Effectiveness of a clinic-based counselling intervention on risk of experiencing intimate partner violence and reproductive coercion: a matched-pair cluster-controlled trial in Ebonyi and Sokoto, Nigeria

Abstract

Introduction

Despite the high rates of intimate partner violence (IPV) globally and the negative impact of IPV and reproductive coercion (RC) on reproductive and maternal health outcomes, few clinic-based interventions to address IPV and RC in clinical contexts have demonstrated effectiveness. The WHO’s LIVES (Listen, Inquire, Validate, Enhance safety, Support), and the Addressing Reproductive Coercion in the Health Setting (ARCHES) intervention have been tested with mixed effects. Our study tested the effectiveness of a combination of LIVES and ARCHES within family planning (FP) and antenatal care (ANC) services at public health facilities in Ebonyi and Sokoto, Nigeria.

Methods

The study used a quasi-experimental, matched-pair, cluster-controlled design to assess the effectiveness of LIVES and ARCHES interventions including empathetic listening and care, safety planning and referrals to other specialised services. The outcomes of experience of IPV, RC and modern contraceptive use (for FP cohort) were measured pre-intervention (baseline) and 9 months post-intervention for participants in 20 intervention clinics compared with participants in 20 comparison clinics. Multilevel models with a random intercept for health-facility and adjusted for individual level variables were used to estimate the intervention effect (time × intervention) on the outcomes.

Results

1617 clients were enrolled in the study. Comparing the intervention participants to the comparison participants, between baseline and endline, there was a relative decrease in exposure to IPV (beta=0.54 (95% CI 0.30 to 0.98)) and RC (beta=0.51(95% CI 0.28 to 0.94)) in the last 6 months. There was no difference in the modern contraceptive use comparing FP cohort participants at intervention clinics to comparison participants over time.

Conclusion

Integration of IPV and RC screening and first-line response in FP and ANC services can reduce risk of violence in women not already experiencing it as well as reduce risk of reoccurrence.

Trial registration number

NCT05331508.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• There is very little evidence on the effectiveness of clinic-based interventions in reducing risk of intimate partner violence or reproductive coercion. To date, there has been no published study that has demonstrated that an intervention comprising a brief one-session clinic-based counselling can prevent intimate partner violence (IPV) and reproductive coercion (RC) in low-resource settings. While there are similar clinic-based intervention studies that have shown reductions in IPV in high-income settings, they consisted of multiple sessions, usually three or more, and some including a mental health component. Just one published study of a similar intervention showed reductions of RC in clients disclosing IPV in family planning (FP) private clinics in the USA; and another study showed reductions of IPV in private FP clinics in Bangladesh.

WHAT THIS STUDY ADDS

• Given that most women interact with health services at some point in their lives, especially for sexual and reproductive health services, the health sector is an important entry point for early identification of women who experience IPV, provision of care, safety planning and referrals to other support services that survivors need.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The results of the study build evidence substantiating that efforts should be made to streamline integration of gender-based violence (GBV) first-line response, including empathetic counselling, safety planning, referral and education on FP methods in the context of GBV, into routine health services, as it can mitigate risk of violence to clients. Future research on client experiences as a result of the intervention is recommended to understand what aspects of the intervention helped reduce risk.

Introduction

Gender-based violence (GBV) is a global public health and clinical problem of epidemic proportions as well as a gross violation of women’s human rights and bodily autonomy. Two of the most common forms of GBV include intimate partner violence (IPV) and reproductive coercion (RC). Worldwide, an estimated one in three women have experienced physical, sexual or emotional violence by their husband/partner.¹ RC, a phenomenon that involves interference with a woman’s autonomous decision-making around contraceptive use and pregnancy decisions,² is highly correlated with IPV, particularly when such interference is meted by a partner.³ While studies are still nascent, an analysis of 2023 household survey data from women of reproductive age (15-49) in need of contraception in ten low- and middle-income countries (LMICs) found rates of RC experienced in the past year to range from 3.1% to 20.1%.⁴ In Nigeria, similar to global estimates, one in three (36%) ever-married women reported that they have ever experienced physical, sexual or emotional violence by their husband/partner.⁵ In some of Nigeria’s 33 states, rates of spousal violence are much higher; for example, in Ebonyi, 54% of women experienced physical, sexual or emotional violence by a husband or partner, while in Sokoto, the rate is 35%.⁵ Rates of acceptability are also high with 28% of women and 21% of men in the country agreeing that wife-beating is justified.⁵ Prevalence of RC among reproductive-aged women in need of contraception in Kano and Lagos, Nigeria was measured at 5% and 2.9%, respectively.⁴

