A livelihood intervention improved nutritional, mental health and sexual health outcomes among HIV-affected adolescents in western Kenya: results from the Shamba Maisha randomised controlled trial

Abstract

Introduction

Interventions that address household-level factors, including food insecurity (FI) and poverty, may reduce HIV risk and improve sexual and reproductive health (SRH) among adolescent girls and young women (AGYW).

Methods

This cluster randomised controlled trial in Kenya assessed AGYW living in households of adults enrolled in Shamba Maisha (SM). SM was a multisectoral agricultural livelihood intervention designed to improve HIV health. At the trial end, we used linear and logistic regressions to compare differences between the intervention and control arms in FI, and mental health among AGYW-caregiver pairs and SRH outcomes among the AGYW, accounting for clusters with robust standard errors. We also conducted a secondary analysis among 15–19 years whose caregivers were SM participants.

Results

The study enrolled 241 AGYW-caregiver pairs (n=131 intervention). The median age of AGYW was 15 years old (IQR: 14–17 years old). At endline, AGYW in intervention households had less FI (OR: 0.45, 95% CI 0.21 to 0.97; p=0.04), lower depressive symptomatology (OR: 0.53; 95% CI 0.24 to 1.15, p=0.11) and no difference in unprotected sex (OR: 1.16; 95% CI 0.63 to 2.16; p=0.63) compared with those in control households. In the secondary analysis, intervention AGYW had higher body mass index (β: 1.25; SE: 0.32; p=0.002), fewer symptoms of depression (OR: 0.42; 95% CI 0.21 to 0.84; p=0.01) and anxiety (OR: 0.23, 95% CI 0.05 to 1.11; p=0.07), and less sexual intimate partner violence (IPV) (OR: 0.28; 95% CI 0.08 to 0.91; p=0.03) than control AGYW. Intervention arm caregivers had less FI than those in the control arm (OR: 0.05; 95% CI 0.0047 to 0.54; p=0.014).

Conclusions

An agricultural livelihood intervention among adults living with HIV improved FI, mental health and IPV outcomes among AGYW in their households. These findings will guide the scale-up of an agricultural livelihood intervention for HIV-affected households with AGYW in sub-Saharan Africa.

Trial registration number

NCT03741634.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• In sub-Saharan Africa, food insecurity (FI) and poverty increase the vulnerability of adolescent girls and young women (AGYW) to HIV infection and worsen their sexual and reproductive health (SRH). Household-level interventions that address underlying structural risk factors may be more effective than individual-focused interventions at improving adolescent girls’ SRH and reducing their HIV risk. Limited research has examined multicomponent interventions implemented at the household level, especially those with combined FI and livelihood components.

WHAT THIS STUDY ADDS

• This study found that a multisectoral agricultural livelihood intervention, consisting of a loan to purchase farming implements, including a human-powered water pump, and education in sustainable farming and financial management, improved FI, mental health and intimate partner violence outcomes among adolescents living in the households of adults enrolled in the intervention, which are important pathways towards improved HIV and SRH outcomes.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our findings support the potential importance of incorporating multicomponent, household-level interventions that address underlying structural risk factors into HIV and other prevention efforts for AGYW. Additional research is needed on this promising topic to reduce the disproportionate burden of HIV and poor SRH among AGYW.

Introduction

Adolescent girls and young women (AGYW), aged 15–24 years, are disproportionately at risk of acquiring HIV in sub-Saharan Africa (SSA). Women and girls accounted for 63% of all new HIV infections in SSA in 2021, with around 4000 AGYW acquiring an infection weekly.¹ Girls account for six out of seven new infections among adolescents aged 15–19, and AGYW are three times more likely to be living with HIV compared with similarly aged young men.¹ Evidence-based HIV prevention interventions, including providing access to condoms and daily oral HIV pre-exposure prophylaxis, HIV risk reduction counselling and routine testing for HIV and sexually transmitted infections (STIs), have not meaningfully reduced HIV incidence among AGYW at the population level in many countries, including Kenya.² These clinic-based interventions are unable to address upstream factors that shape the risk context for AGYW, highlighting an opportunity for additional multipronged HIV prevention approaches that meet the needs of AGYW and thereby support UNAIDS targets to end the HIV epidemic by 2030.³

