Impact of a livelihood intervention on gender roles and relationship power among people living with HIV

Abstract

Objective:

To evaluate the impact of an agricultural livelihood intervention on gender role conflict and sexual relationship power among people living with HIV (PLHIV) in western Kenya.

Design:

Study participants were enrolled in Shamba Maisha, a cluster randomized controlled trial of an agricultural intervention conducted among PLHIV across 16 health facilities during 2016–2020. Intervention participants received a water pump, seeds, and agricultural and financial training; control participants received standard of HIV care.

Methods:

We assessed men’s views on masculinity and gender roles via the validated gender role conflict score (GRCS; range 18–78, higher=greater gender role conflict). We measured gender power imbalances among women via the validated Sexual Relationship Power Scale (SRPS), which combines subscales of relationship control and decision-making dominance (range 1–4, higher=female holds more power). We compared changes over the study period by arm using longitudinal multi-level difference-in-difference linear regression models accounting for clustering of facilities using the intention-to-treat cohort.

Results:

We enrolled 720 participants (366 intervention, 354 control); two-year retention was 94%. Median age was 40 and approximately 55% of participants were female. Among men, after 24-months the decrease in GRCS scores was 4.3 points greater in the intervention than the control arm (p<0.001). Among women, the intervention resulted in 0.25 points greater increase in the SRPS compared to the control arm (p<0.001).

Conclusions:

Shamba Maisha resulted in less gender role conflict in men and greater sexual relationship power for women. Agricultural livelihood interventions may be a powerful tool to improve gender power imbalances, which may subsequently mitigate poverty and food insecurity.

Introduction

Gender power imbalances are associated with adverse sexual health outcomes[1], unreliable contraceptive use[2], poor child health, intimate partner violence[3, 4], HIV transmission risk[5, 6], and suboptimal HIV health outcomes among women in sub-Saharan Africa. As conceptualized by Connell[7], three interdependent structures can be used to describe the gendered nature of relationships between men and women[8]: the sexual division of labor, the sexual division of power, and the structure of cathexis, or social norms around how people express their gender. Power imbalances in these structures of relationships can be exacerbated within the context of economic scarcity[9], where women’s low negotiating power and decision-making in daily matters is further diminished. Further, gender power imbalances are intricately intertwined with poor mental health among both women[10, 11] and men[12], despite the greater power men generally have relative to women. Complex internalizations of this power status and how men “should” act contribute to decreased male engagement in HIV preventative behaviors, such as condom-use and HIV testing[13], increased risk-taking as a form of demonstrating strength[14], as well as poor engagement in HIV care in Sub-Saharan Africa[15]. These internalizations result in an intergenerational cycle of poor mental, physical, emotional, and sexual well-being for both men and women.

Extensive efforts to understand how to address gender power imbalances, particularly in the global response to HIV, have concluded that gender-transformative programming can improve sexual and reproductive health for men and women alike. For example, in Uganda, SASA!, a pair-matched cluster randomized controlled trial worked with key community leaders, activists, and professionals to shift norms at a community-level around gender and power. The SASA! trial resulted in increased joint-decision making and communication as reported by men’s and women’s self-reported ability to refuse sex[16]. A systematic review of 68 studies comprising 36 randomized controlled trials found that most (90%) demonstrated evidence of behavioral and attitudinal outcomes related to gender, with 61% showing a positive effect on primary study outcomes related to gender[17]. Programs successful in promoting gender equity and shifting norms included those offered to both genders and those which were multicomponent in nature. An earlier systematic review of fifteen studies concluded that gender transformative interventions can increase sexual protective behaviors, prevent intimate partner violence, and reduce gender inequitable views[18]. This review noted that just one of the fifteen reviewed studies addressed structural factors such as employment and housing in their design[19]. Both systematic reviews note primary limitations such as insufficient follow-up time to measure the outcomes of interest and concerns about participant selection and differential attrition; taken together, these reviews highlight the need for more research.

