Global Health Action
ISSN: 1654-9716 (Print) 1654-9880 (Online) Journal homepage: https://www.tandfonline.com/loi/zgha20
Evaluation of the rural response system
intervention to prevent violence against women:
findings from a community-randomised controlled
trial in the Central Region of Ghana
Deda Ogum Alangea, Adolphina A. Addo-Lartey, Esnat D. Chirwa, Yandisa
Sikweyiya, Dorcas Coker-Appiah, Rachel Jewkes & Richard M. K. Adanu
To cite this article: Deda Ogum Alangea, Adolphina A. Addo-Lartey, Esnat D. Chirwa, Yandisa
Sikweyiya, Dorcas Coker-Appiah, Rachel Jewkes & Richard M. K. Adanu (2020) Evaluation of the
rural response system intervention to prevent violence against women: findings from a community-
randomised controlled trial in the Central Region of Ghana, Global Health Action, 13:1, 1711336,
DOI: 10.1080/16549716.2019.1711336
To link to this article:  https://doi.org/10.1080/16549716.2019.1711336
© 2020 The Author(s). Published by Informa Published online: 14 Jan 2020.
UK Limited, trading as Taylor & Francis
Group.
Submit your article to this journal Article views: 225
View related articles View Crossmark data
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=zgha20
GLOBAL HEALTH ACTION
2020, VOL. 13, 1711336
https://doi.org/10.1080/16549716.2019.1711336
ORIGINAL ARTICLE
Evaluation of the rural response system intervention to prevent violence
against women: findings from a community-randomised controlled trial in
the Central Region of Ghana
Deda Ogum Alangea a, Adolphina A. Addo-Lartey b, Esnat D. Chirwa c, Yandisa Sikweyiya c,
Dorcas Coker-Appiah d, Rachel Jewkes c and Richard M. K. Adanu a
aDepartment of Population, Family and Reproductive Health, School of Public Health, University of Ghana, Accra, Ghana; bDepartment
of Epidemiology and Disease Control, School of Public Health, University of Ghana, Accra, Ghana; cGender and Health Research Unit,
South African Medical Research Council, Pretoria, South Africa; dGender Studies and Human Rights Documentation Centre, Accra, Ghana
ABSTRACT ARTICLE HISTORY
Background: Intimate partner violence (IPV) affects one in three women globally and under- Received 27 August 2019
mines women’s human rights, social and economic development, and health, hence the need Accepted 08 December 2019
for integrated interventions involving communities in its prevention. RESPONSIBLE EDITOR
Objective: This community-randomised controlled trial evaluated the Rural Response System Jennifer Stewart Williams,
(RRS) intervention, which uses Community Based Action Teams to prevent IPV by raising Umeå University, Sweden
awareness and supporting survivors, compared to no intervention.
Methods: Two districts of the Central Region of Ghana were randomly allocated to each arm. KEYWORDS
Data were collected by repeated, randomly sampled, household surveys, conducted at baseline Intimate partner violence;
(2000 women, 2126 men) and 24 months later (2198 women, 2328 men). The analysis used a violence against women;
randomised control trial;
difference in difference (DID) approach, adjusted for age and exposure to violence in childhood. community intervention; the
Results: In intervention communities, women’s past year experience of sexual IPV reduced rural response system;
from 17.1% to 7.7% versus 9.3% to 8.0% in the control communities (DID = −9.3(95%CI; Ghana
−17.5,−1.0), p = 0.030). The prevalence of past-year physical IPV among women in the
intervention communities reduced from 16.5% to 8.3% versus 14.6% to 10.9% in the
controls (DID = −4.2(−12,3.6), p = 0.289). The prevalence of severe IPV experienced by
women reduced from 21.2% to 11.6% in intervention versus 17.3% to 11.4% in controls
(DID = −3.7(−12.5,5.1), p = 0.408). The direction of impact of the intervention on violence
perpetrated by men was more towards a reduction but changes were not statistically
significant. Emotional IPV perpetration was significantly lower (DID = −15.0(−28.5, −1.7), p
= 0.031). Women’s depression scores and reports of male partner controlling behaviour
significantly also reduced in the intervention arm compared to those in the control arm
(DID = −4.8(−8.0,−1.5), p = 0.005; DID = −2.7(−3.3,−1.0), p = 0.002, respectively).
Conclusion: Our findings indicate that the RRS intervention reduced women’s experiences of
IPV, depression, and partner controlling behaviour and some evidence of men’s reported
reductions in the perpetration of IPV. The RRS intervention warrants careful scale-up in Ghana
and further research.
Background Surveys showed that children born to women who
have been exposed to IPV experience more violence
Violence against women (VAW) is a threat to the
themselves and have poorer nutritional outcomes [2].
rights and wellbeing of women globally. The greater
Additionally, women exposed to IPV have less access
proportion of VAW is perpetrated by intimate part-
to contraception and reproductive health services,
ners. Intimate partner violence (IPV) includes emo-
skilled birth attendance and general healthcare ser-
tional, economic, physical, verbal, and sexual abuse of
vices. Such poorer access to health care, together with
women by a current or ex-husband or boyfriend. It is
a higher prevalence of morbidity and mortality,
a widely recognised phenomenon that affects over
results in a very substantial personal, household and
30% of women with widespread regional variations
wider economic burden on societies [3].
across the world. Sub-Saharan Africa currently has
Even with the high prevalence and the negative
the highest IPV burden [1].
impact of VAW, there are relatively few interventions
The consequences of VAW transcend the negative
proven to reduce VAW. A review of the evidence on
physical, mental, psychosocial and economic impact
interventions to prevent violence against women and
on the victims to impact their children and society at
girls (VAWG) carried out by Ellsberg et al. showed that
large. Evidence from 42 Demographic and Health
interventions that were more effective in reducing
CONTACT Adolphina A. Addo-Lartey aaddo-lartey@ug.edu.gh Department of Epidemiology and Disease Control, School of Public Health,
University of Ghana, P.O.Box LG 13, Legon, Accra, Ghana
© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits
unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
2 D. OGUM ALANGEA ET AL.
VAW often included participatory group sessions, pro- community for effective change to be achieved.
