A Pair-Matched Randomized Evaluation of Faith-Based Couples Counseling in Uganda
Last registered on April 17, 2019

Pre-Trial

Trial Information
General Information
Title
A Pair-Matched Randomized Evaluation of Faith-Based Couples Counseling in Uganda
RCT ID
AEARCTR-0003994
Initial registration date
March 11, 2019
Last updated
April 17, 2019 9:08 AM EDT
Location(s)

This section is unavailable to the public. Use the button below to request access to this information.

Request Information
Primary Investigator
Affiliation
Princeton University
Other Primary Investigator(s)
PI Affiliation
International Rescue Committee and University of Chicago
PI Affiliation
Harvard University
PI Affiliation
Johns Hopkins University
PI Affiliation
Princeton University
Additional Trial Information
Status
On going
Start date
2018-09-21
End date
2019-12-31
Secondary IDs
Abstract
The Becoming One program is a skills-based curriculum designed for faith leaders to use with couples in their congregation seeking counseling prior to or following a wedding. The curriculum uses a biblical framework to communicate about practical skills for improving communication, emotional regulation, shared control over financial resources, financial planning, and sexual consent and pleasure. Faith leaders lead the participatory sessions with groups of couples, asking both members of the couples to attend together. The 12 total sessions are intended to be delivered once a week with each session lasting approximately 90 min. At the heart of the Becoming One intervention is the idea that individuals and couples will change their behavior when they are already self-motivated to do so, and when the source of advice or counsel is a person of social, religious, or political significance for them.

This study measures the effect of the Becoming One (B1) program through a block-randomized control trial. There are three waves of measurement: a pre-randomization baseline, a midline survey approximately five weeks after the "intervention" group has completed the couples counseling program, and an endline survey approximately one year after the "intervention" group has completed the program, but before the "comparison" group has gone through couples' counseling.

The random assignment works by forming couples into pairs within congregations (groups organized by the faith leaders) by matching on their baseline levels of physical violence, and then randomizing one couple in each pair to the intervention group, and one couple to the comparison group. The baseline levels of physical violence include an average of six binary indicators (standardized to have a mean of 0 and standard deviation of 1) for whether the woman in the couple reports having been pushed, slapped, arm-twisted / hair-pulled, punched, kicked, or choked over the preceding 12 months.

The main analyses of the experiment take place at the couple-level. We will construct non-parametric p-values calculated using randomization inference. Using simulation, we will report the alphas (type 1 error risks) one would need to apply at the test-level in order to reject at least one test in the family of main analyses only 5% and 10% of the time, under the global null of no effect on any outcome for any unit.

Our main estimator is a covariate-adjusted estimator below, which, in addition to conditioning on an indicator for the treatment assignment and block-level fixed effects, adjusts for predictive covariates using a mean-centering and interaction approach described in Lin (2013). For each outcome, we use a cross-validated Lasso estimator to select only those covariates that are predictive of the outcome, irrespective of whether those covariates are imbalanced. We justify this approach through a simulation study.

We expect to encounter two-sided non-compliance in this study. For our main analyses, our estimand is an intent-to-treat (ITT) effect.

To reduce the number of tests we are running, our main analyses focus on four outcomes that combine many theoretically-related measures:

1. Intimate Partner Violence: a binary indicator that is 1 if the female partner has experienced any of ten types of physical or sexual violence over the five months preceding the midline survey, and 0 otherwise. We expect that the ITT effect is negative and so will conduct a one-sided test of the null hypothesis of zero or positive effects.

2. Control and Decision-Making: an index of several groups of items, each of which is coded to vary between 0 and 1. The groups include decision-making about household finances and purchasing decisions, financial control, and social control. When at 0, this implies the female experiences all forms of financial and social control, and both partners perceive her to be excluded from all decision-making. When it is at 1, this implies the female partner does not report experiencing any forms of financial or social control, and is involved in all aspects of decision-making. We expect that the ITT effect is positive, and will conduct an upper-tailed test of the null hypothesis of zero or negative effects.