Structural determinants of health—particularly food security and poverty—play a significant role in gender power imbalances and relationship power for men and women in lower income settings. Women play a fundamental part in household food security[20], primarily as farmers in subsistence farming settings. Yet, women have poor access to lines of credit, training[21], and supplies[22] compared to male farmers despite comprising nearly half of farmers in Africa[21, 23]. This gender gap in agriculture, which manifests in food insecurity and poverty, contributes to a vicious cycle of economic gender inequity[24]. These economic gaps in turn create household level gender imbalances, whereby women who contribute largely as subsistence farmers are often not landowners themselves, nor have the economic productivity of their male counterparts who have access to the aforementioned agricultural inputs[24]. These dynamics driven by gender equity have been well documented in western Kenya, the location of our study[25, 26], in addition to the impact of development programs on their economic well-being[27]. Multicomponent interventions can address some of these structural components of health, even in the context of HIV transmission and treatment outcomes[28]. In South Africa, the IMAGE cluster randomized trial, which incorporated a combined microfinance program with a gender and HIV prevention training curriculum, resulted in a 55% reduction in intimate partner violence, yet did not detect differences in condom-less extra-marital sexual encounters[29]. In Burkina Faso, a two-year, cluster randomized controlled trial that only enrolled women demonstrated that the impact of an integrated agriculture, skills-building, empowerment, and nutrition intervention on key child health indicators such as diarrhea and wasting[30]. These studies demonstrated that multicomponent interventions which shift gender norms via upstream determinants of gender power inequities, such as poverty alleviation and skills acquisition, may improve health outcomes. Nevertheless, all these interventions included gender programming; it is unknown whether a livelihood intervention alone can improve outcomes related to gender power imbalances.

Given the multifaceted nature of gender power imbalances and the significant role poverty and food security play in gender imbalances, promoting gender equity through food insecurity alleviation and agricultural development are crucial components to improving gender inequity and overall health[31]. Livelihood interventions that focus on upstream determinants of gender imbalances such as poverty present a promising and sustainable means to advance gender equity. Such interventions simultaneously address the multifaceted underlying structural inequities that foment gender imbalances and improve the wide-ranging health outcomes that follow[32]. This study sought to assess the impact of an agriculture and livelihood intervention, Shamba Maisha, which aimed to improve food insecurity and poverty, on gender role conflict among men and sexual relationship power among women living with HIV in western Kenya.

Methods

Trial Design and Study Sample

Data for this analysis were from the Shamba Maisha trial (NCT02815579). The pilot study preceding Shamba Maisha, as well as the methods for this study, have been described elsewhere[33, 34]. In brief, Shamba Maisha was a cluster randomized controlled trial of a multisectoral agricultural and livelihood intervention conducted across sixteen matched pair health facilities in Kisumu, Homa Bay, and Migori counties in western Kenya from 2016–2019. The study region experiences the highest incidence of HIV cases within Kenya, with Homa Bay having the highest prevalence of people living with HIV nationally at 20%[35]. Further, approximately half of residents of this largely rural region experience some form of food insecurity, despite subsistence farming being the one of the primary livelihood activities for men and women, and 40–63% residents fell within the lowest two quintiles of wealth nationally per the most recent Kenyan Demographic and Health Survey.[36]

The clinical outcomes were HIV load viral suppression (primary) and CD4 count, physical health status, WHO stage of disease, and hospitalizations (secondary outcomes). The study enrolled 720 individuals across the sixteen health care facilities who met the following inclusion criteria: 1) adults living with HIV and currently receiving ART, 2) 18 to 60 years of age, 3) belonging to a patient support group or indicating willingness to join one, 4) having access to arable land and surface water (e.g., lakes, rivers, ponds, shallow wells), 5) evidence of moderate-to-severe food insecurity based on the Household Food Insecurity Access Scale and/or malnutrition (body mass index<18.5) based on medical records during the year preceding recruitment, and 6) agreed to save the required loan down payment (no more than 2,000 KES [~$20 USD at the time]) for the water pump. Exclusion criteria were 1) inability to speak either Dholuo, KiSwahili, or English, and 2) inadequate cognitive and/or hearing capacity to complete planned study procedures, at the discretion of the research assistant. Further, the study purposively recruited women to account for at least 40% of the participants enrolled at each facility.