moting better communication, were multi-component, The overall aim of the intervention was to reduce
often engaged multiple stakeholders, and promoted the incidence of all forms of violence experienced
shared decision-making by couples. Some worked by women and violence perpetrated by men and
through community mobilisation, usually engaging protect women’s human rights through state and
men and women in communities as change agents, community-based structures. The objectives of the
and greater economic empowerment of women, with RRS are to increase knowledge on VAW; change
and without gender-transformative elements of the individual and community attitudes towards gender
intervention, often reduced VAW [4]. There were two equality and violence; positively change social and
evidence-based trials of interventions that sought to gender norms and behaviours that perpetuate gen-
work across the community through approaches that der inequality and VAW; provide counseling and
sought to change social norms on gender and violence. support to couples affected by IPV and other vic-
These evaluated the SHARE (Safe Homes and Respect tims of VAW, and assist victims to seek redress
for Everyone) intervention and the SASA! Activist Kit from state institutions; to develop a referral system
for Preventing Violence against Women and HIV. between the community-based response systems
Both were conducted in Uganda and assessed the and state agencies to encourage a consistent and
impact of IPV at the community level [5,6]. In coordinated response; and to strengthen appropri-
SHARE, there was a statistically significant decrease ate traditional systems of resolution of VAW.
in IPV experienced by women at the community These objectives of the RRS are achieved through the
level, whiles SASA! reported a reduction in sexual and institution of Community Based Action Teams
physical violence experienced by women, although (COMBATs) within the communities, trained and
reduction did not reach statistical significance. These facilitated by the Gender Centre, and supported in
trials showed the promise of this modality of interven- their work over 18 months. The COMBAT members
tion but neither achieved statistically significant are respected male and female members of the commu-
changes reported by men. nity, nominated by the community to play key roles in
In Ghana, the Gender Studies and Human Rights realising intervention goals. Six people were chosen per
Documentation Centre (GSHRDC) designed the Rural community and the study was implemented in 20 com-
Response System (RRS) [7] intervention in 2002 to munities. The teams usually worked together when
reduce violence against women and girls following find- undertaking community sensitisation and awareness-
ings of a survey on VAWG conducted in 1988 [8]. Since raising, but members could also work individually.
then (2002–2008), the RSS intervention was implemen- They were encouraged to use every opportunity, such
ted in four regions (and 18 communities) of Ghana. as community festivals and meetings, weddings, fun-
Although programme reports showed positive improve- erals, Parents Teachers Association meetings, member-
ments in the reduction of VAWG in pilot districts, no ship meetings of social associations, religious
formal impact assessment of this intervention had been groupings, and other meetings of family and friends to
conducted. As part of a global search for interventions carry out this role. They would use these opportunities,
that work to reduce VAWG, the RRS was tested in for example, to share information on wives’ property
Ghana for the first time with funding from UKAid rights after bereavement and the importance of wills, to
from the UK Government via the ‘What works to pre- talk about how to share work in the home and have
vent violence against women and girls?’ Global pro- a non-violent marriage, or to argue against child
gramme. The trial was a collaboration between the neglect. They would also provide counselling for cou-
University of Ghana, Gender Studies and Human ples referred to them because they were known to be
Rights Documentation Centre and South African experiencing violence. More detail about criteria for
Medical Research Council. This paper evaluates the COMBAT selection, training components and imple-
impact of the RRS intervention on intimate partner mentation strategy of the RRS has been previously
violence in four districts of the Central Region of Ghana. described [9]. The RRS intervention was implemented
over an 18-month period.
The RRS also provided training for staff of some
Methods State Agencies, from the police, health, social wel-
fare, Commission on Human Rights and
Intervention description – the rural response
Administrative Justice and National Commission
system (RRS)
on Civic Education. They also provided training
Given the complex interplay of societal and institu- for some other community-based organisations
tional factors in intimate partner violence which and had regular meetings with community tradi-
operate at the individual, interpersonal, commu- tional and religious leaders and other stakeholders
nity, and societal levels, the RRS works with around their roles, responsibilities, and messages in
a broad range of stakeholders within the relation to VAW.
GLOBAL HEALTH ACTION 3
Study design and setting level of 0.05 (two-sided). Factoring in 15% oversam-
pling to allow for incomplete questionnaires, the esti-
This trial was an unblinded community-randomised
mated sample size was 1640 per trial arm (820
controlled study carried out in four districts of the
currently partnered males and 820 currently partnered
Central Region of Ghana: two coastal districts (Abura
females per trial arm at baseline and at post-
and Komenda), and two inland districts (Agona and
intervention). We further oversampled by approxi-
Upper Denkyira). Our trial protocol has been registered
mately 20% to ensure that the minimum sample of
and is available on ClinicalTrials.gov (NCT03237585).
3280 participants that had to be partnered in the
Most (63%) of the region is rural and a population
past year was obtained for our primary outcome ana-
density of about 215 inhabitants per square kilometre
lysis. A total of 4,148 women (pre-intervention = 2000;
[10,11]. About 50% of adults in the region are literate
post-intervention = 2198) and 4,454 men (pre-
with highermale literacy than females (69.8% vs 46.3%).
intervention = 2,126; post-intervention = 2,328) were
The Central Region is predominantly Akan speaking
interviewed. At pre-intervention, 1877 out of 2000
(82.0%) with Fante as the indigenous dialect of most
women and 1973 out of 2126 men sampled were cur-
districts in the region. Agriculture (cocoa, oil palm,
rently partnered. At post-intervention, 1979 out of
pineapple, grains) and fishing are the primary liveli-
2198 and 2200 out of 2328 men were currently part-
hoods and employ more than two-thirds of the work-
nered. More details on participation are shown in the
force in many districts [10].
trial consort diagram in Figure 1.
Eligibility was based on residence within the com-
munity for a least a year and for men having an age
Study population and sampling
of 18 years or older, or for women, being 18-to
Districts in the Central Region within which previous -49 years old. Selected participants had to be able to
VAW research had been carried out were excluded. communicate in the main languages of the study
Two inland and two coastal districts were purposefully (English, Twi, and Fante) and not suffer from
selected from the eligible districts as study sites in order a severe mental deficit (learning difficulty, severe
to allow for at least one-district-wide geographical buf- mental illness or intoxication) that affected their abil-
fer separating the designated sites. The districts were ity to consent. At endline, participation in the study
randomly assigned intervention or control through was solely based on the eligibility of individuals
a blind draw of the four selected districts (district within randomly selected households taking no cog-
names were written on paper and placed in a bag) by nisance of prior involvement in the pre-intervention
the study statistician in Pretoria. Ten localities were survey. We did not ask if participants had partici-
randomly selected in each district resulting in a total pated in the pre-intervention survey since the inter-
of 40 localities considered as clusters in this study. Each vention was both implemented and evaluated at the
cluster (localities) contained one or more enumeration community level. Thus, all adults resident in the
areas (EAs) of varying sizes. The second layer of strati- community had equal chances of coming into contact
fication by gender was done using EAs, (resulting in 42 with intervention as well as being selected for surveys
male-designated and 38 female-designated EAs) for the upon meeting the eligibility criteria.