3. Sexual Consent and Autonomy: an index of items, mostly answered by the female partner but one answered by the male partner, about whether the male takes a non-coercive strategy when his partner refuses sex, and how the female partner feels about her control over decisions about and initiation of sex. The responses are coded to vary between 0 and 1. When the index is at 0, this implies that the female partner reports sexual coercion across all of the domains, and that the male partner reports coercive behavior. When it is at 1, this implies the absence of coercion.

4. Communication and conflict management techniques. An index of items addressing a broad range of domains of communication, answered by both female and male partners. The domains include communication about sex, about daily events and companionship, and techniques for de-escalation when arguing about difficult topics. When the index is at 0, this implies that the partners agree that they have had no supportive or open communication or techniques for peaceful or de-escalating communication across many domains. When it is at 1, this implies the partners agree that they communicate with supportive or open communication or techniques for peaceful or de-escalating communication across all the domains.

We will base our main confirmatory inferences on these four outcomes, which also constitute the family of tests for which we will construct multiple comparison-adjusted alphas as above.

In order to explain why these outcomes move, we will report and describe treatment effects graphically by plotting point estimates on the underlying items. We also have a range of secondary analyses that target either variables that we expect mediate the effect of B1 on these four outcomes, or other outcomes that are not of primary interest. All of these will be reported as analyzed, possibly in an appendix. We will note departures from the analysis plan, and label tests as pre-registered or not.
External Link(s)
Registration Citation
Citation
Annan, Jeannie et al. 2019. "A Pair-Matched Randomized Evaluation of Faith-Based Couples Counseling in Uganda." AEA RCT Registry. April 17. https://www.socialscienceregistry.org/trials/3994/history/45056
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
Interventions
Intervention(s)
FOR MORE DETAILS SEE PAP

Intimate partner violence (IPV) is thought to touch the lives of 30% of women globally and over 44% of married women living in rural Uganda. Efforts to reduce rates of abuse often seek to effect radical normative change through intervention by outside actors. Despite concerns that outside actors often lack legitimacy with local communities, there is comparatively little research on the capacity to reduce violence from within by working with local authority figures.

In 2016, the International Rescue Committee (IRC) Research and Development team began prototyping intervention ideas for the Becoming One (B1) intervention in Liberia. The design team started off with extensive field research to better understand perceptions of different factors relating to intimate partner violence (IPV) as well as existing strategies that communities employ to address IPV. This work identified pastors and other faith leaders as particularly influential sources of authority over a couple’s behavior.

Based on feedback from pastors and the World Vision team, IRC created an initial set of couples counselling prototypes, which could be used in pre-marital counselling by faith leaders. After many iterations based on feedback from both faith leaders and couples, the team conducted a small 8 week pilot in Gulu, Uganda with 40 faith leaders and approximately 900 couples in September and October of 2017. Observations and interviews suggested that framing relationship skills as based in “biblical principles,” seemed to increase the acceptance and practice of new skills. Additionally, providing faith leaders with new counseling methods through instructional videos was an efficient way for many of the faith leaders to digest and then replicate novel approaches.


The purpose of this research protocol is to provide a rigorous assessment of the potential of the Becoming One program to improve couples’ communication and financial transparency, sexual agency for women and other markers of relationship quality.

Through our evaluation of the effectiveness of Becoming One, we also seek to contribute to understanding the causes of intimate partner violence and intrahousehold conflict more generally.

Theory of change

At the heart of the Becoming One intervention is the idea that individuals and couples will change their behavior when 1) they are already self-motivated to do so, and 2) when the source of advice or counsel is a person of social, religious, or political significance for them.

Put in the negative, Becoming One does not attempt to 1) create a motivation to change one’s relationship within couples who have not already expressed some amount of interest in change, or 2) influence couples or individuals using a source of authority external to their pre-existing social networks or community.