The gender disparity in HIV vulnerability among adolescents is strongly shaped by household and structural factors, chief among them food insecurity (FI), poverty and gender inequality.^{4 5} Food-insecure and impoverished AGYW are more likely to be disempowered and to have worse sexual and reproductive health (SRH) outcomes, including higher rates of early sexual debut, inconsistent condom use, transactional sex, forced sex, STIs, HIV and unintended pregnancies.46,11 Living in an HIV-affected household is also associated with HIV-related and other health risks; adolescents that have lost one or both parents to AIDS or reside with an adult caregiver with HIV/AIDS are more likely to experience poverty,¹² sexual risks¹³ and poor psychological well-being.¹⁴ Poverty and FI are both highly prevalent in SSA, with an estimated 490 million people living on less than US$1.90 a day,¹⁵ and over 60% of the population facing moderate or severe constraints on their ability to obtain sufficient food.¹⁶ A growing FI gap by gender highlights that this situation is especially acute among women and girls, particularly those from poor households and those living with HIV and in HIV-affected households.^{17 18} This body of research reinforces that upstream risk drivers, such as poverty and FI, are important targets of intervention to improve SRH and other health outcomes among AGYW.

Recent systematic reviews and studies have demonstrated the need for a shift towards multilevel interventions and away from individual-centred interventions, given that the latter has a limited ability to change the underlying context that shapes vulnerability and risk.19,22 For adolescents, intervening on the risk context may be especially effective at the household level since they may have little power over their decisions on how to use household resources.²³ Yet, few multicomponent interventions to improve adolescent girls’ SRH and reduce their HIV risk have been implemented at the household level, and limited research has examined interventions with combined FI and livelihood components. Many structural interventions with economic components among adolescent girls have combined skills training with economic empowerment, with mixed results. While some showed positive impacts on SRH-related outcomes, including reduced sexual risk-taking intentions,^{24 25} HIV infection,²⁶ unprotected sex,^{27 28} unintended pregnancy and transactional sex,²⁷ and others showed no impact on sexual risk intentions²⁹ or on HIV/STI infections.^{30 31} Some also reported adverse effects, including increased sexual coercion and physical abuse.^{32 33} This highlights the potential hazards of promoting individual asset building with AGYW in precarious economic environments without considering the supports they may need from their household and community. Structural interventions that stimulate income growth at the household level may therefore be ideally suited to support vulnerable AGYW and allow them to realise the full benefits of multicomponent interventions.

Since FI in many regions of SSA is shaped by the combined effects of extreme poverty, environmental change and insufficient agricultural output and distribution, agricultural interventions hold particular promise as a way to improve household FI and wealth.^{34 35} This may be especially so as climate change impacts become more widespread and severe in SSA and increase the risk of substantial agricultural and livelihood disruptions among already vulnerable populations, chiefly the women who constitute almost 50% of the agricultural labour force and largely work as smallholder farmers.³⁶ To our knowledge, the impacts of an agricultural and livelihood intervention on SRH risk-related outcomes among AGYW have not yet been evaluated.

Our study was a cluster randomised control trial (RCT) that evaluated the effects of Shamba Maisha (SM), a climate-adaptive agricultural livelihood intervention aimed at improving HIV-related health outcomes among adults living with HIV in western Kenya. We examine whether this intervention improved the FI and mental health of AGYW-caregiver pairs and the SRH of the AGYW living in SM households. We hypothesised that AGYW living in households participating in the intervention would experience improvements in FI, mental health and SRH.

Methods

Study setting and design

In Kenya, over 40% of new HIV infections are among 15–24 year oldss, with especially high rates among females.³⁷ The SM trial took place in Kisumu, Homa Bay and Migori counties in western Kenya, where the adult HIV prevalence is among the highest in the country.³⁸ This region is predominantly rural, where subsistence farming and fishing are the primary livelihoods, and 65% of the population lives below the poverty line.^{39 40} An estimated 70% or more of PLHIV are moderately or severely food-insecure.^{41 42} Rain-fed agriculture is a key economic driver and livelihood activity in Kenya, and changes in precipitation are affecting poverty and FI levels.⁴³

The methods of the SM trial have been extensively described elsewhere.⁴⁴ Briefly, sixteen healthcare facility catchment areas were randomised to intervention and control arms at baseline. PLHIV aged 18 years and older who were on ART and had access to farmland and surface water were enrolled in the study, and they and their households were followed biannually for 2 years. The SM intervention consisted of (1) a market interest loan (~US$175) from a Kenyan bank for purchasing agricultural implements and commodities, including a human-powered water pump (Super MoneyMaker, Kickstart International, Nairobi, Kenya) and (2) education in financial management and sustainable farming practices. Control arm participants were offered a similar intervention at the conclusion of the trial.

The current study, Adolescent Shamba Maisha (Adolescent SM), was a sub-study of AGYW living in the households of SM participants and who lived in facility catchment areas that were randomised to either the intervention or control arm. The Adolescent SM study is registered on ClinicalTrials.gov (NCT03741634), as well as the SM trial (NCT02815579).