quantitative surveys. The stratification of the EAs was to
ensure the physical separation of male and female desig-
Data collection tools and procedures
nated EAs during data collection to ensure the safety of
women and researchers in accordance with WHO Although qualitative and quantitative methods were
safety recommendations for studies on VAW [12]. used to understand the context of VAW, estimate the
A proportionate sampling of households within EAs burden and evaluate the impact of the RRS on men
was carried out based on EA size, followed by and women in participating districts, only the quan-
a computerised random selection of households from titative components of the trial are presented here.
a list of households in each EA. The list of EAs was The survey tool consisted of questions that had
obtained from the Ghana Statistical Service but listing been validated in a similar trial in South Africa
of households was conducted by the study. Only one [14,15] and included standard measurements devel-
eligible male or female was interviewed per household. oped by the World Health Organisation for its multi-
Balloting was used to select a participant in situations country study on domestic violence against women
where more than one eligible adult was available in the [16]. Questionnaires were initially translated into
household. local dialects (Fante and Twi) by an independent
Sample size for this trial was determined using the consultant and then edited by bi-lingual members of
method of Hayes and Bennet [13] and based on IPV the project team at the University of Ghana. The
estimates (34.9% IPV experience) from the 2008 revised translations were then independently re-
Ghana DHS [11], an expected reduction of IPV by translated by another consultant who had not seen
30% in intervention arms, 90% power and significance the English version of the questionnaire. The project
4 D. OGUM ALANGEA ET AL.
4 Districts selected for the trial
2 Districts randomly allocated to each intervention arm
(Intervention arm: 1 coastal & 1 inland, Control arm : 1 coastal & 1 inland)
Intervention arm: Control arm:
20 communities randomly selected and surveyed (10 per district) 20 communities randomly selected and surveyed (10 per district)
Baseline  survey Baseline  survey
2249 Households screened for eligibility (1201 HHs for men, 1048 HHs  for 2532 Households screened for eligibility (1148 HHs for men, 1384 HHs  for 
women) women)
2004 Households successfully interviewed (1051 HHs for men, 953 HHs for 2122 Households successfully interviewed (1075 HHs for men, 1047 HHs 
women). for women).
2004 Households with complete data for analysis (1051 HHs for men, 953 2122 Households with complete data for analysis (1075 HHs for men, 1047 
HHs for women). HHs for women).
HH Not Interviewed (247: 152 HH for men, 95 HH for women) HH Not Interviewed (414: 56 HH for men; 358 HH for women) 
185 no eligible male/female in HH 222 no eligible male/female in HH
10 refusals 6 refusals
49 vacant 156 vacant
3 other reason 30 other reason
18-month intervention roll-out
Intervention arm: Control arm:
Same 20 communities  surveyed (10 per district) Same 20 communities  surveyed (10 per district)
Follow up survey (24 months) Follow up survey (24 months)
2506 Households screened for eligibility (1291HHs for men, 1215 HHs  for 2781 Households screened for eligibility (1201 HHs for men, 1580 HHs for 
women) women).
2249  Households successfully interviewed 2347  Households successfully interviewed
(1163 HHs for men, 1086 HHs for women). (1177 HHs for men, 1170 HHs for women).
2191 Households with complete data for analysis (1161 HHs for men, 1030 2335 Households with complete data for analysis (1167 HHs for men, 1168 
HHs for women). HHs for women).
HH Not Interviewed (441: 178 HH for men, 263 HH for women) HH Not Interviewed (247: 152 HH for men, 95 HH for women)
330 no eligible male/female in HH 202 no eligible male/female in HH
0 refusals 2 refusals
111 vacant 43 vacant
Figure 1. The RRS trial consort diagram.
team then used a consensus-building translation and perpetration (men); (2) prevalence in past year sex-
method to finalise the translated questionnaire ual IPV experience (women) and perpetration (men); (3)
which was pretested in a population similar to that prevalence of severe physical/sexual IPV experience
of the study population. (women) and perpetration (men). Any participant who
All datawere collected using interviewer-administered responded affirmatively to any one of the five questions
face-to-face interviews recorded onto personal digital on physical IPV or three questions on sexual IPV in the
assistants. The trial followed ethical principles outlined past 12 months was considered to have experienced IPV
by the World Medical Association Declaration of (if female) or perpetration IPV (if male). Severe IPV was
Helsinki [17], and the Belmont report [18]. Ethical assessed by combining the five physical and three sexual
approval for this trial was obtained from the IPV questions. Women were deemed to have experi-
Institutional Review Board at the Noguchi Memorial enced (and men perpetrated) severe IPV if a participant
Institute for Medical Research at the University of responded positively to two or more items, or else
Ghana (#006/15-16) and the South African Medical responded: few or many, to any single item from these
Research Council’s Ethics Committee (EC031-9/2015). eight questions.
Eligible participants were provided with informed con- The secondary outcomes included: (1) change in
sent, assured of confidentiality, anonymity, and mini- gender attitudes; (2) change in individual gender
mised risk of participation. Consenting participants attitudes (3) change in perceived social norms; (4)
(thumb printed or signed) were interviewed in their change in controlling behaviour of male partners; and
homes in their preferred local dialect. Further detail on (5) change in depression.
the methods used in this trial is presented elsewhere [9].
The Pre-intervention survey was conducted from
January till May 2016 and the post-intervention survey Statistical analysis
took place from January till May 2018. Statistical analysis used Stata version 15. The approach to
analysis was an intention to treat, thus, respondents were
included in the arm of analysis according to where they
Study outcomes
lived, whether or not they reported contact with the
The primary and secondary outcomes of the trial are intervention. Among the partnered women and men,
summarised in Appendix 1. These were assessed primary outcomes were derived from items on physical
24 months post-baseline and pre-specified in the trial or sexual IPV experience (women)/perpetration (men).
protocol [9] The primary outcomes of this trial were: (1) The participant was classified as having experienced/
prevalence in past year physical IPV experience (women) perpetrated physical IPV if they responded positively
GLOBAL HEALTH ACTION 5
(once, few times or many times) to any of the 5 items on declined participation) and 99.9% at post-
physical IPV. Similarly, a participant was classified as intervention (2 persons declined participation).
having experienced/perpetrated sexual IPV if they
responded positively to any of the 3 items on sexual
Intervention coverage
IPV. Severe IPV was assessed by combining the five
physical and three sexual IPV questions. Participants After 18 months of implementation of the interven-
were deemed to have experienced (women) or perpe- tion, nearly one-half (48%) of women and a quarter
trated (men) severe IPV if a participant responded posi- (25%) of men in the two intervention districts had
tively to two or more items, or else responded: few or heard of the intervention (the term ‘COMBAT’).
many, to any single item from these eight questions. One-fourth of women and one in ten (11%) men
For secondary outcomes measured using scales such had participated in a sensitisation activity, while
as gender attitudes, social norms, and controlling beha- 16% of women and 10% of men had received educa-
viour, we derived additive scores from the items in the tion on VAW from the information centre messages
scales after testing for internal consistency using (a megaphone broadcast by COMBATs at dawn). In
Cronbach alpha. Before deriving scores, we examined addition, one in eight (12%) women and one in
the data for missing item responses. No imputation of twenty men (5%) reported having received a home
missing data was done as there were no missing data in visit. Figure 2 shows the participation of men and
any of the items for the different scales. women in different intervention activities.