Intervention participants received: a loan to purchase a human-powered water pump after making a small down payment of 1000 Kenyan Shillings (~$10 at the time of study implementation); seeds and farming inputs; and agricultural and financial trainings organized by the study (more information on the training available at: https://shambamaisha.ucsf.edu/events/agriculture-training). Control and intervention participants received standard of care until the study conclusion, when the control participants were offered similar farming inputs and training. Participants were assessed at in-person and home visits semiannually over two years by field research assistants fluent in Dholuo, KiSwahili, and English. The assessments included the collection of data via household and clinic-based surveys, measurements of body mass index and mid-upper arm circumference to assess nutritional status, collection of blood to measure HIV viral load and CD4 count twice per year, and data abstraction from medical records. Participants were reimbursed for participation in accordance with local study compensation amounts and received travel reimbursements for expenses incurred for study visits; no other incentives were given.

The National Institute of Mental Health sponsored the study and had no role in the design and implementation of the trial or interpretation of the results. All study participants provided written informed consent to participate (or a thumbprint with a witness in cases where this was not possible), and all study protocols were approved by the Institutional Review Board at the University of California San Francisco and the Kenya Medical Research Institute’s Scientific and Ethical Review Unit.

Outcome Variables

Gender role conflict was assessed among male participants using the Gender Role Conflict/Stress Scale (GRCS). The GRCS is a scale adapted from the Gender Role Conflict and Masculine Gender Role Stress Scales developed by O’Neil et al[12]. While the original scales from O’Neil et al were developed for men in a Western context, the GRCS was adapted and validated for a population affected by HIV in South Africa[37]. Further, the GRCS extends the original scale by evaluating the gap between men’s perception of what an ideal man is and his own perception of whether he meets these ideals (conflict) and subsequent related stress. We adapted the responses to have four options, compared to three in the original scale, and omitted two questions (“I worry about failing and how it affects my doing well as a man” and “Making more money than a woman is a measure of my value and personal worth”) based on feedback from pilot testing; the new options ranged from strongly agree (1) to strongly disagree (4) and largely concur with the original three options which offer a range of agreement[37]. We reverse-coded the scale such that the possible range of scores was 22–88, with higher scores indicating more gender role conflict. The adapted composite score had 22-items consisting of four subscales: 1) success, power, and competition (n=7; e.g. “I strive to be more successful than others”); 2) subordination to women (n=5; e.g. “having a female boss would be difficult for me”); 3) restrictive emotionality (n=5; e.g. “I do not like to show my emotions to other people”), and 4) sexual prowess (n=5; e.g. “having sex is part of being a successful man”). For each subscale, higher scores indicate less gender equitable views, such as more restrictive emotionality (sub-scale 2). Cronbach’s alpha for the composite score was 0.76, indicating moderately high internal consistency of the GRC/S scale.

The SRPS is comprised of two sub-scales: Decision-Making Dominance and Relationship Control. Developed among a predominantly Latina and African-American population, the SRPS aims to measure relationship power dynamics and power differentials on the health outcomes of women[38]. The SRPS has since been utilized widely internationally; 2015 systematic review found that the overall scale and the Relationship Control Subscale demonstrated high reliability and validity across multiple populations in the domain of HIV/AIDS research; the Decision-Making Subscale had weaker psychometric properties[39]. Decision-Making Dominance was measured via the unmodified eight-item sub-scale of the Sexual Relationship Power Scale (SRPS).[40] The subscale assesses whether a participant or their partner has more say in decision-making in their relationship (e.g.: “Who usually has more say about what you do together?”). Response options include: (1) your partner, (2) both of you equally, or (3) you. Scores range from 8–24, with higher scores representing high decision-making power by the respondent. Both male and female participants answered the subscale questions at each visit. Cronbach’s alpha was 0.93, indicating high internal consistency of the decision-making dominance subscale.