Study framework

Our evidence-based framework (figure 1) was built on our previously published conceptual framework,^{17 45} which was adapted for adolescents and their caregivers using existing theoretical models for understanding the linkages between adolescent HIV risk, food security and socioeconomic well-being. These linkages include psychosocial, empowerment, educational and caregiver pathways. Those living in HIV/AIDS-affected households in SSA may have few vocational skills and resources and high rates of FI and poverty.⁴⁶ Caregivers of adolescents may experience poor physical and mental health, since FI has been associated with a range of adverse clinical effects among persons living with HIV, including worse immunologic, virologic and mental health outcomes.47,58 Parenting practices, such as communication with children and showing affection or closeness, may also be suboptimal due to these stressors.⁵⁹ Adolescents in these households may have poor mental health outcomes, low empowerment (including low self-efficacy, self-esteem) and low educational attainment. These adolescent and caregiver factors, in turn, can contribute to risky sexual behaviours among adolescent girls and lead to subsequent high risk of STIs, HIV and unintended pregnancies.^{8 9 60} This context further cements adolescents and their households into cycles of poverty, FI and poor health.

Figure 1. Intervention framework. HSV-2, herpes simplex virus type 2; SRH, sexual and reproductive health.

Our SM intervention framework thus posits that a multicomponent agricultural livelihood intervention will first improve household food security and wealth, which will, in turn, improve adolescent mental health, school attendance, empowerment and self-esteem, as well as caregiver physical health and parenting practices. Improvements in these adolescent and caregiver pathways will then contribute to decreased unprotected and transactional sex among adolescents and reduced risk of STIs, HIV and unintended pregnancies.

Study population

To determine eligibility for the Adolescent SM study, we reviewed household census data, collected from index (ie, enrolled) SM participants at baseline, to identify all households with AGYW in the eligible age range. Prior to the endline interview for the SM trial, field research assistants contacted index participants to explain the adolescent study and to gain verbal consent for contacting the AGYW residing in their households as well as from the AGYW’s primary caregiver (if it was not the index participant). We defined the primary caregiver as the closest caregiver figure to the AGYW who was responsible for regularly looking after her well-being. This caregiver did not necessarily have to be her parent, nor the index participant in the SM trial.

The eligible age of AGYW was guided by the WHO’s definition of adolescence as those below age 19 years. Eligible participants were any AGYW aged 13–19 years living in the household of an active SM participant who were HIV negative via self-report at the time of SM trial initiation, of adequate cognition to participate, willing to assent (if under age 18) or consent (age 18–19) to participate, able to speak English, Dholuo or Kiswahili, currently unmarried, and who had a caregiver willing to participate. AGYW were not eligible if they had perinatally acquired HIV or an otherwise confirmed HIV diagnosis before the start of the SM trial, would not be aged 13–19 years at study end line, were currently married, were designated heads of households, or did not have adequate cognition to assent or consent. Caregiver eligibility criteria included being the primary caregiver of the AGYW, being eighteen years of age or older, and being willing and able to consent to participate.

In the case of multiple eligible AGYWs in the same household, we randomly selected one AGYW for enrolment in Adolescent SM. Informed consent was gained from all caregivers and AGYW above age 18; for AGYW below age 18, informed assent with the consent of the caregiver was obtained. Study recruitment and enrolment information is illustrated in figure 2.

Figure 2. CONSORT. CONSORT, Consolidated Standards of Reporting Trials.

Data collection

We collected data from each AGYW and her caregiver within 6 weeks before or after the endline study visit of their household’s index SM participant. All visits with the AGYW occurred at healthcare facilities in private locations separate from their caregivers, given the sensitive nature of study topics. Each AGYW participant completed a single survey with a trilingual (English/Dholuo/Swahili) female data collector that included anthropometric measures, demographic information, a food security assessment and questions about mental health, self-esteem, resilience, intimate partner violence (IPV), caregiver relationships and sexual health. We also collected blood tests for the herpes simplex virus type 2 (HSV-2), HIV (with standard pre-and post-test counselling by a Ministry of Health-trained HIV counsellor) and urine pregnancy tests for all AGYW. Structured interviews with caregiver participants took place either at their home or a healthcare facility and included questions on demographics, food security, mental health, communication about sexual health with their child, and caregiver-child conflict. Each AGYW and caregiver participant received 500 Kenyan Shillings (~US$4) at the conclusion of their data collection visit, plus an additional travel compensation of up to 500 Kenyan shillings, determined by distance travelled to the healthcare facility.