Before testing the impact of the intervention on the
outcomes at endline, we assessed for any differences in
Socio-demographic characteristics
the primary outcomes and socio-demographic factors
at baseline. Descriptive statistics were determined for all The samples of male and female participants were very
measured outcomes and summarised in tables compar- similar in their age distributions and the level of educa-
ing study arms. Since the intervention was done at tion across the two data collection points and study
community (cluster) level, we derived summaries for arms. At Pre-intervention, 85.9% and 87.4% of women
all outcomes at the cluster level (percentages for binary in the control and intervention arms were married or in
outcomes and means for continuous outcomes). To a relationship, and 85.4% and 86.3% of men in the
account for differences in primary outcomes at baseline control and intervention arms (Table 1 & Table 2). In
between arms, we used the difference in difference the post-intervention survey, 86.0% of women in the
(DID) method to assess the impact of the intervention control arm and 84.9% in the intervention, and 92.5%
at endline. The DID looks at the differences in out- of men in the control arm and 84.5% in the intervention
comes at endline between intervention and control were married or in relationships, these differences were
arms, taking into account the difference between arms significant for the men at this time point (p = 0.04).
at baseline and the change within arms from baseline to More women and men in the intervention district
endline. For all outcomes where the hypothesised direc- had worked or earned income in the past three
tion of change due to the intervention was a decrease in months compared to their counterparts in control
mean/percentage (see Appendix 1), a negative value of districts at pre-intervention (69.1% vs. 57%, p < 0.05
the model coefficient (adjusted DID) implied a better for women; 66.1% vs. 77.5% p = 0.019 for men).
outcome in the intervention arm relative to the control These types of differences were also seen for men
arm. Similarly, for all outcomes where the hypothesised post-intervention (79.7% vs. 89.5% p = 0.006 for
direction of change due to the intervention was an men). Women and men in control districts, com-
increase in mean/percentage, a positive value of the pared to intervention districts, recorded more severe
model coefficient (adjusted DID) implies a better out- household food insecurity post-intervention (26.1%
come in the intervention arm relative to the control vs. 19.7% p = 0.006 for women, 40.9% vs. 22.0%,
arm. The selection of adjustment variables was based p = 0.02 for men), this was not found pre-
on pre-determined risk factors such as childhood vio- intervention for women and for men there was
lence exposure/experience, education level and food a non-significant trend in this direction pre-
insecurity [19,20], but we also aimed at achieving par- intervention (p = 0.07). Other details of the socio-
simonious models. All models were adjusted for mean demographics of women are shown in Table 1 and of
age at the cluster level and final models for IPV were men in Table 2.
adjusted for age and exposure to violence in childhood
(witnessing the abuse of the mother in childhood).
Primary outcomes
Table 3 presents the intervention effects on the pri-
Results
mary outcomes of the study. Pre-intervention esti-
The response rate for this study was 99.7% at the pre- mates of past year sexual or physical IPV experience
intervention interview round (16 eligible participants were higher among women in intervention
6 D. OGUM ALANGEA ET AL.
48%
27%
25% 25%
16%
13%
11% 12%
10% 11%
5% 4%
Have heard of Overall Community Home visit Information One-on-one
COMBAT Participation sensitisation center chat
Women Men
Figure 2. Coverage of the RRS trial.
Table 1. Socio-demographic characteristics of female participants in the RRS trial pre- and post-intervention.
Pre-intervention Post-Intervention
Control Intervention Control Intervention
(N = 1048) (N = 952) (N = 1168) (N = 1030)
Characteristic n (%) n (%) p-value n (%) n (%) p-value
Age of respondents
<20 yrs 55 (5.2) 53 (5.6) 0.682 80 (6.8) 56 (5.4) 0.526
20–24 yrs 198 (18.9) 178 (18.7) 248 (21.2) 203 (19.7)
25–29 yrs 202 (19.3) 216 (22.7) 231 (19.8) 203 (19.7)
30–34 yrs 198 (18.9) 161 (16.9) 174 (14.9) 184 (17.9)
35–39 yrs 148 (14.1) 128 (13.4) 160 (13.7) 143 (13.9)
40–44 yrs 124 (11.8) 113 (11.9) 125 (10.7) 118 (11.5)
45–49 yrs 123 (11.7) 103 (10.8) 150 (12.8) 123 (11.9)
Marital status
Married 579 (55.2) 489 (51.4) 0.25 525 (44.9) 499 (48.4) 0.365
Separated/Divorced/No relationship 148 (14.1) 120 (12.6) 164 (14.0) 156 (15.1)
Not married but in relationship 321 (30.6) 343 (36) 479 (41.0) 375 (36.4)
Educational level
None 218 (20.8) 216 (22.7) 0.651 226 (19.3) 211 (20.5) 0.934
Primary 234 (22.3) 225 (23.6) 270 (23.1) 246 (23.9)
Junior High School 468 (44.7) 429 (45.1) 516 (44.2) 446 (43.3)
Senior High School or Higher 128 (12.2) 82 (8.6) 156 (13.4) 127 (12.3)
Worked or earned income in past 3 months 596 (57) 657 (69.1) 0.041 677 (58.0) 583 (56.6) 0.531
Years lived in the community
<5 yrs 241 (23.0) 154 (16.2) 0.018 245 (21.0) 155 (15.0) 0.018
5–9 yrs 132 (12.6) 109 (11.4) 187 (16.0) 112 (10.9)
10–19 yrs 198 (18.9) 239 (25.1) 238 (20.4) 245 (23.8)
20–29 yrs 260 (24.8) 295 (31.0) 260 (22.3) 308 (29.9)
≥30 yrs 217 (20.7) 155 (16.3) 238 (20.4) 210 (20.4)
Food insecurity
Low 314 (30.0) 237 (24.9) 0.112 574 (49.4) 452 (43.9) 0.006
Moderate 330 (31.5) 359 (37.7) 286 (24.6) 375 (36.4)
Severe 404 (38.5) 356 (37.4) 303 (26.1) 203 (19.7)
communities compared to controls (16.5% vs. 14.6% The prevalence of physical or sexual IPV perpetra-
for physical; 17.1% vs. 9.3% for sexual). Prevalence of tion by men was higher in intervention districts
physical IPV experience among women reduced from compared to control districts pre-intervention (14%
16.5% to 8.3% post-intervention while a reduction vs. 13% for physical; 18% vs. 16% for sexual). Physical
from 14.6% to 10.9% was observed among controls. IPV perpetration reduced slightly among men in
Sexual IPV experience among women also reduced intervention districts (14% to 12%) but maintained
from 17.1% to 7.7% and 9.3% to 8.0% among women in the control districts (DID = −3.6, 95%CI:-10.6–3.4,
in intervention and control districts respectively p = 0.32), Table 3. Sexual IPV perpetration increased
between intervention and controls (DID = −9.3, among men in both intervention and control districts
95%CI:-17.5–1.0, p = 0.03). Experience of severe post-intervention, although this increase was much
forms of physical or sexual IPV by women in inter- less in the intervention compared to control districts.