Studies across countries in high-, middle- and low-income settings have demonstrated that IPV and RC are associated with poor reproductive and maternal health outcomes.6,8 Women who have experienced IPV and RC are at higher risk for unintended pregnancy due to partner refusal⁶ and partner coercion to become pregnant⁹ or to continue a pregnancy she wants to terminate.¹⁰ IPV during pregnancy is also significantly associated with higher odds of postpartum depression, low birth weight, preterm birth, maternal mortality, miscarriage, premature rupture of membranes, perinatal death and less breastfeeding in the year after birth, as found in recent reviews of research in both high-income and LMIC settings.11,13 Postpartum depression emerges as a highly correlated effect with meta-analysis showing it is three times higher in women experiencing IPV than those who do not in LMICs.¹⁴

In 2013, the WHO released guidelines calling for an improved healthcare response to IPV, calling for health providers to routinely inquire about exposure to IPV when assessing conditions that may be caused or complicated by IPV, including adverse reproductive outcomes. The guidelines also identify antenatal care (ANC) as an opportunity for routine inquiry about IPV, provided a standard protocol, trained providers, privacy, confidentiality and referral services are available, and recommend that pregnant women who disclose IPV be offered empowerment counselling and advocacy/support.¹⁵ In 2019, the guidance was updated to include the identification of RC.¹⁶ Yet, few countries are implementing these guidelines⁸ due to limited evidence and low perceived benefits of the interventions recommended in the guidelines, including routine inquiry and empowerment counselling, which compete with many other interventions in the context of overstretched health systems with limited resources.¹⁷

The latest global reviews and guidelines indicate that the current body of evidence does not demonstrate reductions in ongoing violence through screening and referral of women experiencing it.^{8 15} A handful of studies in high-income settings have found that multi-session empowerment counselling or psychological/mental health intervention in a health setting can reduce participants’ experience of IPV and improve safety-related, self-efficacy and psycho-social outcomes.^{15 18 19} However, of the few studies in LMICs that have evaluated clinic-based GBV screening and support, those demonstrating effects in reducing rates of violence are still forthcoming in the literature.²⁰

In 2021, following increased national attention to GBV due to spikes in reported violence during the COVID-19 lockdowns,²¹ Nigeria adapted the WHO clinical guidelines and handbook for responding to intimate partner and sexual violence, in collaboration with WHO Nigeria, development partners and civil society stakeholders.²² The Nigerian guidelines and accompanying training curriculum recommend routine inquiry, which prioritises specific health services as an entry point for identification of and care for GBV survivors and WHO’s recommended first-line support model, Listen, Inquire, Validate, Enhance Safety and Support (LIVES).²³

A handful of studies have evaluated interventions to address IPV and RC in the health setting, namely Addressing Reproductive Coercion in Health Setting (ARCHES), a reproductive empowerment counselling intervention that aims to help women control their contraceptive use and pregnancy decisions despite opposition from male partners or family members via education and a range of resources, including on IPV and clandestine contraceptives.^{24 25} These have had either null or mixed effects, with either having impact on RC but not IPV, or vice versa (29-31). No studies have included a robust intervention that includes both the WHO-recommended LIVES model and ARCHES.

Upon adaptation of the WHO guidelines in Nigeria, USAID’s MOMENTUM Country and Global Leadership project developed an enhanced GBV first-line response package, herein referred to as LIVES+ARCHES, that combines routine screening for IPV and RC in ANC and FP services using Nigeria’s National Guidelines on Healthcare for those who have experienced GBV, per the WHO LIVES model, and ARCHES. This paper describes results from a study to evaluate the effectiveness of LIVES+ARCHES to reduce experiences of IPV and RC, increase utilisation of modern contraception among the family planning (FP) cohort and improve safety and self-efficacy to use IPV services among ANC and FP clients in public health facilities in Ebonyi and Sokoto States in Nigeria.