What theories justify this approach? Behavioral science theories suggest that behavior change is more likely when interventions seek to allow pre-existing motivations to be expressed, as opposed to when they seek to create a new motivation from scratch (e.g., Prentice & Miller, 2012). Interventions might help pre-existing motivations to be expressed by making it easier to achieve goals (through reminders, forms of social or logistical assistance), or by clarifying ways in which people can achieve goals (through simplified instructions, using culturally relevant explanations or symbols). Becoming One is based on these theoretical principles because it recruits couples who are already interested in religious couples counseling as a first step to getting married (individuals and couples have a pre-existing motivation to examine their relationship), and because Becoming One’s instruction is delivered in the language of religious text (which clarifies ways for couples to develop skills in their relationship using a familiar and revered system of belief) and in couples counseling classes with their peers (which makes it easier for couples to change with peer examples and support).

A second set of social and behavioral theories that justify the Becoming One approach posit that patterns of behavior for individuals and communities are influenced to a relatively large degree by individuals with cultural, religious, social, or political significance or authority (e.g., Rogers & Catano, 1965; Watts & Dodds, 2007). These theories identify “key” influencers as important for activating individual changes in behavior, and under some circumstances, triggering critical mass changes in behavior within communities (e.g., Centola, 2018; Paluck, Shepherd & Aronow, 2016). Becoming One is based on this broad theoretical assertion, because it operates through the influence of faith leaders, who have high status within the Ugandan communities under study, and who are regularly consulted by community members on social and religious matters, including conflicts within households.

We use these basic theories to justify the design of Becoming One, and to predict that participation in the couples counseling will change individual behavior and interactive couple behavior in four primary ways. Becoming One should:

Change couple dynamics through the Becoming One curriculum instruction and example of faith leaders, specifically:

1. Shift couple dynamics that represent zero-sum power struggles over finances, movement and general decision-making, resulting in a more equal balance of power in the couple,
2. Increase the use of skills within couples that do not relate to zero-sum power struggles, involving communication and conflict management techniques.

Reduce violence, through the Becoming One curriculum instruction and example of faith leaders, specifically:

3. reduce intimate partner violence (physical and sexual),
4. increase sexual autonomy and affirmative consent to sex within couples.

On a global level, we predict that these changes will occur because the Becoming One curriculum is delivered by a religious local influencer, the messages encouraging these behaviors will be framed and justified by religious texts, and they will be delivered within the context of peer groups. For this reason, we will collect data (outlined below) to measure the fidelity of program delivery (via program monitors) and to measure treatment participants’ reactions to religious ideas, and perceptions of the peers in their Becoming One group (via a midline and baseline survey). In the survey, we plan to measure effects on a number of potential mediators of the outcomes of Becoming one, described below, in addition to these global predictors of change. They include individuals’ gender attitudes and perceived norms of the community, men's levels of emotional regulation, greater gender-equitable interpretation of religious scripts, increased financial transparency in the couple, couples’ convergence in trust and communication, and individuals’ definitions of violence.

We also predict a number of secondary outcomes from this program related to the primary outcomes, detailed below. These include the degree of convergence in a couple regarding their beliefs about and perceptions of the relationship, a shifted distribution of domestic labor, lowered levels of emotional violence and of incidence of arguments, greater communication about sexual practices, and greater levels of time spent together; on an individual level, shifted perceptions of community norms regarding intimate partner violence, lowered levels of individual depression, and higher levels of confidence in expressing opinions publicly and in perceived social support. As a secondary outcome, we also plan to test separately the effect of Becoming One on different kinds of intimate partner violence, including emotional, physical, and sexual violence.