Relationship Control was measured among women only via a 14-item subscale of the SRPS; the original scale of 15-items was modified to remove the item asking “my partner might be having sex with someone else” given polygyny is a culturally acceptable practice in our study region[40]. Each question was asked regarding a current or most recent intimate relationship and assesses women’s ability to exercise control over her relationship (e.g.: “my partner tells me who I can spend time with”) and perceptions of her relationship (e.g.: “I am more committed to our relationship than my partner is”). Women could respond that the questions were not applicable; the few women who so responded were not included and their number was non-differential by treatment assignment across all visits. The items had four Likert response options ranging from (1) strongly agree to (4) strongly disagree. Average scores, assessed per person-visit, ranged from 1–4, with higher scores indicating more relationship power. The Cronbach’s alpha was 0.87, indicating high internal consistency of the relationship control subscale.

To develop the overall SRPS per the standard scoring algorithm[40], scores for the relationship control and Decision-Making Dominance sub-scale were averaged per person-visit and rescaled to a range of 1–4, thus giving both subscales the same range. They were then combined into an overall scale with equal weighting and a final score ranging from 1–4, with higher scores indicating higher sexual relationship power. As a secondary analysis, the score was trichotomized (low, medium, and high power) as recommended by Pulerwitz el al to ease interpretability[40]. Both SRPS scores were calculated for women participants only given that men were not offered the Relationship Control sub-scale. Given that both subscales of the SRPS are sufficiently reliable to use both independently and combined[40] and may provide insight on specific sub-domains whereby the intervention was most impactful, we assessed the SRPS three ways: 1) Relationship Control sub-scale (women only); 2) Decision-Making Dominance sub-scale (men and women), and 3) the overall SRPS measure which combines the two subscales (women only) as continuous measure. Internal consistency was high for the overall SRPS (Cronbach’s alpha = 0.91).

Statistical Analysis

Summary statistics at each visit were assessed by arm. We conducted intent-to-treat analyses using three-level linear models that specified facilities, individuals, and residual (i.e., visits within individuals) as random effects and arm, visit, and their interaction as fixed effects. The inclusion of the pairs from the pair-matching did not alter results,[33] thus the more parsimonious model excluding pairs is presented. We used interaction terms to compare each outcome between arms and each follow-up visit compared to the baseline visit. A linear contrast was estimated as the interaction of arm and the trend over all visits to estimate the difference between arms in trend (i.e., change) over visits and is expressed as the trend over 24 months for each of the four outcomes. All statistical analyses were conducted using Stata 16 (College Station, TX), and data visualizations were developed in R Studio (Version 4.0.5).

Results

The Shamba Maisha study enrolled 720 participants (366 intervention, 354 control) across 16 HIV-clinic catchment areas. Two-year retention in the intervention and controls arms was 94% and 95%, respectively (Supplemental Figure 1). The median age of participants at study baseline was 40 (interquartile range 34 to 47) and approximately 55% of participants were female by study design (Table 1).

Table 1:

Baseline characteristics of Shamba Maisha study population, by study arm (n=720)