Methods

Study setting and design

The effectiveness study was conducted in a total of 40 public health facilities in Ebonyi and Sokoto states in Nigeria. The selected facilities had functional ANC and FP units and provided at least three modern FP methods,²⁶ one of which was a long-acting reversible contraceptive. We conducted a quasi-experimental, matched-pair, cluster-controlled design with parallel treatment and comparison facilities assigned in a 1:1 ratio to evaluate the effectiveness of the LIVES+ARCHES intervention integrated into ANC and FP services. Enrolment occurred between April first and May 30th, 2022, and follow-up concluded February 2023. The facilities were matched based on urban/rural location and average FP and ANC clients per month. Each facility in a matched pair was assigned to group A or group B, ensuring that within a given local government administration (LGA), there were only A or only B facilities. The A group was assigned to intervention, B to comparison. The matched-pair, cluster-control approach was selected over a randomised cluster-controlled approach to ensure uniform assignment within an LGA, while balancing facility characteristics between groups. This approach helped reduce the possibility of a client being exposed to both intervention and control facilities, control for other community-based interventions that the project was implementing, and made it easier logistically for the study team. The disadvantage to this approach is that because we did not do randomisation, there could be unobserved factors that are different between the groups. The study enrolled clients into two cohorts: an ANC cohort and FP cohort.

Intervention

The intervention tested, adapted and combined LIVES and ARCHES strategies to respond to survivors’ emotional, physical, safety and support needs, help them make informed decisions to seek support and, ultimately, mitigate ongoing violence (for more details on the intervention components, see online supplemental table 1).²³ The study team developed the materials through a participatory process that included stakeholder review (Ministry of Health, State Government officials and health officers), field testing and refining the materials with clients and providers over an 8-week period. This was followed by a 3-day provider training and regular supportive supervision and mentoring visits during the study period to continue to ensure fidelity.

The definition of exposure to the intervention was attendance, at least once, at an intervention facility (ie, those that attended and were enrolled at an intervention facility). All intervention facilities had audio-visual privacy in counselling rooms, documentation materials, job aids and a directory of GBV services to which clients could be referred. In waiting rooms, posters highlighted GBV as a health issue, client rights related to GBV and the fact that clients are routinely screened for GBV in the clinic. Trained health providers routinely screened clients for physical and sexual IPV and RC using a standardised screening form. For individuals who disclosed GBV, providers offered LIVES, including safety planning and referral to other non-clinical services using a referral directory. Regardless of the disclosure of GBV, clients received empowerment counselling and were given materials on IPV, RC and FP options generally and in the context of RC both at ANC and FP clinics. In control facilities, posters that highlighted GBV as a violation of rights and advertised support for women experiencing violence in their relationships were posted.

Participants

Participants included women of reproductive age between 18 and 49 years old seeking FP or ANC services at a study facility. Other inclusion criteria included: current male sexual partner, possessing a mobile phone that could be safely used for follow-up surveys and not accompanied by a male partner or family member above the age of 5 years to their visit. If the woman was accompanied by someone older than the age of five (eg, partner, family member, older child), this may impact their likelihood to disclose experiences of IPV or RC to the provider if this person is present. Therefore, those who were accompanied were not included in the study. For FP clients, an additional inclusion criterion was that they believed they could become pregnant. Women were excluded if they did not meet the above criteria did not provide consent or had impaired cognitive abilities, which may undermine informed consent. Participants were recruited, consented and enrolled in person at the study sites while waiting for their consultation with the healthcare provider over a period of 2 months. See figure 1 for the consort diagram describing the progress of participants through the study.

Figure 1. Consort diagram. ANC, antenatal care; FP, family planning; IPV, intimate partner violence; RC, reproductive coercion; MC, Modern Contraceptives; LFUP, Lost to Follow-Up.

Data collection

There were three tools for data collection with overlap of content between the tools to allow for repeated measurements. The pre-intervention baseline and post-intervention exit survey were conducted on the day the participant was enrolled; the follow-up survey was conducted 9 months later. The pre-intervention baseline interview collected demographic information, measures of FP self-efficacy, current use of modern contraceptives (for FP clients, only), experience of physical and sexual IPV or RC, knowledge of IPV services, self-efficacy to access IPV services and safety measures. The post-intervention exit interview included the same questions on experiences of IPV and RC, as well as questions to assess whether the provider was implementing the intervention with fidelity. The experiences of IPV and RC questions were asked before and after exposure to the intervention to assess whether there were differences in outcome reporting after exposure to the intervention (ie, were women more likely to disclose experiences of IPV and RC after being asked by the provider during their visit). This process was recommended by the ARCHES study investigators, as they observed an increase in disclosure in their post-intervention survey. However, we did not see a difference in reporting of outcomes before and after the intervention (data not shown). The 9-month interview included the same questions as the pre-intervention interview and questions recalling experience at study sites over the implementation period to ascertain intervention fidelity.