Study outcomes and measures

Primary outcomes

The primary outcomes for AGYW participants were FI, depressive symptoms and unprotected sex. FI was measured via the nine-item Household Food Insecurity Access Scale (HFIAS),⁶¹ adapted to ask about the adolescent participant rather than household, with responses ranging from 0 to 3 and a composite score of 0–27; higher scores indicate more severe FI. The HFIAS score was coded dichotomously (food insecure vs food secure) using a standard scoring algorithm.⁶¹ Depressive symptoms were measured via the 15-item Hopkins Symptom Checklist,⁶² and scores were coded dichotomously per standard scoring; row averages of less than 1.75 were categorised as not having depressive symptoms.⁶³ Unprotected sex was operationalised as the percentage of time that condoms were not used during penetrative sexual activity in the prior 6 months.

Secondary outcomes

Secondary outcomes for AGYW participants included sexual IPV, new pregnancies, HIV and HSV-2 seropositivity, body mass index (BMI), physical health, anxiety, resilience, sexual partners in the prior 6 months, and school absenteeism. Questions assessing lifetime experience of sexual IPV were adapted from the WHO Multi-country Study on domestic violence.^{64 65} Answers were scored as 0, never; 1, once; 2, two or three times; and 3, four or more times; scores were coded dichotomously as ever vs never. Current pregnancy was determined by a positive urine HCG test taken during the data collection visit, and the London Measure of Unplanned Pregnancy⁶⁶ was used to assess whether the pregnancy was unintended. Specimens for the HSV-2 tests were collected via blood draw during the data collection visit and were double tested (time-updated and end-of-study batched) with the Kalon Assay (Kalon HSV-2 IgG ELISA, Kalon Biological, U.K.) at a regional lab. AGYW were sent to get HIV tests completed and processed at the national lab. Weight and height were triple measured at the data collection visit and combined to report BMI in kilograms per metre squared. Physical health was assessed with the physical functioning subscale of the Medical Outcomes Study Short form SF-36 (SF-36)⁶⁷ (range 0–100), with higher scores indicative of better health. Anxiety was measured via the Generalised Anxiety Disorder-7 (GAD-7) scale⁶⁸ (range 0–27), with scores of 10 and above indicating symptoms consistent with moderate GAD. Resilience was measured using the validated Child and Youth Resilience (CYRM) measure,⁶⁹ and higher CYRM scores indicated greater resilience. Sexual partners in the prior 6 months were reported as a total number, and school absenteeism was operationalised as the number of school days missed during the previous school term.

Secondary outcomes for caregiver participants included FI (adapted also to ask about the adult caregiver and not the household), anxiety, depressive symptoms, general health and well-being, communication related to sexual health, and caregiver–child conflict. FI, anxiety and depressive symptoms were measured and operationalised using the same scales and coding as those used with AGYW participants. Health and well-being were assessed using eight subscales of the SF-36⁶⁷: physical functioning, bodily pain, role limitations due to physical health problems, role limitations due to personal or emotional problems, emotional well-being, social functioning, energy/fatigue and general health perceptions; subscale scores ranged from 1 to 100, and higher scores indicated better health for that subscale. Caregiver communication related to sexual health was assessed using several subscales from the Families Matter! Programme toolkit⁶⁹ that had been recently adapted to the Kenyan context: sex education communication (range 0–12, higher meaning more communication with child); discussing risk reduction (range 0–12, higher meaning discussed more with child); and responsiveness to being asked about sexual health topics (range 0–12; higher meaning more responsive). The Cronbach’s alpha for these subscales was 0.86, 0.84 and 0.85, respectively. The measure for caregiver–child conflict was also adapted from the Families Matter! Program toolkit⁷⁰ (Cronbach’s alpha of 0.73) (range 0–16, higher meaning less conflict).

Statistical analysis

Sample size

Given that the parent SM trial did not account for the number of adolescents in their recruitment, our Adolescent SM study was not specifically powered to detect the effects of household intervention assignment on adolescent health biomarkers. The data collected in the current study provide estimates of intra-cluster correlation coefficients (ICC) and effect sizes necessary for planning a subsequent larger RCT. To evaluate power for our primary behavioural outcomes, we used NCSS PASS V.15 to compute the minimum detectable standardised effect for the regression of (1) continuous intermediate outcomes (ie, FI and sexual risk composite score) and (2) health outcomes, on intervention group assignment. The effective sample size (ESS) input was computed via the design-based correction 1+(m-1)*ICC, where m was the expected number of adolescents per cluster. We estimated m to be approximately 17, and the ICC for FI is 0.10, thus ESS was 76. Assuming ESS=76, α=0.05, power=0.80, and five additional covariates with a multiple R of 0.30, the minimum detectable R² for the regression of continuous intermediate pathway variables and health outcomes on intervention assignment is 8.7%, which is between small (2%) and medium (13%) effect size benchmarks.

