Objectives
This study aims to evaluate the effectiveness of the locally adapted Group PM+ intervention
in communities affected by adversity in Morang, Nepal. The cluster randomized controlled
trial (c-RCT) will compare Group PM+ to enhanced usual care (EUC) in participants
with high levels of psychological distress recruited from the community. The primary
hypothesis is that at 20 weeks after baseline, people receiving Group PM+ will have
lower psychological distress scores compared to people in the EUC control. The secondary
hypotheses are that people receiving Group PM+ will also report less severity of depression symptoms, posttraumatic stress disorder (PSTD) symptoms, personalized measures of distress, culture-specific symptoms of psychological
distress, somatic symptoms, and higher levels of functioning,social support at the
post-treatment assessments. We also hypothesizehigher levels of skill use related
to the Group PM+ intervention content.
A qualitative component is added to the project with the objective to explore the
feasibility and barriers to scale-up of Group PM+ with relevant stakeholders including
participants, families and Group PM+ facilitators.
Design and setting
The study is designed as a two-arm, single-blind c-RCT that will be conducted in a
community-based setting in Morang, a flood affected district in Eastern Nepal. Outcomes
will be measured on participants’ level at baseline and at two additional time points
mid-line and end line. Mid-line is seven weeks after baseline (for the Group PM+ participants,
this will be approximately one week after concluding the intervention). End-line is
20 weeks after baseline for the control arm which is approximate to 12 weeks after
the 5-sessions are completed in the intervention arm, with this timepoint as the primary
endpoint for the study.
Administrative levels in Nepal are: (1) provinces; (2) districts; (3) nagarpalikas or gaupalikas (municipalities or rural municipalities); (4) wards. Randomization will occur at the ward level, the smallest administrative level in Nepal,
with half of 72 enrolled wards receiving Group PM+ and the other half receiving EUC.
Importantly, given that the groups of the Group PM+ intervention will be of a single
gender (see details below in Group PM+ intervention) and that we do not have resources to enroll more than one group per ward, we will
select a sub-set of 14 of the 72 wards to be those which we enroll male participants
and the remaining 58 wards will enroll female participants. This fraction (14/72),
close to 20% of all wards, was selected to reflect the anticipated uptake of services
which was expected to be lower in this region than in studies conducted by our team
in other regions [15, 16]. Further, we note that the selection of 14 wards will not
be random but instead those 14 wards will be selected to be 14 wards that are close
together and that are, nevertheless, representative of the types of wards in the study
region. More specifically, we selected these 14 “male” wards close together so that
we can best use resources of the male personnel trained to deliver to the Group PM+
intervention. Because of the sub-selection of “male” and “female” wards, randomization
will be stratified by gender and will account for several other baseline cluster-level
covariates using restricted randomization (see details below in Randomization and sample size).
The c-RCT is the design of choice when an intervention is group-based and when the
population is expected to receive clinical and community services according to their
location (i.e. ward) of residence. An alternative design is an individually-randomized group treatment trial (IRGT) in which individuals, rather than clusters are randomized [17, 18]. An IRGT design
is typically expected to have greater power than a c-RCT for the same number of enrolled
individuals and same degree of outcome clustering. However, such a design would not
be suitable given concerns about contamination of the intervention within wards had
there been both Group PM+ and EUC participants in each ward.
Additional enrolment strategies will be employed to minimize the risk of contamination.
Specifically, given that some wards will be contiguous with each other, before participant
recruitment begins, we will map the area and specify a localised area within each
ward from which we will seek to recruit participants. The locations within the wards
will be selected so that recruited participants from each ward are geographically
far from those recruited in neighbouring wards to minimize the chance that participants
from different wards (i.e. from different clusters) interact with each other. Such
a strategy will be used to conserve independence of clusters and to avoid contamination
of EUC clusters with information from the Group PM+ intervention. Figure 1 gives an
overview of the design.
This community-based study is being conducted in 5 municipalities and 3 rural municipalities
that together encompass 72 wards within Morang, a densely-populated district in the
Eastern terai (lowland) region of Nepal. The selected areas have a diverse population with over
20 castes and ethnicities, including Tharu, Brahman/Chhetri, Yadav, and Rai. The national
language of Nepali is spoken by the majority of inhabitants. Morang is flood-affected
annually and in 2017, it was estimated that over 19,000 people were displaced and
over 12,000 homes were partially damaged due to the natural disaster [10]. There are three Primary Health Care Centers (PHCCs) within the selected areas that provide basic healthcare and have an attending health
worker trained in WHO mental health Gap Action Program (mhGAP) and will be used for
EUC referral.