Measures

The primary outcome was reported experience of physical or sexual IPV in the 6 months prior to the interview, assessed at the pre-intervention baseline interview and at the 9-month post-intervention interview. The variable is constructed using two questions adapted from the Revised Conflict Tactics Scale²⁷: ‘Has your current partner ever slapped you, punched you, hit or kicked you, or done anything else to hurt you physically?’ and ‘Has your current male partner ever forced you to have sex or do something sexual when you didn’t want to?’. If they responded ‘yes’ to ever, they were then asked if this had occurred in the last 6 months. At the follow-up interview, they were just asked in the last 6 months. The initial data collection plan included interviews at baseline and three- and 6 months follow-up. Due to a study pause at the 3-month time point, the study shifted to a singular follow-up interview at 9 months. To align with the baseline tool, the endline tool asked about exposure to the IPV and RC outcomes in the last three and 6 months, as well as in the last 9 months to cover the entire implementation period. Exposure in the last 6 months was selected as it was measured at baseline and covered most of the intervention period. A comparison of the 9-month reports of the IPV and RC outcomes was compared with the 6-month reports at endline (online supplemental table 2), but because this was not included in the baseline questionnaire, we are unable to account for differences between the groups at baseline.

Secondary outcomes include reported experience of RC; use of modern contraceptives and safety actions; and self-efficacy to access IPV services and knowledge and use of IPV services. Reported experience of RC was measured using the same process, with a self-report of ‘yes’ to either of the following questions in the last 6 months: ‘Have you ever felt pressured or forced by your current partner to become pregnant when you did not want to be?’ and ‘Has your current partner ever made it difficult for you to get family planning or to use family planning?’.

Modern contraception use among the FP cohort was defined as a responding yes to ‘did you receive a family planning method today from your provider’ and reporting that the method received was a modern contraceptive, which includes the pill, female and male sterilisation, IUD, injectables, implants, male and female condom, diaphragm and emergency contraception in a subsequent question. Self-efficacy to access IPV services included three questions on a 4-point Likert scale, reported separately for each question. Safety was measured using five questions that asked about actions taken to protect themselves or their children, answered ‘yes’ or ‘no’. A question on knowledge of GBV services was adapted from a similar study in Kenya, including the health services identified during the formative GBV services mapping done by the larger project in 2021.²⁴ A similar question on use of the health services was asked at endline only.

To assess intervention fidelity, clients were asked questions to assess the extent to which the provider administered GBV screening at baseline and endline. At the baseline exit interview, clients were asked about the healthcare visit they just attended, and at endline, they were asked whether GBV screening occurred at all, some or none of the visits they attended in the last 9 months.

Covariates

The covariates were hypothesised a priori for model inclusion included cohort (ANC or FP), age (categorised as 18–24, 25–34 and 35–49 years old), education and employment. A random intercept for facility was also included in the models (more details below).

Sample size

Sample size was calculated to detect a 10 percentage point difference in intervention and comparison clients reporting experiences of IPV and/or RC in the past 3 months, with 80% power and 30% loss to follow-up. Calculations were done for each cohort. Assumptions were that 15%–30% of clients would disclose at baseline (based on recent Nigeria Demographic and Health Survey) and a coefficient of variation for cluster sizes set to 0.1 and an intraclass correlation of 0.1% based on a previous trial.⁵ This resulted in a sample size of 440 clients per arm, 880 per cohort, for a total of 1760 clients. Following the pause, we calculated the minimum detectable difference in our primary outcome, experience of physical or sexual IPV or RC, equal to 3.2% and 4.1% for ANC or FP cohort clients, respectively. For modern contraceptive method use, an assumption of 40%–60% modern contraceptive use at baseline and the sample size of 880 FP clients allowed us to detect a 12 percentage point increase with 80% power.

Analysis

All analyses were conducted at the individual level. Descriptive statistics for continuous (ie, age) and categorical variables (t-test and χ2) were used to assess differences in client characteristics between groups at baseline. The STATA command xttrans was used to estimate transition probabilities for the outcomes over time between the two intervention groups, to estimate the probability of an outcome at follow-up, given their outcome status at baseline. Outcomes were analysed using an intent-to-treat approach. Multilevel mixed multivariable logistic models with an interaction term between the time point (pre-intervention vs 9-month follow-up) and intervention group (intervention vs comparison) were used to evaluate the effect of the intervention on experiences of physical or sexual IPV and RC. Due to high prevalence of modern contraceptive use at enrolment, we first attempted to run a multi-level mixed log binomial model, but the full model would not converge. Therefore, we used a multilevel mixed multivariable Poisson model to estimate the intervention effect. The three models included the random intercept for facility, which accounts for the nesting of clients within the facility whereby clients attending a similar facility may be more similar than those who attend another. For the IPV and RC outcomes, we also ran the models for the ANC and FP cohorts separately to assess whether effectiveness differed by service cohort.