Due to the nature of the intervention, neither participants nor study research staff could be blinded to intervention assignment. Laboratory staff, data management staff, all investigators and the study biostatistician remained blinded to study arm until after the database was frozen for analysis. Loss to follow-up was <5%, and the amount of missingness was trivial (<5%) and thus listwise deletion was used.

We compared demographic characteristics of the households with AGYW using baseline data collected during the parent SM trial. We conducted linear and logistic regressions comparing differences between intervention and control arms on the multiple outcomes and controlling for clusters with robust standard errors. Our primary analysis was the intention-to-treat sample. Given that the average age of sexual debut in Kenya is 15 years old, we conducted a secondary set of analyses restricting the sample to those aged 15 years and older and whose caregiver was the index SM participant; this restricted sample theoretically had higher potential to benefit from the intervention. Baseline data (at onset of trial) was not available on the adolescent girls and was only available for household characteristics.

Results

We collected data from 241 AGYW-caregiver pairs (N=131 (54%) intervention) (figure 2) between 28 November 2018 and 5 December 2019. The median age of AGYW participants was 15 years (IQR: 14, 17; table 1). Most AGYW (92%) were still enrolled in school at the time of the study. One-third of the AGYW reported to be either casually dating or in a serious relationship, and three had been previously married, though none were currently married, per eligibility criteria. Approximately 71% of caregivers were index SM participants, and thus by parent study inclusion criteria, living with HIV, and most caregivers were women (88%). Household characteristics at baseline showed covariate balance among primary sociodemographic characteristics between intervention and control groups.

Table 1. Characteristics of households with adolescent girls and young women at Shamba Maisha baseline.

	Control (N=110)	Intervention (N=131)
Age, median (IQR)	42.0 (34.0, 47.0)	39.0 (34.0, 44.0)
Household size, median (IQR)	7.4 (2.7)	7.5 (2.3)
Marital status of index participant, N (%)
Single	2 (1.8)	4 (3.1)
Married	82 (74.5)	100 (76.3)
Widowed	26 (23.6)	25 (19.1)
Separated	0 (0.0)	2 (1.5)
Polygamous household	30 (27.3)	39 (29.8)
Highest educational attainment of index participant, N (%)
None	0 (0.0)	1 (0.8)
Some primary	38 (34.5)	59 (45.0)
Primary	42 (38.2)	34 (26.0)
Some secondary	14 (12.7)	15 (11.5)
Secondary	11 (10.0)	13 (9.9)
Some college/other	5 (6.9)	9 (4.6)
Household food insecurity (FI) status*, N (%)
Mildly FI	1 (0.9)	0 (0.0)
Moderately FI	19 (17.3)	18 (13.7)
Severely FI	90 (81.8)	113 (86.3)
Depression score† of index participant, median (IQR)	1.3 (1.2, 1.8)	1.8 (1.3, 2.1)
Household wealth score‡, median (IQR)	3.0 (2.0, 4.0)	3.0 (2.0, 4.0)

^*The degree of insecure food access in the household in the past 30 days, measured using the Household Food Insecurity Access Scale.

^†Measured using the Hopkins Symptom Checklist; median scores of <1.75 were categorised as not having depressive symptoms.

^‡Quintile of household wealth score that was created using principal components analysis of assets applying the method of the Demographic and Health Surveys.

AGYW living in intervention households had 0.45 the odds of reporting FI at SM endline compared with those in control households (95% CI 0.21 to 0.97; p=0.04) (table 2). The intervention arm had lower depressive symptomatology (OR: 0.53; 95% CI 0.24 to 1.15, p=0.11) in the full sample, with no difference in unprotected sex (OR: 1.16; 95% CI 0.63 to 2.16; p=0.63). In analyses restricted to n=154 girls aged 15–19 whose caregiver was the index SM participant, the odds of having depressive symptoms were lower in the intervention versus control arm (OR: 0.42; 95% CI 0.21 to 0.84; p=0.01). We did not detect a difference in the odds of FI (OR: 0.52; 95% CI 0.15 to 1.79, p=0.30) or unprotected sex (OR: 0.85; 95% CI 0.43 to 1.66, p=0.63) in the restricted sample.

Table 2. Impact of the Shamba Maisha intervention on study outcomes among adolescent girls and young women.

