups did not differ significantly by age,
symptom duration, years of education, or baseline depression or function scale
scores. However, there was a significant difference in the proportions who
met the original depression diagnostic criteria, both among those who completed
the study and all those on the original lists of eligible participants (Table 1). (Tests for differences in baseline
characteristics were performed using standard significance tests and were
not adjusted for cluster effects. However, because we found a positive correlation
between clusters, adjusting for cluster effects would tend to reduce variance
and cause group differences to be even less significant than the values reported
herein.)
After completion of IPT, the point prevalence of major depression (those
who met the DSM-IV algorithmic criteria) was significantly
higher among the control groups than the intervention groups, regardless of
which participants were included in the analysis (Table 2). Similarly, the cluster-adjusted odds ratio for major depression
among controls compared with the intervention groups was also highly significant
and remained so in the intention-to-treat analyses. Adjustment for covariates
had very little effect (Table 3).
Change in Depression and Function Scores
The decline in depression scores was substantially greater among the
intervention groups (Table 4).
Mean changes were 17.47 and 3.55 for the intervention and control groups,
respectively (11.59 and 2.38, respectively, for the entire group of eligible
persons). Within the intervention groups, the mean change among men was 14.43
(95% CI, 12.32-16.55) compared with 20.46 (95% CI, 18.09-22.84) among women.
There was no significant sex difference among controls.
Improvement in function scores was also greater among the intervention
groups across all 3 analyses. When the change in depression score is included
in the model with change in function as the outcome, the latter is statistically
the same for both the intervention and control arms, indicating a close correlation
between these 2 variables.
We examined how change in ability to perform individual tasks varied
between the intervention and control groups. Table 5 shows all tasks assessed for both sexes and the mean change
between preintervention and postintervention interviews. We found a significant
difference in the amount of change comparing intervention participants with
controls for all female tasks except "consoling the bereaved." In contrast,
there were no statistically significant differences in the amount of change
for the individual male tasks.
This study demonstrated that IPT-G-U was effective for depressionlike
illness, depression symptoms, and associated dysfunction among persons in
our study sample. There was a highly significant decline in overall severity
of depression symptoms, the proportion of persons with major depression, and
dysfunction among those who received IPT-G-U compared with the control groups.
Unlike most function assessment instruments, ours was not restricted to health-related
problems but refers to total dysfunction due to all causes (including lack
of resources or assistance). Moreover, the tasks assessed were those previously
chosen by a sample of the population as being particularly important to them.6 Therefore, we believe that a large improvement in
the ability to do these tasks could significantly affect community welfare
and development, depending on how many people are affected by depression and
experience improvement. Although the effects on depression and function were
less when we based the analysis on intention to treat, they remained substantial
and highly significant.
The experimental nature of the study permitted conclusions about cause
and effect between depression and dysfunction. In our first assessment in
Uganda in 2000, we noted a strong association, but that cross-sectional study
could not determine whether depression was the cause or result of dysfunction.
This is important in developing countries like Uganda, where dysfunction is
common and due to multiple other causes, including physical illness. Since
IPT-G-U focuses on depression only, and function improved in this study only
in concert with depression (Table 4),
we conclude that depression is causing dysfunction. This is in accord with
longitudinal data from other countries14 and
suggests that targeting depression may be a useful way of improving function
among depressed persons. Why the intervention appears to have benefited women
in almost all areas of functioning considered separately whereas among men
no such benefits for any particular functional area were evident requires
further study.
Our results illustrate the importance of a control study arm in assessing
effectiveness. In this study, the control groups improved significantly over
the course of the intervention. Since the severity of mental symptoms and
function can vary with time, these changes are likely due to regression to
the mean. They may also be related to other actions taken by those in the
control groups, such as treatments by traditional healers (although those
we interviewed in 2000 thought they were unable to help with these problems).
Whatever the cause, a noncontrolled study of any intervention in this population
would likely have shown an effect regardless of its true efficacy.
Because we did not assess whether the control groups took any other
actions to relieve their depression, the trial comparison is not IPT vs nothing
but IPT vs the usual treatment, whatever that may be. (We did not assess whether
the intervention groups took any action to relieve their depression other
than attending the IPT sessions.) This approach was appropriate here because
our purpose was not to investigate replacing existing approaches. Rather,
it was to determine whether IPT-G-U would be a useful addition by World Vision
International to the current methods of coping with depression by local people.
When informing participants of their allocation to the intervention
or control arms, we reread the consent form and asked them to confirm their
continuing participation in one of the study arms. This contravened standard
procedures by which allocation is performed only after obtaining consent (or
renewing it, as here). This was done because of our prior experience in sub-Saharan
Africa, where random allocation is not well understood or accepted. Many previous
attempts by one of us (P.B.) to explain that people "might" receive something
were misinterpreted, resulting in resentment and withdrawals from the study.