A complete case analysis was used, following sensitivity analyses for IPV, RC and use of modern contraceptives, to confirm there were no systematic differences in those with missing data by intervention arm and individual characteristics. For the outcome on use of modern contraceptives, there were differences in missingness by individual characteristics. Therefore, we used the mi suite of commands in Stata to run a multiple imputation analysis and found this did not change the intervention effect interpretation (online supplemental table 3), so we retained the complete case analysis in the main results.

The mean and SD of the self-efficacy scores were analysed descriptively by survey and intervention group. To assess whether the change over time by group was statistically significant, an ordinal logistic model with an interaction term for survey and intervention group plus the cohort (ANC or FP) and random intercept for facility was run to estimate the p value. For each safety measure, the percentage of clients reporting taking the measure was reported by survey and intervention group and a multilevel logistic model with the same covariates as the self-efficacy model was used to estimate the p value for the intervention effect for each measure. Use of services at endline was compared using χ2 analyses. All analyses were run in Stata V.18.

Patient or general public involvement

Patients provided input into the design of communication materials that were used for the intervention. Patients/general public were not otherwise involved in the design or recruitment of the participants.

Ethical clearance

The study was approved by the Institutional Review Board at Johns Hopkins Bloomberg School of Public Health (IRB #17494) and (Ebonyi and Sokoto State Ethical Clearance Committee ref # MH/1580/V.IV). All participants completed written informed consent at baseline and the 9-month follow-up, due to the significant time lag and sensitive topic. The study was registered in clinicaltrials.gov (NCT05331508).

Results

In total, 1617 women completed both baseline and the 9-month follow-up surveys, resulting in a 92% retention rate. In both cohorts, the comparison and intervention groups were fairly similar in age and education (table 1). Due to missing data for each of the outcome models, the total number included in the model is smaller than the overall sample (See table 2 for totals by outcome). In the FP cohort, a larger proportion of women were employed in the intervention group (p<0.01).

Table 1. Participant characteristics of entire sample that completed endline survey.

Variable	ANC cohort			FP cohort
	Comparison (N=402)	Intervention (N=395)	P value	Comparison (N=416)	Intervention (N=404)	P value
Woman’s age (years) Mean (SD)	26.7 (0.28)	26.6 (0.25)	0.678	28.9 (0.30)	28.3 (0.30)	0.244
Education N (%)
None	163 (40.6%%)	147 (37.2%)	0.150	128 (30.8%)	156 (38.6%)	0.054
Primary	63 (15.6%)	71 (18.0%)		88 (21.2%)	77 (19.2%)
Secondary	155 (38.6%%)	142 (35.9%)		160 (38.5%)	126 (31.3%)
College/University	21 (5.2%)	35 (8.9%%)		40 (9.5%%)	44 (10.9%)
Woman employed N (%)	116 (28.9%)	109 (27.6%)	0.156	139 (33.4%)	94 (23.3%)	p<0.01

Table 2. Distribution of and intervention effect for (i) physical and sexual IPV, (ii) RC and (iii) current modern contraceptive use in the last 6 months, by survey and intervention group.

Physical or sexual intimate partner violence*	% Reporting outcome		Intervention effect
	Baseline N (%)	Endline N (%)	β (time × intervention) (95% CI)
Comparison facilities (N=779)	54 (6.93%)	64 (8.21%)	0.54 (0.30 to 0.98)*
Intervention facilities (N=754)	74 (9.81%)	58 (7.69%)
Reproductive coercion*			β (time × intervention) (95% CI)
Comparison facilities (N=784)	51 (6.51%)	66 (8.41%)	0.52 (0.29 to 0.96)†
Intervention facilities (N=768)	72 (9.38%)	59 (7.68%)
Current modern contraceptive use (FP cohort only)‡			β (time × intervention) (95% CI)
Comparison facilities (N=366)	259 (70.68%)	321 (87.67%)	1.04 (0.82 to 1.32)
Intervention facilities (N=372)	252 (67.92%)	327 (87.89%)

^*Covariates in model: age categories, education, employment, service group (ANC or FP) and random intercept for facility.

^†p<0.05.

^‡Covariates in model: age categories, education, employment and random intercept for facility.

ANC, antenatal care; FP, family planning; IPV, intimate partner violence; RC, reproductive coercion.