	Control (N=110)	Intervention (N=131)	Model 1: Full sample (N=241)			Model 2: Restricted to those aged 15–19 years and whose caregiver was index participant (N=154)
			OR/β	95% CI/SE	P value	OR/β	95% CI/SE	P value
Primary outcomes, N (%)
Food insecurity (ref: secure)	99 (90.0)	105 (80.2)	0.45	0.21 to 0.97	0.04	0.52	0.15 to 1.79	0.30
Depressive symptoms (ref: none)	19 (17.3)	13 (9.9)	0.53	0.24 to 1.15	0.11	0.42	0.21 to 0.84	0.01
Unprotected sex (ref: none)	27 (24.6)	36 (27.5)	1.16	0.63 to 2.16	0.63	0.85	0.43 to 1.66	0.63
Secondary outcomes, N (%)
Pregnancy (ref: negative)	6 (5.5)	10 (7.6)	1.42	0.41 to 4.94	0.58	0.60	0.16 to 2.33	0.46
HIV positive (ref: negative)	2 (1.9)	7 (5.3)	2.94	0.63 to 13.67	0.17	1.49	0.25 to 8.75	0.66
HSV-2 positive (ref: negative)	10 (9.1)	18 (13.8)	1.61	0.55 to 4.74	0.39	1.21	0.38 to 3.87	0.75
BMI, median (IQR)	21.0 (18.8, 22.8)	21.0 (18.6, 22.9)	−1.43	(1.28)	0.28	1.25	(0.32)	0.0002
# Sexual partners in prior 6 months, median (IQR)	1.0 (1.0, 1.0)	1.0 (1.0, 1.0)	−0.058	(0.065)	0.39	−0.18	(0.10)	0.10
Other outcomes
Anxiety symptoms (ref: none), N (%)	7 (6.4)	3 (2.3)	0.34	0.07 to 1.67	0.19	0.23	0.05 to 1.11	0.07
Child Youth Resilience score*, median (IQR)	48.0 (44.0, 52.0)	51.0 (47.0, 54.0)	1.76	(0.96)	0.09	1.41	(1.29)	0.29
# School days missed in prior term, median (IQR)	1.0 (0.0, 4.0)	0 (0.0, 2.0)	0.46	(0.69)	0.51	0.69	(1.05)	0.52
Sexual IPV† (ref: none), N (%)	14 (16.9)	9 (8.6)	0.47	0.19 to 1.17	0.10	0.28	0.08 to 0.91	0.03

^*Higher scores mean more resilience.

^†Among those in a relationship.

BMI, body mass index; HSV-2, herpes simplex virus-2; IPV, intimate partner violence; β, beta coefficient.

Secondary and other outcomes

The intervention effect on sexual IPV in the full sample was OR: 0.47 (95% CI 0.19 to 1.17; p=0.10) and was greater in the restricted sample, with the odds of experiencing sexual IPV almost four times lower in the intervention arm compared with the control arm (OR: 0.28; 95% CI 0.08 to 0.96; p=0.03). We detected neither differences in pregnancy (OR: 1.42, 95% CI 0.41 to 4.94; p=0.58), HIV (OR: 2.94; 95% CI 0.63 to 13.67; p=0.17) nor in HSV outcomes (OR: 1.61; 95% CI 0.55 to 4.74; p=0.39) (table 2) by study arm. These findings were consistent when using the restricted sample. Likewise, there were no detectable intervention effects for BMI (β: −1.43; SE: 1.28; p=0.28), the number of sexual partners in the prior 6 months (β: −0.058; SE: 0.065; p=0.39), or school absenteeism (β: 0.46; SE: 0.69; p=0.51). In the restricted sample, there was evidence of intervention impact on BMI (β: 1.25; SE: 0.32; p=0.002), with girls in the intervention arm having a higher BMI.

While we did not detect effects of the intervention on anxiety in the full sample, there was evidence for an impact in the restricted sample (OR: 0.23; 95% CI 0.05 to 1.11; p=0.07). There was evidence for intervention impact on youth resilience (β: 1.76; SE: 0.96; p=0.09) but not for reducing school absenteeism (β: 0.46, SE: 0.69, p=0.51); there were no effects on these outcomes in the restricted sample.

Caregiver outcomes

Caregivers of AGYW in intervention households had a reduced odds of FI (OR: 0.05, 95% CI 0.0047 to 0.54; p=0.02) (table 3). We did not detect intervention effects on anxiety (OR: 0.70, 95% CI 0.19 to 2.60; p=0.53) or depressive symptoms (OR: 0.72; 95% CI 0.28 to 1.89; p=0.51). Median scores for the subscales of the SF-36 were consistently above the midpoint of 50; no differences were detected by arm in any of the subscales. While we did not detect a difference in the attitudes towards sex education, sex education communication topics or sexual risk reduction subscales of the Families Matter! Subscales, we did detect a difference in responsiveness to being asked about sex (β: −0.78, SE: 0.34, p=0.03); we did not detect differences in the caregiver–child conflict score.