Study arms
Problem Management Plus (PM+) is a WHO trans-diagnostic psychological intervention
that is delivered by trained non-specialist lay-providers in 5 sessions to adults
impaired by distress [1, 2]. The manual comprises of the following evidence-based
techniques: (a) problem solving, (b) stress management, (c) behavioural activation,
and (d) accessing social support.
The Group PM+ intervention consists of five 2.5 to 3 hour sessions in which participants
are taught techniques to manage their stressors and problems. Table 1 gives an overview
of the content of the 5 sessions. The aim is to have six to eight participants per
group, with separate groups for men and women and with gender-matched facilitators.
Information on seeking services from local health facilities with mhGAP-trained health
care staff trained in providing mental health care and/or psychosocial support is
provided to the Group PM+ participants as well as to the EUC participants.
Community psychosocial workers (CPSWs) are trained as Group PM+ facilitators [10]
. CPSWs are a cadre of community health workers that have a long track record in
providing psychosocial support in Nepal [19]. For this study individuals from the
community will be recruited to become new CPSWs. Fifteen local community women and
men who have completed higher secondary school (equivalent of 12th grade education) from the study region will be selected based on their basic communication
skills as reflected through the interviews., management and organization skills, interest
and motivation to serve community people, and commitment to work in the given time.
They are then given a 10-day basic CPSW training, with a standard curriculum developed
by TPO Nepal. The CPSW training includes an overview of psychosocial concepts, cause
and effects of psychosocial issues, basic communication skills, common mental health
problems in communities, group facilitation skills and psychoeducation. Competency
is evaluated before and after the CPSW trainig with a standardized role play assessment
tool (ENACT) that has been developed in Nepal and used for non-specialists in humanitarian settings
[20].
The CPSW training is followed by a 10-day Group PM+ training using the adapted manual
and other intervention materials. Group PM+ is named Khulla Man (“open heart-mind” in Nepali), which is consistent with Nepali ethnopsychological
models of distress, trauma, and recovery. The Group PM+ training includes learning
about the impact of adversity on mental health, basic counselling skills, how to deliver
the content of the Group PM+ manual, group managment skills and self-care. Competency
is assessed with ENACT again at the conclusion of the PM+ training, and fidelity is
assessed with a PM+ specific checklist.
After completing PM+ training, three rounds of practice sessions will be completed
by each CPSW in an adjoining district that is not a part of the study area. Competency
assessments and supervision will be conducted during these practise sessions. Based
on ENACT pre and post scores, clinical judgement during the PM+ practice sessions,
assessments using the fidelity sheet, and the PM+ competency criteria, twelve CPSWs
(ten female and two male) out of fifteen will be selected. In regards to ENACT, the
CPSW, who scores the lowest points i.e. 1 (Need improvement) for each item, will be
removed from the study.
Facilitators are supported by assistants called ‘Group PM+ helpers’ who receive a
basic 1-day training on assisting Group PM+ delivery and participate alongside CPSWs
in practice PM+ groups. They help with the logistics and organizational aspects of
the group sessions, such as reminding participants when sessions take place, reminding
those that do not show up for the sessions, and providing child care. Additional tools
such as calendars, session cards and reminders, all developed specifically for the
Nepal implementation of Group PM+, are used to increase retention of the material
and attrition by participants.
Enhanced Usual Care (EUC)
In rural regions of Nepal, care-as-usual for most people with mental health problems
until recently consisted of no psychological or psychiatric treatment in local health
facilities. People with severe mental conditions would often, after a long delay between
onset of symptoms, be taken to tertiary psychiatric services in the Kathmandu valley,
or other urban settings with psychiatric services, by family members [21]. The Programme
for Improving Mental Health Care (PRIME) has been implemented in Chitwan district,
in southern Nepal, and has implemented and evaluatedthe WHO mental health Gap Action
Programme (mhGAP) Intervention Guide since 2012 [22, 23]. The mhGAP Humanitarian Intervention Guide [24] was contextualized for Nepal after the 2015 earthquakes and
Nepali primary care workers in many districts, including Morang, have since been trained
using mhGAP. Both the EUC and intervention arm will receive a referral to mhGAP trained
primary health care worker providing treatment when needed (e.g. severe psychiatric
disorder or suicidality) .