In both cohorts, at baseline and endline, a significantly greater proportion of clients at intervention sites reported the provider administering the intervention components (online supplemental table 4). At endline, intervention clients reported being screened by a provider about experiences of abuse by a partner at some (27.7%) or all (66.5%) of their visits, compared with 21.8% and 2% among comparison clients, respectively (p<0.01).

For physical or sexual IPV and RC outcomes, the percentage reporting at baseline was slightly higher in the intervention group than in the comparison group but decreased at endline for intervention clients and increased for comparison clients (table 2). The current use of modern contraceptives increased over time in both groups. The intervention effect showed a significant relative reduction in IPV and RC outcomes (p<0.05). For both IPV and RC outcomes, ANC clients were at a lower risk compared with FP clients, though this was only significant for the IPV outcomes (IPV AOR=0.63 (0.46–0.85), RC AOR=0.83 (0.62–1.12), online supplemental table 5). There was no observed intervention effect for the current use of modern contraceptives in the FP cohort comparison. For full models, including covariates and random effects (see online supplemental table 5). The cohort-specific sub-analysis of the IPV and RC outcomes showed a similar trend; the intervention effect showed a decrease in exposure to the outcomes over time comparing intervention to comparison clients in both the FP and ANC cohorts, except for the physical or sexual IPV outcome in the FP cohort, where the intervention effect is no longer statistically significant (online supplemental table 6).

The probability of transitioning from ‘no experience of IPV’ to ‘experienced IPV’ from baseline to endline is greater in the comparison (6.5%) than intervention group (3.7%). The same trend was seen for the RC outcome (comparison=6.8% vs. intervention=2.9%). However, the probability of transitioning from experiencing IPV at baseline to not experiencing it at endline was greater in the comparison group in both outcomes. In the intervention arm, the probability of transitioning to ‘no experience of IPV’ was still greater than continuing to be exposed to violence but not for RC (see table 3).

Table 3. Transition probabilities for (i) IPV and (ii) RC.

	Baseline	Comparison (n=779)		Baseline	Intervention (n=754)
		Endline			Endline
		No	Yes		No	Yes
IPV	No (n=725)	678 (93.5%)	47 (6.5%)	No (n=680)	655 (96.3%)	25 (3.7%)
	Yes (n=54)	37 (68.5%)	17 (31.5%)	Yes (n=74)	41 (55.4%)	33 (44.6%)

	Baseline	Comparison (n=784)		Baseline	Intervention (n=768)
		Endline			Endline
		No	Yes		No	Yes
RC	No (n=733)	683 (93.2%)	50 (6.8%)	No (n=696)	676 (97.1%)	20 (2.9%)
	Yes (n=51)	35 (68.6%)	16 (31.4%)	Yes (n=72)	33 (45.8%)	39 (54.2%)

IPV, intimate partner violence; RC, reproductive coercion.

The GBV service awareness varied, with Health Services being the most well-known service, followed by law enforcement and legal aid. The intervention effect for the client’s knowledge of all services (online supplemental table 7), except for shelter/accommodation, was significantly positive. Among women reporting IPV or RC at endline, a greater proportion of women in the intervention, which included referral to services, reported use of legal, psychosocial, shelter and economic reintegration services than those in the comparison arm, which did not include referral to services (table 4).

Table 4. Use of services among GBV survivors (IPV or RC) at endline, by intervention group.

Service	Comparison N (%) (n=112)	Intervention N (%) (n=84)	χ2 P value
Health services	42 (37.5%)	27 (32.7%)	0.437
Law enforcement	14 (12.5%)	14 (16.7%)	0.409
Legal aid	2 (1.8%)	8 (9.8%)	p<0.05
Psychosocial support	2 (1.8%)	17 (20.2%)	p<0.01
Shelter/temporary accommodation	3 (2.7%)	15 (17.9%)	p<0.01
Economic reintegration	0 (0.0%)	11 (13.1%)	p<0.01

GBV, gender-based violence; IPV, intimate partner violence; RC, reproductive coercion.

For the three self-efficacy questions, both groups saw increases in scores from baseline to endline (table 5). Intervention clients had significantly greater scores in thinking of a solution and developing a plan for seeking help (p<0.01). When asked, ‘Have you ever done any of the following to protect yourself or your child/children for fear that your partner would become violent’ in the last 3 months, for all safety measures except for identifying a friend or relative from whom to seek help, there was a significantly greater increase among intervention clients over time than comparison clients.

Table 5. Self-efficacy scores and safety actions taken by survey and intervention group (n=1617).