Table 3. Impact of the Shamba Maisha intervention on study outcomes among caregivers.

	Control (N=110)	Intervention (N=131)	Full sample (N=241)
			OR/β	95% CI/SE	P value
Food insecurity (ref: secure), %	99.0	83.7	0.05	0.0047 to 0.54	0.014
Anxiety symptoms (ref: none), %	10.6	7.7	0.70	0.19 to 2.60	0.60
Depressive symptoms (ref: none), %	22.1	17.1	0.72	0.28 to 1.89	0.51
Health and well-being indicators,* median (IQR)
Physical functioning	100.0 (90.0, 100.0)	95.0 (80.0, 100.0)	−10.30	(9.05)	0.27
Role limitations due to physical health	100.0 (50.0, 100.0)	100.0 (0.0, 100.0)	−11.28	(6.82)	0.12
Role limitations due to emotional health	100.0 (33.3, 100.0)	100.0 (0.0, 100.0)	−10.50	(8.54)	0.24
Energy/fatigue	75.0 (60.0, 85.0)	75.0 (55.0, 85.0)	−0.93	(3.45)	0.79
Emotional well-being	76.0 (64.0, 88.0)	76.0 (60.0, 88.0)	−1.44	(2.91)	0.63
Social functioning	53.0 (53.0, 75.0)	53.0 (53.0, 62.5)	−4.82	(3.51)	0.19
Pain	90.0 (67.5, 100.0)	90.0 (55.0, 100.0)	−4.24	(5.47)	0.45
General health	75.0 (65.0, 85.0)	75.0 (60.0, 85.0)	−1.97	(3.52)	0.58
Caregiver communication about sexual health,† median (IQR)
Attitudes towards sexual education‡	6.0 (5.0, 7.0)	7.0 (5.0, 7.0)	0.37	(0.25)	0.20
Sexual education communication topics§	9.5 (5.5, 11.5)	8.0 (4.0, 12.0)	−0.75	(0.67)	0.28
Sexual risk reduction topics¶	8.0 (5.0, 10.0)	8.0 (4.0, 11.0)	0.13	(0.52)	0.81
Responsiveness to being asked about sexual topics**	12.0 (11.0, 12.0)	11.0 (10.0, 12.0)	−0.78	(0.34)	0.03
Caregiver-child conflict score*	16.0 (15.0, 16.0)	15.0 (14.0, 16.0)	−0.24	(0.34)	0.49

^*Measured using eight subscales of the Medical Outcomes Study Short form SF-36; range 0–100, higher scores meaning better health.

^†Measured using four subscales from the Families Matter! Program toolkit. Range 0-16, higher scores meaning less conflict

^‡Range 0–7, higher scores meaning more positive attitudes.

^§Range 0–12, higher scores meaning more communication.

^¶Range 0–12, higher scores meaning topics discussed more.

^**Range 0–12, higher scores meaning more responsive.

SF-36, Short Form-36; β, beta coefficient.

Discussion

A household-level agricultural and finance intervention among adults living with HIV improved food security, mental health and SRH among AGYW living in their households, including less FI, higher BMI, fewer symptoms of depression and anxiety, and less sexual IPV. Associations were more pronounced among AGYW aged 15–19 years and among those whose primary caregiver was the index participant in the intervention trial. These findings are above and beyond the intervention’s impact on the health of adult PLHIV enrolled in SM.⁴⁴

We did not find any differences between the intervention and control arms in terms of incident HIV, STIs, unprotected sex or unintended pregnancy, although our study was not powered to detect these associations. Attesting to the potential promise of this intervention for SRH, in a concurrent qualitative study among a purposive sample of adolescent SM participants, we previously reported that adolescents in the intervention arm described reduced transactional sex and fewer concurrent sexual partnerships, while no similar changes were reported among control participants.⁷¹ These positive SRH changes were perceived to be related to several intersecting mechanisms, including that the intervention improved food security at the household level, thereby helping adolescents meet their basic needs like food, helping them stay in school and increasing their confidence and self-efficacy in overcoming existing reciprocity norms and inequities in sexual relationship power.⁷¹ Additionally, the intervention increased caregiver confidence in talking about adolescent SRH issues. Our quantitative results supporting positive impacts on similar pathways that shape SRH, including IPV, mental health and nutrition, add further support to the potential for the SM intervention to improve adolescent SRH.