vention districts reduced from 21% to 12% while Although the intervention effect showed overall
control districts recorded a reduction from 17% to a reduction in IPV perpetration, no significant
11% (DID = −3.7, 95%CI:-12.5–5.1, p = 0.41). impact was observed overall (all p > 0.05).
GLOBAL HEALTH ACTION 7
Table 2. Socio-demographic characteristics of male participants in the RRS trial pre- and post-intervention.
Pre-intervention Post-intervention
Control Intervention Control Intervention
(N = 1075) (N = 1051) (N = 1167) (N = 1161)
Characteristic n (%) n (%) p-value n (%) n (%) p-value
Age of respondents
<20 yrs 40 (3.7) 40 (3.8) 0.69 49 (4.2) 85 (7.3) 0.227
20–24 yrs 141 (13.1) 142 (13.5) 164 (14.1) 179 (15.4)
25–29 yrs 170 (15.8) 154 (14.7) 191 (16.4) 172 (14.8)
30–34 yrs 135 (12.6) 125 (11.9) 161 (13.8) 154 (13.3)
35–39 yrs 141 (13.1) 115 (10.9) 165 (14.1) 134 (11.5)
40–44 yrs 110 (10.2) 114 (10.8) 108 (9.3) 111 (9.6)
45–49 yrs 81 (7.5) 81 (7.7) 82 (7.0) 77 (6.6)
≥50 yrs 257 (23.9) 280 (26.6) 247 (21.2) 249 (21.4)
Marital status
Married 597 (55.5) 674 (64.1) 0.219 636 (54.5) 628 (54.1) 0.040
Separated/Divorced/No relationship 157 (14.6) 165 (15.7) 87 (7.5) 178 (15.3)
Not married but in relationship 321 (29.9) 212 (20.2) 444 (38.0) 355 (30.6)
Educational level
None 282 (26.2) 126 (12.0) 0.118 270 (23.1) 77 (6.6) 0.059
Primary 171 (15.9) 188 (17.9) 172 (14.7) 219 (18.9)
Junior High School 283 (26.3) 406 (38.6) 365 (31.3) 445 (38.3)
Senior High School or Higher 339 (31.5) 331 (31.5) 360 (30.8) 420 (36.2)
Worked or earned income in past 3 months 605 (66.1) 690 (76.5) 0.019 798 (79.7) 917 (89.5) 0.006
Years lived in the community
<5 yrs 76 (7.1) 170 (16.2) <0.003 70 (6.0) 152 (13.1) <0.001
5–9 yrs 72 (6.7) 119 (11.3) 74 (6.3) 124 (10.7)
10–19 yrs 142 (13.2) 226 (21.5) 173 (14.8) 269 (23.2)
20–29 yrs 278 (25.9) 250 (23.8) 356 (30.5) 316 (27.2)
≥30 yrs 507 (47.2) 286 (27.2) 494 (42.3) 300 (25.8)
Food insecurity
low 365 (34.0) 422 (40.2) 0.077 384 (32.9) 582 (50.1) 0.020
moderate 197 (18.3) 284 (27.0) 306 (26.2) 324 (27.9)
severe 513 (47.7) 345 (32.8) 477 (40.9) 255 (22.0)
Table 3. Differences in primary outcomes evaluation for men and women in the RRS trial.
Baseline Baseline End-line End-line
IPV mean IPV mean
Outcome Study Arm numbers percentage numbers percentage adjusted DID (95% CI) p-value
Women’s experience
Physical IPV Control 151/989 14.6 114/1043 10.9
Intervention 139/888 16.5 89/936 8.3 −4.2 (−12.0–3.6) 0.289
Sexual IPV Control 84/989 9.3 71/1043 8.0
Intervention 138/888 17.1 85/936 7.7 −9.3 (−17.5- −1.0) 0.030
Severity of IPV (Physical or Sexual) Control 174/989 17.3 115/1043 11.4
Intervention 176/888 21.1 124/936 11.6 −3.7 (−12.5–5.1) 0.408
Men’s perpetration
Physical IPV Control 108/1008 12.6 141/1122 13.2
Intervention 127/965 13.8 131/1078 12.3 −3.6 (−10.6–3.4) 0.318
Sexual IPV Control 161/1008 15.9 222/1122 20.5
Intervention 167/965 17.6 210/1078 20.6 −5.1 (−0.16.3–6.1) 0.368
Severity of IPV (Physical or Sexual) Control 171/1008 17.6 218/1122 20.0
Intervention 201/965 21.7 240/1078 23.3 −4.1 (−15.3–7.1) 0.463
Secondary outcomes Women in the intervention arm experienced slightly
higher levels of emotional IPV compared to women in
Women in the intervention arm reported higher
the control arm (30% vs 27%) pre-intervention. Post-
levels of depressive symptoms than their counterparts
intervention, women’s experience of emotional IPV
in control districts pre-intervention (mean depression
reduced (from 30% to 22%) in the intervention arm
score 19.6 vs 17.4). Post-intervention, a significant
but was little changed in the control arm (27.3% vs.
reduction was observed in depression levels for
26.6%). There was some evidence that the intervention
women in the intervention districts compared to con-
may have had an impact on women’s experience of
trols (DID = −4.75; 95% CI:-7.98–−1.52, p = <0.01).
emotional IPV (DID = −9.6, 95% CI:-20.4- −1.2,
Similarly, men in control districts reported higher
p = 0.0.08) although this difference did not achieve
levels of depression pre-intervention compared to
conventional statistical significance. Contrarily, the
men in control districts. No reduction in depression
absolute prevalence of men’s perpetration of emotional
level among men in intervention districts versus con-
IPV was higher in both intervention and control arms
trol districts was found (DID = −0.58, 95% CI: −3.-
post-intervention than pre-intervention. However,
58–2.42, p = 0.70), Table 4.