Participants in the EUC control clusters will receive a time-restricted (between 30
and 45 minutes) family meeting conducted by local Community Informants, that will
consist of; (a) basic information on adversity and mental health, (b) benefits of
getting support, (c) information on seeking services from local health facilities
with mhGAP-trained health care staff trained in providing mental health care and/or
psychosocial support[10]. The mhGAP training that these health care staff recivied
consists of a 6-day training, focusing on a selected number of mental disorders including
common mental disorders, including an additional module on anxiety disorders (excluding
PTSD). This family meeting will be conducted with family members of the participant
or the participant only based on participants’ preferences. Both arms will receive
the same family meeting format and referral information to primary care-based treatment.
Randomization
The unit of randomisation is the ward (i.e. the cluster), as this is the smallest
unit of administration in Nepal. This unit was selected to ensure sufficient number
of clusters, as there are only 17 municipalities/ villages in the district, which
would the next possible level of randomization. Municipalities with mainly non-Nepali
speaking inhabitants will be excluded. A total of 72 wards will be selected for participation
with a target sample size of 8 participants enrolled per ward (see rationale below
in Sample size justification). Then, for the 36 wards randomly allocated to Group PM+, a single group of 8 participants
will be formed in each ward. As indicated above, of these 72 wards, 14 will be selected
as “male” wards and 58 as “female” wards to reflect differences in uptake of services
by males compared to uptake by females, as observed in earlier studies conducted by
our team (see above). As such, the overall estimated intervention effect will reflect
such a 1:4 ratio of males: females should the intervention be scaled up more broadly.
Furthermore, as noted above, we will not take a random sample of 14 wards as “male”
since it is important that the selected wards are such that whichever 7 are randomly
allocated to Group PM+ are sufficiently close in proximity so that it will be reasonably
straightforward for two male CPSWs to lead the 7 male Group PM+ groups (i.e. 1 in
each of the “male" Group PM wards).
Restricted randomization will be used. Specifically, we will first use stratification
by “ward gender” (i.e. randomization separately within 14 “male” wards and within
58 “female” wards). Then, within each “ward gender”, we will use covariate constrained
randomization to account for three baseline cluster-level covariates that are expected
to be related to participant outcomes and for which it is important for us to achieve
balance between the two study arms. Those three covariates, all defined as binary,
are: (1) access to mental health services (high or less than 1 hour to reach nearest
PHCC vs. low or less than 1 hour to reach nearest PHCC), (2) disaster risk (high or
landslides or flooding in the last 3-years vs. low-to-moderate or minimal landslides
or flooding in the last three years) and (3) rural/urban status (rural defined as
wards that do not touch a major highway, majority of homes made of wood/straw/mud,
and no local markets and urban defined as wards close to highways, majority of homes
made of concrete and access to local markets). Covariate constrained randomization
is a generalized form of stratification which can be used to simultaneously balance
on multiple baseline covariates without the need to formally define strata based on
the cross-classification of those covariates [25]. In practice, in order to perform
covariate-constrained randomization within the two strata defined by the 14 “male”
wards and the 58 “female” wards, we will separately implement covariate constrained
randomization in Stata software (version 14 [26]) using the cvcrand procedure [27]. Randomization will be performed in advance of enrolment of participants
and will be conducted by the study statistician who does not know the study region.
The statistician will use a simple data set with only the ward codes and 3 relevant
covariates to ensure that there is no room for bias in the implementation. Moreover,
a seed will be set so that the implentation is reproducible in Stata statistical software.
Sample size justification
The c-RCT was designed to have at least 90% power to detect moderate effect sizes
of 0.46 for the primary outcome of individual psychological distress, measured by
the GHQ-12 questionnaire (see details below in Outcome Measures) at the primary time point of 20 weeks follow-up (i.e. endline). An effect size of
0.46 would correspond to between-arm differences of 3.2 units in mean GHQ-12 for an
overall standard deviation of 7 units, a conservative assumption based on data from
our pilot c-RCT [28]. Power was calculated in R software (version 3.4.2) by programming
a standard calculation for a comparison of two means in a c-RCT with 72 clusters assuming
a two-tailed 5% significance level [29]. It was additionally assumed that 8 participants
would be enrolled in each ward, and that up to 2 participants per ward would drop-out
before outcomes were measured (a conservative assumption for the purposes of the power
calculation). Clustering of outcomes by ward was assumed to be relatively large with
an interclass correlation coefficient (ICC) of 0.2 based on baseline data from a cohort study in the Chitwan district used in
the PRIME study [22]. Although clustering in the EUC wards is anticipated to be lower
than the assumed 0.2 in the Group PM+ wards because EUC participants will not meet
in groups, we conservatively assume the same levels in both arms for the purposes
of the power calculation.