Faith leaders

The faith leaders recruited to deliver the Becoming One program completed one or two video-based trainings that prepared them to lead couples in sessions designed to develop practical skills for reducing IPV. During the training they received a faith leader guide, couples guides, an attendance and planning log, marketing materials, “I love my wife/husband” bracelets, and a tablet with copies of the videos and additional instructional materials in a branded bag to help them jump start and administer the program. They were also connected with other faith leaders who would be leading Becoming One sessions and were invited to join a WhatsApp group where they could share insights and learn from one another as they implemented the program.

Becoming One syllabus

The Becoming One program includes a skills-based curriculum which is meant to provide couples with practical skills for improving communication, emotional regulation, shared control over financial resources, financial planning, and sexual consent and pleasure. Sessions are meant to be participatory with the couples often asked to complete interactive activities together. Each couple is also given a guide book with additional home practice activities that they are asked to complete and report back on during subsequent sessions. The content of the program is infused with gender-equitable religious and biblical materials to inspire and reinforce concepts while also providing a relatable social frame though which to integrate them. The design materials also showcase aspirational couple identities through vivid photos of Ugandan couples.

Number of sessions

The Becoming One program is comprised of 12 sessions delivered once a week with each session lasting approximately 90 min. The groups were responsible for selecting a time and place that was convenient for them to meet. The sessions are organized into three thematic modules: communication, finance, and sex. These are preceded by an introductory session and the program concludes with a final ceremony in which the couples are recognized.
Intervention Start Date
2018-11-01
Intervention End Date
2019-02-28
Primary Outcomes
Primary Outcomes (end points)
Our main analyses focus on four outcomes that combine many theoretically-related measures:

1. Intimate Partner Violence. A binary indicator that is 1 if the female partner has experienced any of ten types of physical or sexual violence over the five months preceding the midline survey, and 0 otherwise. We expect that the ITT effect is negative and so will conduct a one-sided (lower-tailed) test.

2. Control and Decision-Making. An index of several groups of items, each of which is coded to vary between 0 and 1. The groups include decision-making about household finances and purchasing decisions, financial control, and social control. When at 0, this implies the female partner experiences all forms of financial and social control, and both partners perceive her to be excluded from all decision-making. When it is at 1, this implies the female partner does not report experiencing any forms of financial or social control, and is involved in all aspects of decision-making. We expect that the ITT effect is positive, and will conduct an upper-tailed test of the null hypothesis.

3. Sexual Consent and Autonomy. An index of items, mostly answered by the female partner but one answered by the male partner, about whether the male partner takes a non-coercive strategy when his partner refuses sex, and how the female partner feels about her control over decisions about and initiation of sex. The responses are coded to vary between 0 and 1. When the index is at 0, this implies that the female partner reports sexual coercion across all of the domains, and that the male partner reports coercive behavior. When it is at 1, this implies the absence of coercion.

4. Communication and Conflict Management Techniques. An index of items addressing a broad range of domains of communication, answered by both female and male partners. The domains include communication about sex, about daily events and companionship, and techniques for de-escalation when arguing about difficult topics. When the index is at 0, this implies that the partners agree that they have had no supportive or open communication or techniques for peaceful or de-escalating communication across many domains. When it is at 1, this implies the partners agree that they communicate with supportive or open communication or techniques for peaceful or de-escalating communication across all the domains.
Primary Outcomes (explanation)
PRIMARY OUTCOME 1: INTIMATE PARTNER VIOLENCE
Binary indicator that any one of the following is true: physical_push_5mo_w > 0; physical_slap_5mo_w > 0; physical_twist_5mo_w > 0; physical_punch_5mo_w > 0; physical_kick_5mo_w > 0; physical_choke_5mo_w > 0; physical_weapon_5mo_w > 0; sexual_forced_intercourse_5mo_w > 0; sexual_forced_other_acts_5mo_w > 0; sexual_threaten_5mo_w > 0.