8 D. OGUM ALANGEA ET AL.
Table 4. Differences in depression scores among respondents in the RRS trial.
Baseline End-line
Outcome Intervention Arm mean score mean score Adjusted DID p-value
Women
Depression score Control 17.36 16.85
(high = more depressed) Intervention 19.62 15.02 −4.75 (−7.98–−1.52) 0.005
Men
Depression score Control 18.91 17.60
(high = more depressed) Intervention 15.11 13.34 −0.58 (−3.58–2.42) 0.703
adjusted estimates show a significantly lower prevalence primary outcomes in women and men pointed
at endline in the intervention arm relative to the control towards a reduction, and this was statistically signifi-
arm (DID= −0.15, 95%CI:-28.5–−1.7, p = 0.03) at post- cant for sexual violence reported by women. In
intervention. women, there were also statistically significant reduc-
Women’s report of the controlling behaviour of male tions in partner controlling behaviour and less
partners pre-intervention was slightly higher in the inter- depression. In men, there was evidence of lower
vention arm compared to the control arm, Table 5. At reported perpetration of emotional/economic IPV.
post-intervention, women in the intervention commu- Although the overall direction of effect was that of
nities recorded significant reductions in partner control- lower IPV, in both intervention and control commu-
ling behaviour compared to those in the control arm nities the sexual IPV and economic and emotional
(DID = −2.66, 95% CI: −3.30–−1.02, p = 0.002). No IPV reports of men were higher at endline, but com-
significant changes were observed in men’s reports of paratively the increase was less than that among the
controlling behaviour post-intervention, Table 5. men in the intervention arm communities, and the
There was some evidence that women in the control degree to which it was less was statistically significant
arm may have had somewhat more gender-equitable for economic and emotional IPV. In interpreting this
scores compared to women in the intervention arm pre- it is important to note that this study used repeat
intervention (DID = 1.55, 95% CI:-0.26–3.36, community surveys so it was not the same men and
p = 0.094), Table 6. Social norms and individual atti- women responding at different time points.
tudes were similar among women in both arms pre- It is not clear why we have different results for
intervention. Although there were slight improvements men and women. They were interviewed in different
in gender attitudes, social norms, and individual atti- communities and were not in couples, so exactly the
tudes post-intervention, these improvements were not same results would not be expected, but we must
statistically significant. Similar to the women, there consider how the intervention impacted one gender
were no significant effects of the intervention on without as much effect on the other. It is possible that
men’s gender attitudes and perceived social norms the women gave socially desirable responses and
towards gender relations. exaggerated change, however, if that is the case it is
surprising that the change was seen in some IPV
measures (sexual violence and controlling behaviour)
Discussion
and no significant reductions in all of them. The
This trial sought to investigate the impact of the veracity of women’s reports are also supported by
Rural Response System (RRS) intervention in redu- the correlated health outcome, depression, also redu-
cing VAW and its effectiveness in enabling women to cing. An alternative explanation is that men gave
reduce their exposure to IPV and men to reduce their socially desirable responses, exaggerating their use
IPV perpetration. The direction of change in all the of violence in their efforts to be seen as tough men,
Table 5. Differences in emotional/economic IPV and partner controlling behaviour among respondents in the RRS
trial.
Pre-intervention Post-Intervention
mean percentage / mean percentage/
Outcome Study arm mean score mean score adjusted DID p-value
Women
Experience of emotional/economic IPV Control 27.3 26.6
Intervention 30.2 21.5 −9.6 (−20.4–−1.2) 0.080
Partner controlling behaviour score Control 19.66 19.56
(high = more controlling) Intervention 21.53 19.01 −2.66 (−3.30–−1.02) 0.002
Men
Perpetration of emotional/economic IPV Control 21.7 34.1
Intervention 31.9 34.3 −15.0 (−28.5- −1.7) 0.031
Controlling behaviour score (high = more Control 22.46 22.39
controlling) Intervention 21.30 21.83 0.50 (−1.22–2.23) 0.562
GLOBAL HEALTH ACTION 9
Table 6. Differences in gender attitudes and norms among respondents in the RRS trial.
Pre-intervention Post-Intervention
Outcome Intervention arm mean score mean score adjusted DID p-value
Women
Gender attitudes score Control 15.45 15.68
(high = equitable) Intervention 13.92 15.70 1.55 (−0.26–3.36) 0.094
Individual gender attitudes Control 19.54 19.79
score (high = equitable) Intervention 19.05 20.05 0.75 (−0.80–2.30) 0.337
Social norms score Control 20.04 21.16
(high = equitable) Intervention 18.47 20.17 0.58 (−1.11–2.26) 0.498
Men
Gender attitudes score Control 16.72 15.73
(high = equitable) Intervention 16.45 16.66 1.20 (−0.55–2.95) 0.177
Individual gender Control 25.20 23.85
attitudes score (high = equitable) Intervention 24.82 24.68 1.22 (−0.45–2.88) 0.152
Social norms score Control 19.96 18.49
(high = equitable) Intervention 19.90 18.29 −0.14 (−3.342–3.14) 0.932
or else that the intervention sensitised them more to time to change and it is possible that more impact
their use of violence and the baseline under-reported could have been seen if there had been a longer
it. These are both possible. Men’s levels of physical intervention [21]. It is also possible that the measures
IPV reporting at baseline were lower than women’s, of gender attitudes and social norms did not capture
although their sexual IPV reporting levels were the the full range of indicators of women’s empowerment
same. A notable change in their reporting was in in the community. The improvement in women’s
much higher levels of sexual IPV reported at endline, depression may have resulted from actual support
this was not at all similar to women’s reports. Given received from the COMBATs or the perceived socie-
the discrepancy, our inclination is to trust women’s tal support stemming from the visibility of VAW as
reports of violence experiences more than men’s as it a social issue created by the intervention. Women’s
is hard to know why women would exaggerate phy- exposure to information on how to handle cases of
sical IPV at baseline, and it is more likely that VAW and the perceived affirmation of a woman’s
women’s reports of sexual IPV are more reliable as power to seek help or redress could also have con-
they would be the ultimate arbitrators of whether tributed to a reduction in depression.