Participants
People living in the 72 selected wards in Morang district are eligible to participate
when they are over 18 years old, and understand and speak Nepali. Inclusion criteria
to be eligible for the trial are (1) answering affirmative to the heart-mind screener
and for functional impairment [30] and (2) scoring above 16 on the WHO Disability
Assessment Schedule for functional impairment (WHODAS) [31]. The heart-mind screener is locally developed (sensitivity of 0.94) and will
be used to determine the acceptability of local idioms of distress and impairment
due to these problems [30]. The WHODAS is a generic instrument assessing health and
disability that can be used with adult populations across cultures. Additionally,
only males will be eligible for enrolment in the 14 “male” wards and similarly, only
females will be eligible for enrolment in the 58 “female” wards. Exclusion criteria
for participation in the trial are (1) being at imminent risk of suicide (2) presence
of a severe mental disorder (e.g., psychosis) or cognitive impairment identified by
a score above 2 on an adapted version of the WHO Ten Questions Screen (TQS) for disability
detection [32] and (3) alcohol use disorder (score =>16 on the alcohol use disorders
identification test (AUDIT).
Imminent risk of suicide will be determined through a structured screening questionnaire.
Persons with current suicidal ideation and suicide plans or recent attempts will be
referred immediately to a psychosocial counsellor but will not be excluded from participating
in the study. Observable symptoms of psychosis and severe cognitive impairment will
be assessed using an observation checklist. Four items are included to examine the
client’s ability to comprehend questions and follow basic instructions, and the degree
to which the client can communicate with the assessor. A positive response above 2
on any of these behavioural items is an indication for exclusion and is discussed
with a supervisory team. Alcohol dependency will be assessed by the alcohol use disorders
identification test (AUDIT) [33]. According to WHO’s guidelines for AUDIT use in primary
care, people that score below 16 can benefit from simple advice [34]. Those with a
score of 16 or over would benefit the most from advice plus brief counselling and
continued monitoring and therefore, those that score 16 or above on the AUDIT will
be excluded from the study and referred to a near-by mhGAP trained health professional[10]
.
Procedures
Each ward of participating municipalities in Morang district will have 1 community
informant (CI) who will conduct recruitment through the use of the Community Informant Detection Tool (CIDT) and community sensitization activities. CIs are often Female Community Health Volunteers
(FCHVs), mother’s group members, or social mobilizers within their respective communities.
CIs will, as much as possible, also be gender-matched for the “gender” of their wards.
CIs from intervention and control wards will be trained separately to maintain blinding.
Control ward CIs will not be given any information on Group PM+ or any other information
about the existence of an intervention arm. Intervention CIs will additionally be
given a 1-day training to become Group PM+ ‘helpers’ for the sessions.
The community informants (CI) will be trained on the CIDT to identity people with
common mental disorders in the community. The CIDT is a pro-active case detection
approach aimed to increase helpseeking using a vignette-based tool designed for the
ease of use by lay people. It has been developed and tested in Nepal [35], with positive
results on the positive predictive value (0.68) and increasing the utilization of
mental health services [36]. A general distress CIDT version had been adapted for
this trial (cite 16), which includes gender-matched vignettes for the “gender” of
the wards.
After the community informant identifies a person in the community who matches the
symptoms described in the vignettes, they will be asked if they would like support
for their problems. If so, the research assistant (RA) will then conduct the consent
and screening procedures.
People who are identified as meeting the exclusion criteria initially by the RAs will
be referred to health workers trained in mhGAP, hospitals with psychiatric services,
or counselors. People that meet the inclusion criteria for the study, in both the
intervention and control wards, will receive a visit from the CI for a family meeting.
Based on the preference of the participants this can either be with or without their
family.