To avoid this, we thought it essential to explain at each stage exactly what
would be happening to participants, particularly since one study arm would
be receiving nothing beyond usual treatment at this time. We expected this
approach to produce better data than the standard clinical trial procedure
because, in our experience, few people refuse enrollment when they believe
that they accurately understand what will happen. If refusals among both control
and intervention groups are few, then there is no significant participation
bias on the basis of study arm allocation. In this study, only 9 persons in
intervention villages and none in control villages refused to participate,
and demographic data (Table 1)
suggest that the intervention and control groups were not significantly different
overall. Once the study began, only 4 persons withdrew from the intervention
groups while none of the controls who were recontacted after the intervention
refused reinterview.
We are currently uncertain how long the effects of IPT will last. A
postintervention assessment of both the intervention and control groups will
be conducted 6 months after the intervention ended. We may find that it will
be necessary to add a maintenance component to prevent recurrence. This is
usually provided on a monthly basis.
To reduce interviewer bias, each participant was interviewed by the
same interviewer at baseline and follow-up. This makes it more likely that
an interviewer will know whether a person had the intervention, since they
may remember some interviewees and where they are from. We are uncertain how
often this occurred and, therefore, the extent to which blinding was compromised.
We still do not know the mechanism by which IPT was effective. Our study
design could not suggest which elements were essential and which were not.
For example, we cannot separate IPT per se from the group dynamics of simply
meeting together. Testimonials from participants suggest that the group problem-solving
element of IPT was vital, but this is not conclusive. We plan to investigate
this in future studies.
In conclusion, we note that this was the first time that the trainers
(H.V. and K.F.C.) had been in Africa or trained local people. This was also
the first experience of the IPT trainees; many were hesitant at first and
took weeks to gain confidence in the method. Under these circumstances, the
effects of this intervention impressed us. We might expect even greater impact
with more local experience with this approach.
1. Mental Health: New Understanding, New Hope. Geneva, Switzerland: World Health Organization; 2001.
2.Murray CJ, Lopez AD. Global mortality, disability, and the contribution of risk factors:
Global Burden of Disease Study.
Lancet.1997;349:1436-1442.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=9164317
Google Scholar 3.Ekin I, Shea M, Watkins J.
et al. National Institute of Mental Health Treatment of Depression Collaborative
Research Program: general effectiveness of treatments.
Arch Gen Psychiatry.1989;46:971-982.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=2684085
Google Scholar 4.Weissman M, Markowitz J, Klerman G. Comprehensive Guide to Interpersonal Psychotherapy. New York, NY: Basic Books; 2000.
5.Bolton P, Tang AM. An alternative approach to cross-cultural function assessment.
Soc Psychiatry Psychiatr Epidemiol.2002;37:537-543.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=12395144
Google Scholar 6.Bolton P. Cross-cultural validity and reliability testing of a standard psychiatric
assessment instrument without a gold standard.
J Nerv Ment Dis.2001;189:238-242.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=11339319
Google Scholar 7.Verdeli H, Clougherty K, Bolton P.
et al. Adapting group interpersonal psychotherapy (IPT-G-U) for a developing
country: experience in Uganda.
World Psychiatry.In press.Google Scholar 8.Derogatis LR, Lipman RS, Rickels K, Uhlenhuth EH, Covi L. The Hopkins Symptom Checklist (HSCL): a self-report symptom inventory.
Behav Sci.1974;19:1-15.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=4808738
Google Scholar 9.American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders,
Fourth Edition. Washington, DC: American Psychiatric Association; 1994.
10.Mollica RF, McInnes K, Sarajlic N, Lavelle J, Sarajlic I, Massagli MP. Disability associated with psychiatric comorbidity and health status
in Bosnian refugees living in Croatia.
JAMA.1999;282:433-439.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=10442658
Google Scholar 11.Bolton P, Neugebauer R, Ndogoni L. Prevalence of depression in rural Rwanda based on symptom and functional
criteria.
J Nerv Ment Dis.2002;190:631-637.Google Scholar 12.Omar RZ, Thompson SG. Analysis of a cluster randomized trial with binary outcome data using
a multi-level model.
Stat Med.2000;19:2675-2688.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=10986541
Google Scholar 13.Diggle PJ, Liang KY, Zeger SL. Analysis of Longitudinal Data. New York, NY: Oxford University Press; 1994.
14.Ormel J, VonKorff M, Ustun TB, Pini S, Korten A, Oldehinkel T. Common mental disorders and disability across cultures: results from
the WHO Collaborative Study on Psychological Problems in General Health Care.
JAMA.1994;272:1741-1748.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=7966922
Google Scholar