Self-efficacy (measured on 4-point Likert scale)	Intervention group	Baseline Mean (SD) (max=3)	Endline mean (SD) (max=3)	P value
If I am in trouble, I can usually think of a solution	Comparison	2.11 (0.88)	2.21 (0.76)	p<0.01
	Intervention	2.16 (0.94)	2.39 (0.84)
I am confident that I could develop a plan for seeking help	Comparison	2.03 (0.89)	2.16 (0.76)	p<0.01
	Intervention	2.19 (0.88)	2.42 (0.80)
I am confident that I would be able to execute my plan	Comparison	1.97 (0.91)	2.14 (0.79)	p=0.110
	Intervention	2.16 (0.91)	2.36 (0.87)

Safety actions (measured with Y/N)	Intervention group	Baseline (%)	Endline (%)
Identified a safe place to go in case you need to leave your home?	Comparison	66.4%	61.7%	p<0.01
	Intervention	59.7%	65.2%
Identified a friend or relative to whom you could seek help?	Comparison	69.8%	60.1%	p=0.055
	Intervention	65.3%	64.8%
Set aside some things you may need, such as clothes, documents	Comparison	38.3%	28.6%	p<0.01
	Intervention	26.4%	31.4%
Set aside funds	Comparison	37.6%	22.7%	p<0.01
	Intervention	22.8%	25.1%
Made a plan for children	Comparison	48.4%	42.3%	p<0.01
	Intervention	29.7%	50.4%

Discussion

This study demonstrated that the LIVES+ARCHES intervention can effectively reduce experience of IPV and RC and improve knowledge, self-efficacy and safety planning. To our knowledge, no other interventions comprising a brief, one-session clinic-based counselling intervention have demonstrated effectiveness in preventing both IPV and RC, nor have similar intervention studies been tested and proven effective in low-resource, public health facilities. Earlier interventions in the USA²⁸ and Mexico²⁹ that integrated IPV and RC screening, supportive care, safety planning and referrals did not prove effective in reducing IPV or RC. One study in private FP clinics in the USA that integrated RC screening and counselling into IPV screening was able to reduce risk of RC in women also reporting other forms of IPV by 71%.³⁰ An intervention study on ARCHES in private FP clinics in Kenya found that women in intervention facilities had lower odds of physical IPV at follow-up but mixed results on RC and sexual IPV.³¹ Another forthcoming publication of a study testing ARCHES in private, Non-Governmental Organization-run abortion clinics in Bangladesh found modest effects on reduced risk of IPV but no significant effects on RC.³²

Similar clinic-based IPV screening and counselling intervention studies that have shown reductions in IPV in high-income settings consisted of multiple sessions, usually three or more, with some including a mental health component,^{33 34} such as interpersonal psychotherapy to improve women’s social support.³⁵ We have no measure of how frequently women sought care during the study period and therefore are not able to determine how many times each woman in the intervention site was exposed to counselling. However, in our study, the entirety of the intervention tested was designed to be delivered in one session, and over 90% of intervention clients report the providers administering the screening steps in some or all visits during the study period.

The transitional probability analysis showed that the LIVES+ARCHES intervention was more likely to prevent IPV, both physical and sexual, as well as RC in women who had not already been exposed to it, than mitigating ongoing violence in women among GBV survivors. For women already experiencing IPV or RC at baseline, we saw a greater probability for women in comparison groups than intervention groups to move to not experiencing violence at endline, thereby making the reduced level of IPV or RC among this subset of women unattributable to our intervention, though this comparison is qualitative, and therefore we cannot say for certain there is a significant difference. Violence is not necessarily continuous, and thus, in both groups, the transition from yes to no may have been due to the relatively short 6-month recall period. As is well established, IPV does not always happen continuously in relationships, and there may be ‘honeymoon’ periods with no violence. In women not already experiencing IPV or RC at baseline, the probability of continuing to not experience violence or RC after the intervention was greater in intervention than control groups. These findings indicate that the clinic-based GBV intervention may have served more as a primary prevention intervention than a secondary prevention intervention, as it was conceived. While our study did not ascertain clients’ knowledge about GBV before and after the intervention, we hypothesise that this intervention may have been the first time the women were exposed to the concept that violence is not acceptable and, thus, responded positively to the information that they could access services to prevent it. Studies on GBV in Ebonyi³⁶ and Sokoto³⁷ suggest wide acceptance of wife-beating as justified if a woman disobeys her husband and something to be tolerated by both men and women to keep the family together. During routine supervision visits, study staff also heard from health providers that for many clients, the GBV screening and counselling made them realise for the first time that the abuse they faced by their partners was violence. Past studies also suggest that the sheer act of being asked about IPV within a healthcare setting can cause awakening and behaviour change.^{29 38} While screening for IPV has not demonstrated effectiveness on subsequent IPV in the USA or other high-income countries,^{20 39} settings where such violence is generally not socially acceptable,⁴⁰ it may have impacted utilisation of support services and experience of violence in Ebonyi and Sokoto, Nigeria, where there is a wide acceptance of violence. We posit that the intervention may have been a vehicle to intercept these norms for study participants by bringing realisation that they do not have to accept violence, particularly because it was delivered by a trusted authority. Whereas, in settings where IPV is generally not accepted, such approaches may be less relevant, as much of the research to date has shown. This aligns with the stages of change model around which a body of evidence has been collated to substantiate that women experiencing IPV move from pre-contemplation to contemplation before taking action.⁴¹