	Control (n=354)	Intervention (n=366)
	N (%) or Median (IQR)
Sex
Women	194 (54.8%)	202 (55.2%)
Age, mean (SD)	40.4 (9.3)	40.3 (8.9)
Religion
Christian	353 (99.7%)	363 (99.2%)
Muslim	1 (0.3%)	0 (0.0%)
Other	0 (0.0%)	3 (0.8%)
Marital status
Single	6 (1.7%)	12 (3.3%)
Married	251 (70.9%)	271 (74.0%)
Widowed	86 (24.3%)	75 (20.5%)
Divorced	3 (0.8%)	3 (0.8%)
Separated	8 (2.3%)	5 (1.4%)
Household size, mean (SD)	6.1 (2.7)	6.5 (2.6)
Wealth index, quintiles
Lowest	80 (22.8%)	62 (17.3%)
Second	59 (16.8%)	83 (23.1%)
Third	67 (19.1%)	76 (21.2%)
Fourth	71 (20.2%)	70 (19.5%)
Highest	74 (21.1%)	68 (18.9%)
Food security score, median (IQR)	20.0 (17.0, 24.0)	22.0 (20.0, 25.0)
Categorical food security score
Mild FI access	2 (0.6%)	1 (0.3%)
Moderate FI access	77 (21.8%)	72 (19.7%)
Severe FI access	275 (77.7%)	293 (80.1%)
Viral load, detection limit ≤200 copies/mL
Undetectable	291 (82.4%)	314 (85.8%)
Detectable	62 (17.6%)	52 (14.2%)
Years since ART initiation, median (IQR)	4.9 (2.6, 6.9)	5.1 (2.7, 7.2)

Gender Role Conflict Scale (men only)

Among men, average baseline GRCS scores were 47.6 and 47.0 in the intervention and control arms, respectively (Figure 1; by arm and visit means and standard deviations are available in supplementary material (Supplementary Table 1). At 24-months, intervention and control average endline scores were 42.2 vs. 45.5, which translates to a 4.3 points greater decrease in GRCS in the intervention than the control arm over the study period (95% CI: −5.90, −2.69; p<0.001; supplementary Table 2). For the success, power and competition subscale, there was no difference between intervention and control arms over the study period (difference in trends by arm: 0.072; 95% CI: −0.75, 0.60; p=0.83). For the subordination to women subscale, intervention participants had 2.05 points lower scores than control participants over the study period (95% CI: −2.74, −1.37; p<0.001). For the restrictive emotionality subscale, intervention participants had 1.46 points lower scores than control participants over the study period (95% CI: −2.13, −0.80; p<0.001). For the sexual prowess subscale, intervention participants had 0.75 points lower scores than control participants over the study period (95% CI: −1.30, −0.2; p=0.008) (Supplemental Figure 2).

Figure 1:

Mean Gender Role Conflict Scale scores among men by study arm and visit over the two-year study period

Decision-Making Dominance (men and women)

Among all participants, average baseline decision-making dominance scores were 15.86 and 15.59 in the intervention and control arms, respectively. At 24-months, average end line scores were 16.18 vs 15.93, with −0.10-point difference in intervention trend and control trend (95% CI: −0.74, 0.53; p=0.75). Throughout the study period the scores fluctuated. (Figure 2).

Figure 2:

Mean Decision-Making Dominance Scale scores by study arm over the two-year study period.

Relationship Control (women only)

Among women, average baseline relationship control scores were 2.51 and 2.50 for intervention and control arms, respectively. At 24-months, average endline scores were 2.89 and 2.60 for intervention and control arms, respectively, which translates to a 0.33 greater Relationship Control score in the intervention arm compared to the control arm (95% CI: 0.23, 0.43 p<0.001; Figure 3).

Figure 3:

Mean Relationship Control and Combined SRPS scores among women by study arm and visit over the two-year study period

Full (combined) SRPS (relationship control and decision-making dominance)

Among women, average baseline SRPS scores were 2.20 and 2.25 for intervention and control arms, respectively. At 24-months, average endline scores were 2.52 and 2.36 for the intervention and control arms, respectively. The intervention resulted in 0.25 points greater SRPS scores (95% CI: 0.16, 0.35; p<0.001) in the intervention compared to the control arm (Figure 3).

