A systematic review of income generation interventions, including microfinance and vocational skills training, for HIV prevention

Abstract

Income generation interventions, such as microfinance or vocational skills training, address structural factors associated with HIV risk. However, the effectiveness of these interventions on HIV-related outcomes in low- and middle-income countries has not been synthesized. We conducted a systematic review by searching electronic databases from 1990-2012, examining secondary references, and hand searching key journals. Peer-reviewed studies were included in the analysis if they evaluated income generation interventions in low- or middle-income countries and provided pre-post or multi-arm measures on behavioral, psychological, social, care or biological outcomes related to HIV prevention. Standardized forms were used to abstract study data in duplicate and study rigor was assessed. Of 5,218 unique citations identified, 12 studies met criteria for inclusion. Studies were geographically diverse, with 6 conducted in sub-Saharan Africa, 3 in South or Southeast Asia, and 3 in Latin America and the Caribbean. Target populations included adult women (N=6), female sex workers/bar workers (N=3), and youth/orphans (N=3). All studies targeted females except 2 among youth/orphans. Study rigor was moderate, with 2 group-randomized trials and 2 individual-randomized trials. All interventions except 3 included some form of microfinance. Only a minority of studies found significant intervention effects on condom use, number of sexual partners or other HIV-related behavioral outcomes; most studies showed no significant change, although some may have had inadequate statistical power. One trial showed a 55% reduction in intimate partner violence (adjusted risk ratio 0.45, 95% confidence interval 0.23-0.91). No studies measured incidence/prevalence of HIV or sexually transmitted infections among intervention recipients. The evidence that income generation interventions influence HIV-related behaviors and outcomes is inconclusive. However, these interventions may have important effects on outcomes beyond HIV prevention. Further studies examining not only HIV-related outcomes, but also causal pathways and intermediate variables, are needed. Additional studies among men are also needed.

Background

Poverty, unemployment and lack of economic opportunity are structural factors that shape HIV risk (Dworkin & Blankenship, 2009; Parker, Easton, & Klein, 2000). Globally, interventions that operate at the structural level to address such factors are increasingly seen as key to effectively reducing HIV transmission (Fenton, 2004; Parker et al., 2000).

Microfinance and vocational training are two approaches to increasing people's ability to generate income and secure livelihoods, helping to address structural factors associated with HIV risk. Microfinance encompasses a range of financial services, including credit, savings, insurance, and fund transfers, provided to individuals or groups who normally would not be reached by traditional financial institutions. Microcredit programs provide small loans to those who would not generally qualify for traditional loans. Terms of microcredit loans vary by their initial and subsequent amounts, interest rate, and length of repayment. While microcredit programs often use group lending, since poor clients frequently have little collateral, there are wide variations: loans may be dispersed individually and payments made individually, dispersed to a group with group repayment, or dispersed individually with group repayment. In addition to microfinance, other income generation interventions train participants in vocational skills, either for positions within existing industries or to develop small businesses. Both microfinance and vocational skills training programs may include additional components, such as health education, gender awareness, critical thinking, or communication skills, and many provide strong social support.

Two primary mechanisms have been proposed to explain how income generation interventions might affect HIV-related outcomes. First, they may increase financial independence, feelings of self-confidence, and power within relationships, making participants more able to negotiate safer sex and less dependent on exchanging sex for money or material goods (Dworkin & Blankenship, 2009). Second, these programs are often conducted in supportive groups that may provide participants with opportunities to build social capital, increasing access to resources and increasing knowledge and self-efficacy for HIV prevention-related behaviors. Third, income generation programs may serve as convenient platforms for adding HIV education and skills training (Dworkin & Blankenship, 2009).

Income generation interventions aim to address structural factors associated with HIV risk. However, their effectiveness on HIV-related outcomes has not been synthesized. We conducted a systematic review of the effect of income generation interventions, including microfinance and vocational skills training, on HIV prevention in low- and middle-income countries.

Methods

This review is part of the Evidence Project, a series of systematic reviews of HIV behavioral interventions in low- and middle-income countries following established guidelines (Moher, Liberati, Tetzlaff, & Altman, 2009).

Definition and inclusion criteria

We defined income generation interventions as interventions which impart vocational skills or provide capital or commodities that enhance the capacity of individuals or groups to generate income.

Articles included in the review had to meet the following criteria:

1. Published in peer-reviewed journals.

2. Evaluate income generation interventions as defined above.

3. Use pre/post or multi-arm designs comparing individuals who received the intervention to those who did not.

4. Measure behavioral, psychological, social, care or biological outcomes related to HIV prevention.

5. Conducted in low- or middle- income counties (World Bank, 2012).

We excluded studies examining conditional or unconditional cash transfer interventions with no expectation of repayment as these have been reviewed elsewhere (Pettifor, Macphail, Nguyen, & Rosenberg, 2012). No restrictions were placed on intervention setting, target population, or language of publication.