Informed consent
The consent procedures consists fo two steps, first informed consent for screening
and then informed consent for participation in the Group PM+ trial [10] . After identification
by the CI, potential eligible people will be approached by the research assistant
for informed consent for screening. If a participant screens positive, the CI will
give more information about the research project and will conduct the full trial informed
consent during the family meeting.
All respondents who decide to participate will provide written consent, if possible.
Full information on the study will be provided in local, lay Nepali language before
obtaining consent from each participant. Given high rates of illiteracy, the consent
form will be read to all participants. After providing verbal consent, literate participants
will be asked to acknowledge the process with a signature. For illiterate participants,
verbal consent or adding a symbol or sign will be sufficient. We will make sure that
potential participants fully understand what participation entails and that they,
at any time and withour any consequences, can withdraw their consent without having
to give an explanation.. Participants will be made aware that refusal to participate
will not have an impact on any type of support they receive outside the study. For
the qualitative interviews, separate written informed consent will be taken at the
time of the interview.
Outcome measures
The primary outcome is levels of individual psychological distress, measured by the
GHQ-12 [37, 38] at end-line, 20 weeks after baseline. The GHQ-12 consists of 12 questions
that are scored on a 4-point Likert scale ranging from 0 to 3, with higher total scores
representing higher levels of distress. The GHQ-12 has been translated and clinically
validated in Nepal (Cut-Off: 1/2, Sens 85.6%, Spec 75.8%, PPV 86.7%, NPV 84%) [39].
Secondary outcomes include levels of depressive symptoms measured by the Primary Health
Questionnaire (PHQ) [40]; general functioning measured with the WHO Disability Assessment
Scale (WHODAS) [31]; post-traumatic stress disorder (PTSD) symptoms measured by the Post-traumatic
stress disorder Check List (PCL-5) [41]; levels of perceived social support measured
by the Multi-dimensional Scale of Perceived Social Support (MSPSS) [42]; somatic complaints
by the Psychosocial Mental Health Problems (PMHP)[5], the Somatic Symptom Scale –
8 (SSS-8) [43]; Self-report wellbeing outcomes are assessed using the Psychological
Outcomes Profile Instrument (PSYCHLOPS)[44]. Please see table 2 for an overview of
the different measures on different time-points.
The WHO Disability Assessment Scale (WHODAS) is a generic instrument assessing health
and disability in adults. It assesses difficulties that people are experiencing during
the last 30 days, due to their illness, across six domains of functioning (cognition,
mobility, self-care, getting along, life activities, and participation). Difficulties
are scored on a 5-point Likert scale of: not at all difficult, a little difficult,
sometimes difficult, very difficult, or always difficult. The WHODAS can be used with
all diseases and across cultures. The scale has been previously used in Nepal and
has an good internal consistency between items (α = 0.90) and validity with multiple
mental health measures for depression (r = 0.616, p < 0.001), anxiety (r = 0.624,
p < 0.001), and PTSD (r = 0.499, p < 0.001)[4, 45].
The Patient Health Questionnaire (PHQ-9) is a 10-item instrument measuring symptom
depression [40]. It has been translated and clinically validated in a primary care
population in Chitwan, Nepal: the validated cut-off score of ≥10 (sensitivity =0.94,
specificity = 0.80, positive predictive value (PPV) = 0.42, negative predictive value
(NPV) = 0.99, positive likelihood ratio = 4.62 and negative likelihood ratio = 0.07)
[30].
The original Post-traumatic stress disorder Check List PCL-5 is a 20-item checklist
corresponding with the 20 DSM IV PTSD symptoms. To diminish the burden of questionnaires
administered by participants in this study the 8 -item version will be used. This
was shown in a recent study to have comparable diagnostic utility to the 20-item PCL-5
[46] and has been used in Nepal and will be used in this study to diminish the burden
of questionnaires administered by participants [47].
The Multidimensional Scale of Perceived Social Support (MSPSS) [42] is a self-rating
tool of perceived social support from three categories of support: family, friends,
and significant other. It has been locally adapted [48] and validated to use in Nepal
[49]. The MSPSS consists of 12 questions that are rated on a 5-point Likert scale
ranging from 1 “very strongly disagree” to 5 “very strongly agree”. Higher scores
indicate higher perceived levels of social support.
The Somatic Symptom Scale (SSS) is an 8-item patient-reported outcome measure of somatic symptom burden [43]
that has been translated and adapted using standard cross-cultural approach [50].