Discussion

This cluster randomized controlled trial of an agriculture and livelihood intervention among individuals living with HIV in western Kenya demonstrated positive impacts on several domains of gender power imbalances, including gender role conflict among men and relationship control among women over a two-year study period. Our findings are consistent with other livelihood interventions that addressed gender equity. The IMAGE cluster randomized controlled trial of loans to economically insecure women in South Africa improved household communication about sex.[29] A randomized controlled trial in Cote d’Ivoire that combined “gender dialogue groups” with a women’s economic empowerment group savings program demonstrated reduced odds of economic abuse, a dimension of household financial decision-making, among its female participants[41]. These findings are also consistent in direction and magnitude to non-livelihood and non-HIV, gender specific programming, such as an quasi-experimental study among Ethiopian men which demonstrated an increase in gender equity views[42] as well as a violence prevention study among young South African males which reported a “subtle” shift in masculinity views;[43] both of these studies used the GEMS[12]—which informed the development of the GRC/S used in this study[37]—to measure gender equity. Shamba Maisha extends this literature by including men and women as primary beneficiaries of the intervention, comprehensively measuring various domains of gender equity via validated scales, and including a substantial sample size and longer follow-up period for a novel, multisectoral livelihood intervention developed with a guiding conceptual framework outlining the causal path from study inputs to expected outcomes[44].

There are various theories regarding the structural drivers of gendered relationships with overlapping domains[45]; some prominent structural drivers of gendered relationships include social norms, food security, labor force participation, and political participation[46]. With the exception of political participation, each of these drivers were integrated into the Shamba Maisha intervention. Improving women’s relative income or employment, such as through Shamba Maisha, is a demonstrated method for improving household food security[47], which in turn may improve gendered power in relationships. This was demonstrated in the pilot study which preceded this larger trial, whereby female participants reported improved financial power and sexual decision-making, such as condom negotiations and the ability to refuse or initiate sex[32]. The enrollment of at least 40% women in each cluster translated to women’s presence in the regular trainings and meetings among the intervention arm participants and thus exposed them to the process of joint agricultural and financial decision-making. This presence challenges gendered social norms around engagement in agricultural labor force participation in our specific study setting, as such spaces for training and resource allocation in the agricultural setting are typically male=dominated[48, 49]. Among men, the regular participation of women in the trainings may have served to demonstrate that women are as capable and productive farming and business counterparts. Among women, the agricultural and business knowledge gained may have translated into more financial capital that can drive shared decision-making, respect and equity in the household, thereby improving her relationship power.

The decision-making dominance scores fluctuated during the study period. We conjecture that this may be related to the way the scale was developed and standardized. Shared decision making for household and financial matters is scored lower than sole decision making by the woman, even though it may be the most desired option, and the option most likely to result from our intervention[50]. This nuance is difficult to assess statistically per the standard scoring algorithm.

This study included important strengths such as a randomized assignment of the intervention, purposive inclusion of women, a substantial sample size, a two-year follow up period, and a very high study retention (about 94%) in both arms. Further, the study implemented several validated tools to examine different dimensions of gendered power among women and men in acknowledgement that gender inequity adversely affects all genders. These different scales allow a more nuanced understanding of how agricultural and livelihood interventions are of potential benefit to all. The Shamba Maisha intervention did not include any explicit gender transformative programming components, such as gender equity focused sessions in the training. While the intervention still impacted gender equity, the potential reach of this transformation might have been increased had specific programming been integrated. Further, understanding the potential impact of Shamba Maisha on intra-household dynamics was not possible as we did not collect data among the partners of participants; future studies may wish to collect such data among partners to evaluate the potential effect of interventions that hold promise in changing gender dynamics. A limitation of our study is its potential generalizability given that Shamba Maisha enrolled PLHIV with access to land and surface water.

Conclusion

We found that an agricultural and livelihood intervention had positive impacts on several measures of gender power among men and women living with HIV in western Kenya. These results may be generalizable to other settings where subsistence farming and gender inequity intersect. Livelihood and agricultural interventions should purposively enroll women in addition to men, as such interventions not only serve to improve food security and health, but improve gender power and equity among women. Addressing gender inequity among men and women is crucial and can improve health for both genders. Future studies might also consider livelihood interventions for populations at-risk for HIV transmission, given the role that gender power imbalances and livelihood insecurities play in HIV acquisition[51, 52]. In order to maximize impact, interventions should combine economic empowerment with educational components that explicitly address gender equity.