they regarded an act as forced or otherwise. The difference in men and women’s reports on
Our findings showed relatively lower engagement male controlling behaviour may be reflective of dif-
and exposure to all intervention activities among ferences in male and female perception of controlling
men, with the lowest levels of reporting for one-on- behaviour among many other reasons including feel-
one chats and home visits which allow better dialogue ings of powerlessness [22]. The reduction in women’s
on key issues compared to other sensitisation activ- report of male controlling behaviour suggests
ities. Given that exposure to the intervention is neces- increased power and assertiveness among women in
sary to effect change, further research is needed to their relationships post-intervention. Conversely, the
better understand the pathways to impact, with a goal increased controlling behaviour reported by men may
of the intervention adaptation to enhance the effect also be in response to the likely shift in power rela-
on men. This is especially important, considering that tions at home following intervention effects on
lower participation by men in similar programmes is women [22]. It is also possible that the increased
well documented (insert refs). controlling behaviour reported by men may be due
The intervention had an overall direction of posi- to increased sensitisation regarding what constitutes
tive impact on gender attitudes and norms except for controlling behaviour post-intervention.
social norms reported by men in communities that This trial had some strengths and weaknesses
were perceived as more conservative. It is likely that worth mentioning. It involved very large samples of
the latter was due to changes in perceptions of com- men and women in both pre-and – post-intervention
munities after awareness raising i.e. previously men surveys allowing for more precise estimates of out-
didn’t really reflect on how conservative their com- comes and intervention effects. Secondly, the study
munities were. It does appear that improvements in outcomes were measured using standardised tools
gender attitudes and norms were not needed to effect and thus comparable to work done in other settings.
changes in IPV and controlling behaviour experi- The ability of this trial to adjust for pre-intervention
enced by women. It is possible that the direct coun- estimates in our assessment of intervention effects on
selling of couples and messages given by traditional study outcomes makes our findings more robust. The
leaders about violence were particularly impactful in analysis of intervention effects was based on intention
the absence of evidence of overall gender attitudes to treat which is a reflection of real-life situations and
and social norm change. Social norms take much can be generalised to populations with similar
10 D. OGUM ALANGEA ET AL.
settings. A possible weakness of this study is its reli- other stakeholders who played diverse roles in supporting
ance on self-reported study outcomes that can be this trial.
accompanied by recall bias and social desirability,
this is inevitable in violence research. We did not Author contributions
interview the same women and men on the two
occasions so we cannot comment on the change in All authors contributed to the design of the study and the
individuals’ behaviour but we have the benefit of development of study tools. AAL, DOA, RMKA, and YS led
interviewer training. DOA, AAL, and RMKA co-directed
reports not being due to multiple assessments. field activities. DOA and AAL drafted the initial manuscript.
There was one important area of difference DOA, AAL, YS, EC, RJ, DCA, and RMKA critically revised
between time points and arms for women, and this the manuscript for intellectual content. All authors reviewed,
was in their levels of food insecurity. There was much edited and approved the final manuscript.
less severe food insecurity reported in the post-
intervention interviews in both study arms than in
Disclosure statement
the pre-intervention interviews. At both time points,
severe food insecurity was greater in the control arm No potential conflict of interest was reported by the
than the intervention arm. Food security was identi- authors.
fied at baseline as a risk factor for the experience of
past year IPV, after adjusting for other variables, but Ethics and consent
it was not directly correlated, in that the intervention
arm at baseline had less severe food insecurity than This trial obtained ethical approval from the Institutional
Review Board at the Noguchi Memorial Institute for Medical
the control arm and higher reported physical IPV. An Research at the University of Ghana (#006/15-16) and the
adjustment was made for food insecurity in the initial South African Medical Research Council’s Ethics Committee
analysis but was excluded in the final model due to its (EC031-9/2015). All participants were provided with and
nonsignificant effects on the estimates, and for the signed informed consent forms before participation.
purpose of achieving an efficient model.
Funding
Conclusion This trial was funded under the What Works to Prevent
Violence? A Global Programme on Violence Against
Our evaluation has shown that the Rural Response Women and Girls (VAWG), funded by the UK Aid from
System (RRS) intervention model has apparently had the UK’s Department for International Development
an impact on experience and perpetration of violence (DFID). The funding was managed by the South African
in rural communities in Central Ghana. This inter- Medical Research Council. However, the contents of this
vention worked through teams of trained and sup- publication do not necessarily reflect the views or official
policies of the UK Government. The funders played no role
ported community activists, which seems to have in the trial as well as its outputs.
critically provided support to couples experiencing
violence as well as its work to sensitise communities
to VAW. To our knowledge it is the third such Paper context
intervention to work in this way, joining the ranks Intimate partner violence is a global menace. We evaluated
of SHARE and SASA! (discussed above) that showed the Rural Response System intervention which uses com-
evidence of impact in Uganda. The findings support munity-based action teams to prevent violence against
the needs for further investment in delivering this women and intimate partner violence in the Central
intervention to rural communities in Ghana and sug- Region, Ghana. Our findings indicate that the intervention
reduced women’s experiences of intimate partner violence,
gest that much value could be gained from further depression, and partner controlling behavior, and some
research aiming to better understand the observed evidence of men’s reported reductions in the perpetration
disparities in intervention effects on men and of intimate partner violence. The intervention warrants
women and the pathways to change, as well as its careful scale-up in Ghana and further research.
impact in the context of scale-up.
ORCID
Acknowledgments Deda Ogum Alangea http://orcid.org/0000-0001-7496-
8514
We are grateful to all our study participants for their Adolphina A. Addo-Lartey http://orcid.org/0000-0001-
voluntary participation in the trial and agreeing to share 8852-7189
their experiences with us. We are also thankful to the Esnat D. Chirwa http://orcid.org/0000-0003-0471-4978
intervention implementation team from the Gender Yandisa Sikweyiya http://orcid.org/0000-0001-7153-
Studies and Human Rights Documentation Centre, the 5828
COMBAT members, Traditional and Local authorities of Dorcas Coker-Appiah http://orcid.org/0000-0002-6409-
participating Districts, our field research staff as well as 1330
GLOBAL HEALTH ACTION 11
Rachel Jewkes http://orcid.org/0000-0002-4330-6267 [15] Gibbs A, Washington L, Willan S, et al. The Stepping
Richard M. K. Adanu http://orcid.org/0000-0001-8053- Stones and Creating Futures intervention to prevent
6793 intimate partner violence and HIV-risk behaviours in
Durban, South Africa: study protocol for a cluster
randomized control trial, and baseline characteristics.
References BMC Public Health. 2017;17:336.
[16] García-Moreno C, Jansen H, Ellsberg M, et al. WHO
[1] Devries KM, Mak JY, García-Moreno C, et al. The multi-country study on women’s health and domestic
global prevalence of intimate partner violence against violence against women. Vol. 204. Geneva: World
women. Science. 2013;340:1527–1528. Health Organization; 2005. p. 1–18.