PSYCHLOPS is a self-report measure that covers the main problem that participants
are currently facing in one question. Participants are asked to give free text responses
to the problem domain. PSYCHLOPS has been validated in primary care populations across
several countries [51, 52] and will be administered at baseline, mid and end line.
Other measures and further data Competency and fidelity will be assessed with a modified version of the Enhancing Assessment of Common Therapeutic Factors (ENACT) tool tailored for Group PM+ [53]. The ENACT scale is an 18-item assessment for common
factors in psychological treatments that can be used with non-specialist in different
settings.
At baseline demographic characteristics of participants will be recorded, including
age, years of education, occupation and living situation. Traumatic events will also
be assessed with the Traumatic Events Inventory (TEI), an 11-item assessment of traumatic
exposure associated with poor mental health outcomes [54]. The TEI has previously
been used in Nepal [55] . A natural disaster questionnaire has also been developed
for this trial. This consists of five questions on if participants were affected by
floods, earthquakes, landslides, fires or other natural disasters in the last 5 years.
Participants will be asked if their property were damaged and if they themselves of
any relatives and friends were hurt by such natural disasters. Behavioral and psychosocial
skills related to coping with emotional distress will be assessed with the Reducing
Tension Checklist, that contains 10-item assessment of behavioural and psychological
skills to evaluate skill acquisition of PM+ skills. It has been adapted based on PM+
content and findings in phase 1 of the project (cite 16).
During PM+ sessions the Subjective Units of Distress Scale (SUDS) will be used. The
SUDS, a scale of 0 to 10 for measuring the subjective intensity of disturbance or
distress currently experienced by an individual [56], will be used for each participant
during the second to fifth PM+ sessions. The scale has been previously used in Nepal
[57].
Masking
In this project, research assistants administering all interviews, community informants,
research supervisors, and study statisticians will be blinded. The intervention does
not allow for the intervention facilitators and participants to be blind to treatment
allocation. Blinding of assessors will be ensured by minimizing the chance of contact
between assessors and facilitators and having two separate offices for the research
and clinical staff. Assessors will also prompt participants not to share any information
on the type of treatment that they receive and explain that they are not supposed
to know. After each assessment, assessors will be asked to indicate what treatment
they think each participant will or has received (e.g. medication, one-on-one counselling,
group counselling, referral etc). This will provide some data on the amount of unblinding
that might occur in the RCT. Furthermore, each of the research assistants sign a contract
in which they agree to not share any details of the study with others.
Given the challenges of blinding in c-RCTs and the concerns about the potential for
selection bias given that participant recruitment occurs after randomization of the
wards in which the participants reside [58], we have used the “timeline cluster” to
visualize procedures in relation to blinding and participant recruitment [59]. Specifically,
we generated Figure 2 using an online open-access tool developed by the “timeline
cluster” authors [59]. This figure provides additional details to complement the overall
study flow chart (Figure 1), including information on whether a specific stage of
the process pertains to clusters, to participants or to both. The dark boxes indicate
stages in the procedure when both participants, and the study personnel who will interact
with those participants, will blinded to which arm the cluster has been allocated
to. We will use a design so that study participants are recruited by trained RA’s
who do not know which arm the ward (cluster) has been assigned to (see up to stage
7 in each arm, Figure 2). During service delivery (stages 8-9a in Group PM+ and stage
9b in EUC), participants cannot be blinded to study arm. However, as noted above,
we have designed the midline and endline data collection procedures such to try to
ensure that the RA’s conduct the interviews are blinded to study arm (Stages 10-11
in both arms), which is indicated by the light grey shading (i.e. indicating partial
blinding because the participants are no longer blinded at this stage). Importantly,
when commencing the interview, the RA will emphasize to the participants how important
it is that the participants does not reveal details about what kind of services they
have received. We recognize that, within a specific ward, if an RA is inadvertently
unblinded while conducting the interview with a participant before the final interview
in that ward (i.e. before interviewing the 8th of the 8 enrolled participants), that RA would therefore be unblinded for the interviews
of remaining participants. We will record data as to whether such unblinding occurred
and therefore will be able to report on any threats to data validity. And, even in
such a case, the RA’s receive rigorous and comprehensive training on procedures to
objectively record responses to our instruments and measures and therefore, we expect
to be able to mitigate any potential for measurement bias that could arise as a result
of unblinding.