Our findings support previous literature showing that targeting low household economic empowerment can improve SRH outcomes among AGYW, even though, as above, most of the existing literature focuses on AGYW directly, rather than on the household level. For example, the SUUBI Study in Uganda found that HIV-affected adolescents randomised to a comprehensive microfinance intervention composed of matched savings accounts, financial management workshops and mentorship had improved HIV treatment adherence²⁹ and reduced sexual risk-taking intentions²⁴ compared with those in the control arm. The SHAZ! Study in Zimbabwe, an RCT of a multicomponent intervention package delivered to adolescent girls that included life skills and health education, vocational training, micro-grants and social services, showed efficacy in reducing transactional sex and unintended pregnancy.²⁷ While these and similar studies have demonstrated positive effects on some SRH-related outcomes, other types of economic-strengthening interventions targeted at the individual level can pose unique risks for adolescents who need guidance and support from family and community. For example, implementation of the Adolescent Girls Empowerment Program (AGEP) in Uganda—a social, health and economic-asset building intervention for vulnerable adolescent girls aged 10–19 years—resulted in increased odds of sexual touching and harassment by men among girls who received the economic asset-building component on its own without the social protection components of the full intervention.³³ Results from a cluster RCT evaluating the AGEP in Zambia concluded that economic constraints at the household level may need to be addressed to yield change in SRH risk behaviours for girls.⁷² Our findings further support the promise of household-level economic and livelihood interventions as a strategy to improve health and SRH outcomes for AGYW.

That SM decreased anxiety and depression among AGYW is supported by a growing body of literature showing that economic strengthening and livelihood interventions can positively affect mental health among adolescents.^{73 74} In our concurrent qualitative study, adolescent girls and caregiver pairs in the intervention arm reported improved psychological well-being that they perceived to have occurred through several pathways that included: (1) reduction of social isolation; (2) reduction of shame and stigma; (3) increased attendance and concentration in school; (4) improved caregiver mental health and (5) reduced parental aggression and improved household communication.⁷⁵

SM intervention impacts were more likely among AGYW aged 15–19 compared with the full sample that included their younger counterparts (aged 13–14 years). Since the average age of sexual debut in the rural study setting tends to be younger than in other parts of Kenya, around age 15 or younger,76,78 this is a time when the interplay between poverty, FI, women’s disempowerment and sexual risk taking is likely to become more pronounced. These findings suggest that it may be prudent to focus on this older age range in future household-level livelihood and agricultural interventions aiming to improve SRH among AGYW.

Our study had several important strengths and limitations. Only households that had access to farming land and available surface water or a shallow well were eligible to participate in the study, limiting the generalisability of these findings to other populations in western Kenya. Our trial was focused on HIV-affected households and not specifically designed with AGYW as the main target population. While this was by design, as the goal of the study was to assess whether a household-level livelihood intervention can benefit AGYW, the study was not optimised to improve SRH in this group. For example, most of the girls in the trial were at day school and less commonly at boarding school, so were less exposed to the day-to-day activities of SM, and this may have decreased the potential impacts of the intervention on these girls. Similarly, the primary caregiver of many adolescents in the study was not the index SM participant, and hence those adolescents may have been less closely affected by the intervention. Consistent with this hypothesis, we found that the intervention had greater impacts among older adolescents for whom their primary caregiver was the index participant. Future iterations of SM could consider alternative designs to maximise demonstration of benefits on AGYW, such as limiting to adolescents who are living at home rather than at boarding school, enrolling the adolescent together with her primary caregiver in the intervention, and including some SRH-specific education components as part of the intervention package. That we found benefits on nutrition, mental health and IPV outcomes without inclusion of these adolescent-specific components strengthens the conclusion that household-level interventions can improve health and well-being for AGYW. It is likely that an enhanced version of SM targeted specifically to AGYW may show even more promise in improving SRH in this population. Additional key strengths of our study were the relatively large sample of adolescents, data collected on both AGYW and caregivers (finding similar effects on FI in both), and use of rigorous measures, including biomarkers, to assess our study outcomes. Power for some outcomes was limited by the number of clusters.

In summary, a household-level multisectoral agricultural and finance intervention improved nutritional, mental health and SRH in AGYW in western Kenya. These findings can help guide the development of scalable, household-level FI interventions with the goal to improve mental, physical and sexual health among AGYW in SSA. As the intervention included irrigation as a tool to help households cope with seasonal droughts that are increasingly affected by climate change, this work also supports the benefits of climate-adaptive livelihood interventions to tackle the intersecting problems of poverty, FI, climate change, poor mental health and HIV risk among adolescent girls. If proven effective in larger trials, this would strongly argue for the importance of incorporating such interventions into HIV and other prevention efforts for AGYW.

Data availability statement

Data are available on reasonable request.