PRIMARY OUTCOME 2: CONTROL AND DECISION-MAKING
Mean of the following variables, coded between 0 and 1: fin_control_w_i = 1 if "Self/Own Choice", .5 if "Give Part to Husband/Partner", 0 if "Give All To Husband/Partner";
fin_control_work_w_i = 1 if no, 0 if yes;
fin_control_take_money_w_i = 1 if no, 0 if yes;
fin_control_keep_money_w_i = 1 if no, 0 if yes;
control_friends_w_i = 1 if no, 0 if yes;
control_family_w_i = 1 if no, 0 if yes;
control_whereabouts_w_i = 1 if no, 0 if yes;
control_mobile_w_i = 1 if no, 0 if yes;
dm_earnings_resp_w_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_earnings_partners_w_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_large_purchase_w_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_health_w_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_visit_family_w_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_windfall_resp_w_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_windfall_partner_w_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_earnings_resp_m_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_earnings_partners_m_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_large_purchase_m_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_health_m_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_visit_family_m_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_windfall_resp_m_i = 1 if female partner makes or is involved in decision-making; 0 otherwise;
dm_windfall_partner_m_i = 1 if female partner makes or is involved in decision-making; 0 otherwise.
Income_r_sep_amt_m = 1 if no, 0 if yes;
control_general_w_i = 1 if I have control over all such decisions, .66 if I have control over most such decisions, .33 if I have control over some of these decisions, 0 if I don’t control any of these decisions.

PRIMARY OUTCOME 3: SEXUAL CONSENT AND AUTONOMY
Mean of the following variables, coded between 0 and 1:
no_to_sex_strategies_m_i = 1 if male partner indicates he will take a non-coercive path when refused sex by his partner (don't have sex or talk about it); 0 otherwise (persuade her, she always wants to have sex, have sex anyway);
control_sex_w_i = 1 if female partner feels she has control over all decisions about when to have sex with husband; .66 if "most such" decisions; .33 over "some of these decisions"; 0 if female partner expresses not having any control;
initiate_sex_w_i = 1 if female partner indicates always, .66 if sometimes; .33 if rarely; 0 if never;
no_to_sex_w_i = 1 if female partner feels very confident saying no to sex, .5 if somewhat confident, 0 if not at all confident;
no_to_sex_threatens_w_i = 1 if female partner indicates no, 0 if yes;
ever_pressured_w_i = 1 if female partner never feels pressured into sex, .66 if not often, .33 if sometimes, 0 if often.