Data management
The research team will keep the identifying key, linking the name to code numbers,
in a secure location and only the study principal investigators (PI) of the study
will have access to primary data. Research assistants will not enter any personally
identifying details into the data set. Data will be collected using a password-protected
tabled, from where data will be synchronized and uploaded in the Open Data Kit (ODK),
saved on a private server, and transferred to a data-analytic computer program (STATA)
without the identifying key. The site PI will conduct quality assurance checks on
data collected by the research assistants on the tablet.
The data collected with other means, like qualitative data and other documentation
(e.g. supervision forms and training reports), will be safely stored in locked cabinets
at the site office. The qualitative data will be fully anonymized and coded and will
not contain any identifying information. Results of this project will be published
regardless of being negative or positive results and submitted to peer-reviewed scientific
journals.
Data analyses plan
Analysis of quantitative outcomes, including the primary outcome of GHQ-12, will adopt
the intention-to-treat approach whereby all participants will be analysed according
to the arm to which their ward was randomized. That is, even if intervention arm participants
did not attend all Group PM+ sessions, the primary analysis will include them in the
Group PM+ arm. The linear mixed effects modeling approach will be used to model participant-level
score outcomes. More specifically, the two follow-up time points (mid-line 7 weeks
after baseline and end-line 20 weeks after baseline) will be analysed within the same
model. The following design variables will be included as fixed effects: arm, time
(an indicator for the 20-week follow-up time-point), the arm-by-time interaction (to
allow for different intervention effects at each of the two follow-up time-points),
ward gender (to account for the stratified design) and the three covariates used in
the constrained randomization procedure (i.e. access to mental health services, disaster
risk and rural/urban status). To increase statistical power, each participant’s baseline
measure of the outcome will be adjusted for as a fixed effect [60]. To account for
clustering by ward, a random intercept will be included for which the degree of clustering
is allowed to differ for intervention and control arm clusters. Due to the repeated
follow-up measurements on participants, a random intercept will be included for participant.
In the event that baseline outcome data are missing, we will use a constrained longitudinal
analysis approach whereby the baseline measure is also modelled as an outcome (rather
than a covariate) and the baseline mean level is constrained to be equal between arms
[60]. In this case, we will allow for changing correlation of outcomes over time by
additionally including a random slope for each individual or by using an unstructured
residual correlation matrix. For score outcomes for which the assumptions of the linear
mixed model are violated, we will transform the outcomes (e.g. log-transformation)
or adopt a bootstrap approach to estimate confidence intervals. Binary outcomes will
be analyzed within the generalized estimating equations framework. Specifically, we
will use the modified Poisson approach [61] assuming a Possion outcome distribution,
with an exchangeable working correlation matrix and robust standard errors to account
for the outcome model misspecification (i.e. Poisson instead of binomial). Such an
approach has been shown to be preferable to a binomial regression model for clustered
outcome data [61]. A log link will be used to obtain risk ratios and an identity link
to obtain risk differences and the mean model will include the same terms as the models
for the continuous outcomes.
Additional supportive analyses will test robustness to missing outcomes, to baseline
covariate imbalance and to the combination of both. Specifically, the supportive analyses
will include the following three approaches: (1) analyses that account for any baseline
covariates that are predictive of missing outcomes, (2) analyses that account for
any baseline covariates identified to be imbalanced between treatment arms, and, (3)
analyses that combine both approaches (1) and (2), i.e. that account for all baseline
covariates identified to be predictive of missing outcomes or to be imbalanced. For
approach (1) to assess robustness to missing outcome patterns, if the probability
of missingness is only related to the baseline covariates in the model, then these
adjusted analyses will provide valid estimates of the intervention effect having accounted
for the missing data patterns.
Sub-group analyses will assess whether there are differing intervention effects according
to the following variables: gender and baseline depressive symptoms. To do so, the
model will include an indicator for the sub-group variable and interactions between
that indicator and intervention arm and time-point. Baseline depressive symptoms will
be included in the model as a binary variable indicating whether the participant met
the cutoff score for depressive disorder, specifically a a baseline PHQ-9 score of
10. These analyses are exploratory in nature as the study is not powered to detect
such effects. Adherence in the intervention arm will be quantified through the number
of sessions attended. Similarly, within the intervention arm, we will examine potential
differences in intervention due to different facilitators. To do so, we will analyze
outcomes in intervention arm only and see its relationship with facilitator. Likewise,
within the intervention arm, we will examine whether estimated outcomes are different
for those who completed all five sessions vs. those who completed fewer sessions.