Similar to other recent studies,^{25 28 32} the LIVES+ARCHES intervention significantly increased overall knowledge of GBV services, including health, legal, psychosocial support and economic reintegration. There were few exceptions, including knowledge of shelters or intention to access law enforcement, shelters or economic re-integration services. This is not surprising given that a mapping of GBV service availability conducted by the same project team implementing the study found that there was only one shelter in Sokoto at the time of the intervention and that law enforcement and economic support services are scarcely available and of poor quality. Moreover, LIVES+ARCHES in Nigeria also significantly increased use of services, unlike most other studies,^{20 32} which we attribute to the fact that many follow-up services were offered on-site through a GBV focal point. Given the fact that 45% of women in Nigeria who have experienced violence never sought help,⁵ these findings show an important role the health sector can play in linking survivors of violence to support they may need to mitigate risk of violence.

We found increases in utilisation of modern contraceptives in both intervention and comparison groups. This may be attributable to the work of other development partners also supporting control sites to improve FP counselling and services. Uysal and colleagues found ARCHES was effective in increasing modern contraceptive use in private FP clinics in Nairobi.²⁵ Given that both intervention and control sites were private in the Uysal study, they were likely better able to control for other intervening strategies to increase FP use. Additionally, modern contraceptive use was higher at baseline than the assumptions for the sample size. Therefore, it is possible our sample size was not powered to detect differences adequately.

Strengths and limitations

Two study strengths are the relatively high retention rate of participants in the intervention, as measured by the client report of exposure to the screening intervention. However, we did not ask about exposure to every component; therefore, we do not have a complete picture of implementation. A second limitation is that the study pause meant that the 6-month recall period for the IPV and RC outcomes did not cover the entire study period. Although the intervention effect demonstrated a significant relative decrease for the 6-month time period, we were unable to assess the 9-month time period because this time frame was not asked at baseline. A key limitation is the potential of social desirability bias with respect to experiences of IPV and RC, possibly leading to under-reporting of the outcomes. A limitation for the ANC and FP cohort-specific sub-analysis is that we may not have been powered to detect a difference in intervention effectiveness. Although the sample size was achieved, the baseline levels of IPV and RC were lower than the sample size assumptions, and ultimately, we did not see the minimum detectable difference estimated post-enrollment. Additionally, increased modern contraceptive use may not have been detected due to sampling of women who were already seeking FP services and thus, likely predisposed to use modern contraceptives, as the high baseline rate of modern contraceptives showed. Furthermore, as with many real-world studies, there were other actors strengthening GBV services across facilities in Nigeria, meaning the study did not have pure comparison or intervention sites. However, given the presence of other actors in both intervention and comparison sites and efforts to match based on facility characteristics, we believe the sites to be comparable. Finally, a major limitation to our study is that it did not include a component to understand the client’s experience with the intervention, which would provide insights into why exposure to violence was reduced.

Conclusion

Integration of GBV first-line response in health services may allow providers to prevent and respond to violence that women face in their relationships. Given that most women interact with health services at some point in their lives, especially for reproductive health services, the health sector is an important entry point for early identification of women who experience GBV, provision of care, safety planning and referrals to other support services that survivors need. This study found that the integration of screening and first-line response in FP and ANC services can effectively reduce experiences of violence, particularly among those not already experiencing it, even in low-resource settings. Future efforts should be made to streamline integration of GBV screening and response into routine services throughout Nigeria. Future research on client experiences as a result of the intervention is also recommended. In particular, it would be useful to understand if the intervention acted as a secondary prevention strategy, for example, by linking women experiencing violence to services, or a primary prevention strategy, by indeed, raising awareness that abuse in relationships is violence and a violation of rights.

Data availability statement

Data are available upon reasonable request.