PRIMARY OUTCOME 4: COMMUNICATION AND CONFLICT MANAGEMENT TECHNIQUES
Mean of the following variables, coded between 0 and 1:
disc_day_m_i = average of disc_day_p_m_i and disc_day_r_m_i, where the first concerns the male partner's perception of how often the couple discusses things in the female partner's day, and the second concerns the male partner's perception of frequency of discussing things in their own day: both sub-items coded 1 for many times, .66 for a few times, .33 for once, 0 for never;
disc_worry_m_i = average of disc_worry_p_m_i and disc_worry_r_m_i, where the first concerns the male partner's perception of how often the couple discusses the female partner's worries or feelings, and the second concerns the male partner's perception of frequency of discussing his own worries or feelings: both sub-items coded 1 for many times, .66 for a few times, .33 for once, 0 for never;
disc_sex_m_i = 1 if male partner reports discussing sexual relationship many times, .66 if a few, .33 if once, 0 if never;
disc_day_w_i = average of disc_day_p_w_i and disc_day_r_w_i, where the first concerns the female partner's perception of how often the couple discusses things in the male partner's day, and the second concerns the female partner's perception of frequency of discussing things in their own day: both sub-items coded 1 for many times, .66 for a few times, .33 for once, 0 for never;
disc_worry_w_i = average of disc_worry_p_w_i and disc_worry_r_w_i, where the first concerns the female partner's perception of how often the couple discusses the male partner's worries or feelings, and the second concerns the female partner's perception of frequency of discussing her own worries or feelings: both sub-items coded 1 for many times, .66 for a few times, .33 for once, 0 for never;
disc_sex_w_i = 1 if female partner reports discussing sexual relationship many times, .66 if a few, .33 if once, 0 if never;
comm_1_m_i = 1 if male partner reports female partner never interrupts, .66 if rarely, .33 if sometimes, 0 if always;
comm_2_m_i = 1 if male partner reports female partner always listens, .66 if sometimes, .33 if rarely, 0 if never;
comm_3_m_i = 1 if male partner reports female partner always comforts when he is having problems, .66 if sometimes, .33 if rarely, 0 if never;
comm_4_m_1 = 1 if male partner reports female partner always thanks him for things he does, .66 if sometimes, .33 if rarely, 0 if never;
comm_5_m_i = 1 if male partner reports female partner never says things that make him feel small, .66 if rarely, .33 if sometimes, 0 if always;
comm_1_w_i = 1 if female partner reports male partner never interrupts, .66 if rarely, .33 if sometimes, 0 if always;
comm_2_w_i = 1 if female partner reports male partner always listens, .66 if sometimes, .33 if rarely, 0 if never;
comm_3_w_i = 1 if female partner reports male partner always comforts when she is having problems, .66 if sometimes, .33 if rarely, 0 if never;
comm_4_w_i = 1 if female partner reports male partner always thanks her for things she does, .66 if sometimes, .33 if rarely, 0 if never;
comm_5_w_i = 1 if female partner reports male partner never says things that make her feel small, .66 if rarely, .33 if sometimes, 0 if always;
how_arg_calm_m_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner often expressed feelings in calm and respectful way, 0 otherwise;
how_arg_listened_m_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner often tried to see respondent's side of things, 0 otherwise;
how_arg_yell_m_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner never yelled, insulted or swore, 0 otherwise;
how_arg_threat_m_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner never threatened in some way, 0 otherwise;
how_arg_left_m_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner often left so that she could calm down when argument was heated, 0 otherwise;
how_arg_calm_w_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner often expressed feelings in calm and respectful way, 0 otherwise;
how_arg_listened_w_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner often tried to see respondent's side of things, 0 otherwise;
how_arg_yell_w_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner never yelled, insulted or swore, 0 otherwise;
how_arg_threat_w_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner never threatened in some way, 0 otherwise;
how_arg_left_w_i = 1 if either there is no category of arguments the respondent reports having had or they had at least one kind of argument and respondent's partner often left so that he could calm down when argument was heated, 0 otherwise;
Conflict_2_m_i = 1 if the partner disagrees strongly that they’ve held back feelings to avoid a conflict, .66 if they disagree, .33 if they agree, and 0 if they agree strongly.
Conflict_3_m_i = 1 if the partner strongly agrees they have good strategies for resolving disagreements, .66 if they agree, .33 if they disagree, and 0 if they disagree strongly.
Conflict_2_w_i = 1 if the partner disagrees strongly that they’ve held back feelings to avoid a conflict, .66 if they disagree, .33 if they agree, and 0 if they agree strongly.
Conflict_3_w_i = 1 if the partner strongly agrees they have good strategies for resolving disagreements, .66 if they agree, .33 if they disagree, and 0 if they disagree strongly.
Secondary Outcomes
Secondary Outcomes (end points)
We group secondary outcomes into two categories: those that we hypothesize are possible mediators between Becoming One and any effects on our primary outcomes, and those that we consider to be additional outcomes that participating in Becoming One may influence, but are not of primary concern.