We hypothesize that skills acquired will mediate any impact of the intervention.To
this end, we will perform a mediation analysis within the framework outlined by Zhang
et al [62] that accounts for the multilevel (i.e. clustered) data structure. We will
use the midline measure of the Reducing Tension Checklist as the mediating variable and the endline timepoint for outcomes of interest. We
note two important features of this analysis: (1) we have selected the midline measure
for the hypothesized mediating variable to ensure that it precedes the outcome measure
in time in order to be able to make stronger causal claims than we would were the
mediator and outcome measured at the same point in time, and, (2) we will ensure that
potential confounders of the mediator-outcome relationship are accounted for in the
analysis.
Qualitative evaluation
Semi-structured interviews will be conducted with a subsample of Group PM+ participants
(equal number of completers and non-completers); Group PM+ facilitators; control arm
participants; research assistants; family members of Group PM+ participants (equal
number of intervention completers and non-completers); community informants; and local
decision makers. The interviews will be conducted by trained interviewers that are
familiar with the key principles of qualitative interviewing. Interviews will follow
a semi-structured topic guide that address themes around barriers and facilitators
in implementing PM+, satisfaction with the intervention, barriers and facilitators
to adherence, and barriers and facilitators to scale up and integrating Group PM+
into other services.
All interviews will follow the same process: Group PM+ participants and other KIs
will be selected through convenience sampling. Informed consent will be obtained using
a single step procedure where participants are provided oral and written information
about the study and its purpose in the local language. The number of KI interviews
in each category of respondent will be determined by empirical saturation, with a
minimum of 2 - 16 participants per each category. FGDs will also be conducted in relevant
categories.
Qualitative data analyses
The qualitative data collected from FGDs, key informant interviews and notes during
the process evaluation will be coded in NVIVO [63] and analysed using content analysis[64]on
the translated transcripts of the original language. Coding will be conducted by multiple
independent raters, and inter-rater reliability will be calculated using Kappa scores.
Ethical considerations
Throughout the different study phases participants in both arms will have access to
mhGAP trained health staff in the districts. When necessary they will be referred
to a specialist for further assessment or management of severe psychiatric problems.
If a participant experiences psychological problems after the project, they will be
offered additional support.
All adverse events (AE) and serious adverse events (SAEs) that are reported spontaneously by the participant or observed by either research
or intervention staff and will be recorded. All staff will be trained in the TPO Nepal
Adverse Events Reporting Mechanism which guides the process of reporting and supporting/referral
in case of any adverse events.
All AEs and SAEs will be reported to a local independent Data Safety Management Committee (DSMC). The DSMC includes psychiatrists, non-governmental organization experts in psychosocial
programs, and researchers and is established specifically for oversight of the trial
and review of SEs and SAEs. The chair or a nominated person from the DSMC will review
SAEs within 48 hours, deciding if an SAE is likely related or unrelated to the intervention.
The DSMC will review all AEs once a month. In both instances the committee will where
necessary determining any appropriate action in respect of ongoing trial conduct (i.e.
referral to specialised care). All changes in treatment resulting from Adverse Events
or Serious Adverse Events will be reported to the DSMC in Nepal. TPO Nepal is responsible
for data collection and storage and making data available to the DSMC, funders, and
IRBs for audits when appropriate.
The project has been approved locally by the Nepal Health Research Council, Kathmandu,
Nepal and by the WHO Ethical Review Committee (Version 3; Protocol ID: 2817, October
25, 2018).
Dissemination
Findings from the c-RCT will be published through various channels. In Nepal the results
will e disseminated to key stakeholders, including district, provincial, and national
government, through Nepali and English reports and presentations. Internationnaly,
the findings will be published in academic journals, reports to the research funder
(Office of U.S. Disaster Foreign Assistance/USAID) and disseminated through the Mental
Health Innovation Network (www.mhinnovation.net) For authorship eligibility we will comply with guidelines of the International
Committee of Medical Journal Editors. Also,additional attention will be given to recommendations
for equitable representation of researchers from LMIC for academic authorship [65].
After publication of the primary analyses, the data will be made publicly availablto
keep with transparency recommendations.