[2] Chai J, Fink G, Kaaya S, et al. Association between [17] World Medical Association. World Medical
intimate partner violence and poor child growth: Association declaration of Helsinki. Ethical principles
results from 42 demographic and health surveys. Bull for medical research involving human subjects. Bull
World Health Organ. 2016;94:331. World Health Organ. 2001;79:373.
[3] Klugman J, Hanmer L, Twigg S, et al. Voice and [18] Department of Health, Education. The Belmont
agency: empowering women and girls for shared pros- report. Ethical principles and guidelines for the pro-
perity. Washington, DC: World Bank Group; 2014. tection of human subjects of research. J Am Coll Dent.
[4] Ellsberg M, Arango DJ, Morton M, et al. Prevention of 2014;81:4.
violence against women and girls: what does the evi- [19] Chirwa ED, Sikweyiya Y, Addo-Lartey AA, et al.
dence say? Lancet. 2015;385:1555–1566. Prevalence and risk factors of physical or sexual inti-
[5] Wagman JA, Gray RH, Campbell JC, et al. Effectiveness of mate violence perpetration amongst men in four dis-
an integrated intimate partner violence and HIV preven- tricts in the Central Region of Ghana: baseline
tion intervention in Rakai, Uganda: analysis of an inter- findings from a cluster randomised controlled trial.
vention in an existing cluster randomised cohort. Lancet PloS One. 2018;13:e0191663.
Glob Health. 2015;3:e23–e33. [20] Alangea DO, Addo-Lartey AA, Sikweyiya Y, et al.
[6] Abramsky T, Devries K, Kiss L, et al. Findings from the Prevalence and risk factors of intimate partner vio-
SASA! Study: a cluster randomized controlled trial to lence among women in four districts of the Central
assess the impact of a community mobilization inter- Region of Ghana: baseline findings from a cluster
vention to prevent violence against women and reduce randomised controlled trial. PloS One. 2018;13:
HIV risk in Kampala, Uganda. BMCMed. 2014;12:122. e0200874.
[7] Coker-Appiah D. A guide to developing a community [21] World Health Organization. Changing cultural and
response to violence against women in Ghana. Accra, social norms that support violence. Geneva,
Ghana: Gender Studies and Human Rights Centre; Switzerland: WHO Press, World Health
1999 Organization; 2009.
[8] Coker-Appiah D, Cusack K. Breaking the silence and [22] Petrik ND, Olson REP, Subotnik LS. Powerlessness
challenging the myths of violence against women and and the need to control: theMale Abuser’s Dilemma.
children in Ghana: report of a national study on violence. J Interpers Violence. 1994;9:278–285.
Accra, Ghana: Gender Studies and Human Rights [23] WHO. Global and regional estimates of violence
Documentation Centre; 1999. against women: prevalence and health effects of inti-
[9] Addo-Lartey AA, Ogum Alangea D, Sikweyiya Y, et al. mate partner violence and non-partner sexual vio-
Rural Response System to prevent violence against lence. Geneva: WHO; 2013.
women: methodology for a community randomised [24] Radloff LS. The CES-D scale a self-report depression
controlled trial in the Central Region of Ghana. Glob scale for research in the general population. Appl
Health Action. 2019;12:1612604. Psychol Meas. 1977;1:385–401.
[10] Ghana Statistical Service. 2010 population & housing [25] Pulerwitz J, Gortmaker SL, DeJong W. Measuring
census: summary report of final results. Accra: Ghana sexual relationship power in HIV/STD research. Sex
Statistical Services; 2012 Roles. 2000;42:637–660.
[11] Ghana Statistical Service (GSS), Ghana Health Service [26] Jewkes R, Nduna M, Levin J, et al. Impact of Stepping
(GHS), ICF Macro. Ghana demographic and health Stones on incidence of HIV and HSV-2 and sexual
survey 2008. Accra, Ghana: GSS, GHS, and ICF behaviour in rural South Africa: cluster randomised
Macro;2009. controlled trial. BMJ. 2008;337:a506–a506. PubMed
[12] World Health Organization. Putting women first: PMID: ISI:000258669200026.
ethical and safety recommendations for research on [27] Jewkes R, Gibbs A, Jama-Shai N, et al. Stepping
domestic violence against women. Geneva: World Stones and Creating Futures intervention: shor-
Health Organization; 2001. tened interrupted time series evaluation of
[13] Hayes R, Bennett S. Simple sample size calculation a behavioural and structural health promotion
for cluster-randomized trials. Int J Epidemiol. and violence prevention intervention for young
1999;28:319–326. people in informal settlements in Durban, South
[14] Jewkes R, Levin J, Penn-Kekana L. Risk factors for Africa. BMC Public Health. 2014;14:1325. PubMed
domestic violence: findings from a South African PMID: 25544716; PubMed Central PMCID:
cross-sectional study. Soc Sci Med. 2002;55:1603–1617. PMCPMC4320600.
12 D. OGUM ALANGEA ET AL.
Appendix 1. Outcome measures in the RSS trial
Number of Method of Cluster-level Hypothesised direction of change
Outcome Source Indicator items scaling summary because of the intervention
Primary outcomes
Physical IPV WHO VAW scale (adapted) [23] One or more episode of physical IPV in the past 12 months (men perpetrate; 5 Binary Proportion Decrease
women experience)
Sexual IPV One or more episode of sexual IPV in the past 12 months (men perpetrate; women 3 Binary Proportion Decrease
experience)
Severe IPV More than one episode of physical or sexual intimate partner violence in the past 8 Binary Proportion Decrease
12 months (men perpetrate; women experience)
Secondary outcomes
Depressive Centre for Epidemiologic Studies Depression in the past week, continuous 20 Mean Mean Decrease
symptomology Depression (CESD) scale [24]
(CESD)
Controlling Modified Sexual Relationship Power Experiences of control in primary relationships by a male partner (men perpetrate; 8 Mean Mean Decrease
Behaviours (SRP) scale [25] women experience)
Emotional/economic WHO VAW scale (adapted) [23] One or more episodes of emotional IPV in the past 12 months (men perpetrate; 5 Binary Proportion Decrease
IPV women experience)
Gender attitudes Modified Gender Equitable Men’s Scale Agreement with statements on gender attitudes (4 point Likert scale, ≥ more 8 Mean Mean Increase
(GEMS) [26] equitable)
Individual gender Scale developed for Stepping Stones Agreement with statements on gender attitudes (4 point Likert scale, ≥ more 9 Mean Mean Increase
attitudes Creating Futures study [27] equitable)
Social norms Scale developed for Stepping Stones Agreement with statements on gender attitudes (4 point Likert scale, ≥ more 9 Mean Mean Increase
Creating Futures study [27] equitable)