We hypothesize that the following variables are potential mediators: Gender attitudes and norms, men’s emotional regulation, religious interpretation, financial transparency, convergence in couples’ trust and communication, and individuals’ definitions of violence

The list of secondary outcomes is:
The degree of convergence in a couple regarding their beliefs about and perceptions of the relationship, a shifted distribution of domestic labor and of perceptions of community norms regarding intimate partner violence; within the couple, lowered levels of emotional violence and of incidence of arguments, greater communication about sexual practices, higher levels of time spent together; on an individual level, shifts in perceived community norms, lowered levels of individual depression and higher levels of confidence in expressing opinions publicly and in perceived social support. As a secondary outcome, we also plan to test separately the effect of Becoming One on different kinds of intimate partner violence, including emotional, physical, and sexual violence.
Secondary Outcomes (explanation)
SEE PAP.
Experimental Design
Experimental Design
This study measures the effect of the B1 program through a block-randomized control trial.
There are three waves of measurement: a pre-randomization baseline, a midline survey approximately five weeks after the "intervention" group has completed the couples counselling program, and an endline survey approximately one year after the "intervention" group has completed the program, but before the "comparison" group has gone through couples' counselling.

Every couple in the study is invited to participate in B1. Intervention and comparison groups are formed from the three "cohorts" that go through the program. Specifically, those randomly assigned to the first cohort of B1 go through the program before the midline and endline surveys, whereas those assigned to the third cohort of B1 go through the program after the midline and endline. The design is not a typical "stepped-wedge," insofar as the outcomes of the second cohort, who go through B1 some time after the midline but before the endline, are not directly measured as part of the main experimental comparison. Rather, the creation of a second cohort makes feasible the one-year delay between the baseline survey and the eventual entry of the third cohort into B1.
Experimental Design Details
Not available
Randomization Method
The random assignment works by forming couples into pairs within congregations by matching on their baseline levels of violence, and then randomizing one couple in each pair to the intervention group (first cohort), and one couple to the comparison group (third cohort).

Baseline violence was measured as follows: first, we clean the data and impute any missing items; second, we code six binary indicators for whether the woman in the couple reports having been pushed, slapped, arm-twisted / hair-pulled, punched, kicked, or choked over the preceding 12 months; third, we take the mean of these binary indicators, and standardize it to have mean 0 and standard deviation of 1.

We form two-couple blocks within congregations using the blocktools algorithm in R. Some faith leaders were able to gather a bigger group than others, and some respondents recruited by faith leaders did not want to do the survey. However, the sample selection procedure described above ensures that each congregation has eight (n = 1), ten (n = 30), twelve (n = 77), or fourteen (n = 32) couples to randomize among (the numbers in parentheses indicate the frequencies of the different congregation sizes in the study).

We create blocks and randomize within them using the following code in R:

# Create blocks -----------------------------------------------------------
block_on <- c("prop_physical_z")
bls <- blockTools::block(
data = blwc,
n.tr = 2,
id.vars = "cup_id",
groups = "fl_id",
block.vars = block_on)
blwc$blocks <- createBlockIDs(obj = bls,data = blwc,id.var = "cup_id")

# Randomize ---------------------------------------------------------------
set.seed(5862007)
blwc$Z <- with(blwc, randomizr::block_ra(blocks = blocks, prob = .5))


Randomization Unit
Couples are the randomization unit. Main analyses are conducted at the couple-level, so they are the main unit of analysis. Some analyses are conducted at the individual level, implying clustered
Was the treatment clustered?
No
Experiment Characteristics
Sample size: planned number of clusters
1680 couples in 140 congregations
Sample size: planned number of observations
3,360 individuals
Sample size (or number of clusters) by treatment arms
840 couples in first cohort (treatment), 840 couples in third cohort (control)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Supporting Documents and Materials

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
Mildmay Uganda Research Ethics Committee
IRB Approval Date
2018-09-15
IRB Approval Number
0508-2018
IRB Name
Poverty Action Institutional Review Board
IRB Approval Date
2018-09-14
IRB Approval Number
14916
Analysis Plan
Analysis Plan Documents
B1 Midline PAP

MD5: 4adeb793c33b6847334a1fa147c42de0

SHA1: 714a5f0b2b17f65a2bb81182766f3d4845d12e9a

Uploaded At: March 11, 2019