Search strategy

We searched PubMed, PsycINFO, Sociological Abstracts, CINAHL, EMBASE, and EconLit from January 1, 1990 to August 15, 2012. We also reviewed the table of contents of the journals AIDS, AIDS and Behavior, AIDS Education and Prevention, and AIDS Care and reference lists of included articles.

Search terms

The following terms were entered into computer databases: [(“micro-credit” OR “micro credit” OR microcredit OR “job training” OR “income generation” OR “income generating” OR “job skills” OR employment OR “economic empowerment” OR cooperatives OR “micro-finance” OR “micro finance” OR microfinance OR “micro-enterprise” OR “micro enterprise” OR microenterprise OR “small business” OR “small loans” OR “micro loans” OR microloans OR “micro-loans” OR “vocational training” OR “business training” OR livelihood) AND (HIV OR AIDS)].

Screening abstracts

Titles, abstracts, citation information, and descriptor terms were screened by study staff. Full text articles were obtained of selected abstracts and assessed for final eligibility by two independent reviewers. Differences were resolved through consensus. Articles presenting relevant qualitative, cost-effectiveness, or review information were included as background material.

Data extraction and analysis

Data were extracted independently by two reviewers using standardized forms. Differences were resolved through consensus. Corresponding authors were contacted when clarification was needed.

The following information was gathered from each study: location, setting and target group; time period; intervention description; study design; sample size and characteristics; ; follow-up; outcome measures; comparison groups; results; and limitations. Study quality (rigor) was assessed using the following items: (1) prospective cohort; (2) control/comparison group; (3) pre-/post-intervention data; (4) random assignment to intervention; (5) random selection for assessment; (6) follow-up>=80%; (7) socio-demographic equivalence; and (8) baseline outcome measure equivalence. Data were extracted from background articles using a simplified form.

Meta-analysis was not conducted due to heterogeneity across studies in intervention modalities, target populations, and measured outcomes.

Results

Study Descriptions

We identified 7,611 citations through database searching and 39 through secondary and hand searching (Figure 1). After removing duplicates, 5,218 citations were screened and 72 full-text articles were pulled for review. Of these, 5 did not meet study design criteria and 48 were included as background. Two articles were excluded after extensive discussion. One intervention “collapsed” three months after implementation for not adequately incentivizing participants or addressing local issues; this article was excluded because it was unclear whether the evaluation compared participants who received the intervention to those who did not (Boungou Bazika, 2007). Another intervention, which included post office savings accounts, was excluded because the savings program was not directly linked to income generation, and it was unclear which participants joined the program (Basu et al., 2004). Seven articles were excluded for being related to other articles in the review; for example, we identified a pilot study and larger trial of the same intervention in Uganda (Ssewamala, Alicea, Bannon, & Ismayilova, 2008; Ssewamala et al., 2010), so we included the larger trial results (Ssewamala et al., 2010). The remaining 12 articles were eligible for inclusion. One article (Sherer, Bronson, Teter, & Wykoff, 2004) reported data from three countries with different programs, which we analyzed as three different studies. Another study, the IMAGE project, reported sexual behavior and HIV-related outcomes in three articles (Kim et al., 2009; Pronyk et al., 2006; Pronyk et al., 2008), but were analyzed as one study. Thus we ultimately included 12 studies reported in 12 articles (Ashburn, Kerrigan, & Sweat, 2008; Dunbar et al., 2010; Kim et al., 2009; Lee et al., 2010; Odek et al., 2009; Pronyk et al., 2006; Pronyk et al., 2008; Rosenberg, Seavey, Jules, & Kershaw, 2010; Rotheram-Borus, Lightfoot, Kasirye, & Desmond, 2012; Sherer et al., 2004; Sherman et al., 2010; Ssewamala et al., 2010).

Figure 1.

Flow chart for included/excluded records

The 12 included studies were geographically diverse (Table 1). Six were conducted in sub-Saharan Africa (2 in Uganda and 1 each in Zimbabwe, Kenya, Malawi, and South Africa), three in South or Southeast Asia (1 each in India, Cambodia and Thailand), and three in Latin America and the Caribbean (1 each in the Dominican Republic, Haiti, and Guatemala). Target populations included female sex workers or bar workers (N=3), youth or orphans (N=3), and adult women (N=6). All studies targeted women except two with youth/orphans (Rotheram-Borus et al., 2012; Ssewamala et al., 2010).

Table 1.

Characteristics of included articles

Study name and country	Citation	Population characteristics	Intervention description	Study design	Outcomes
Microfinance alone (2 studies)
Mujeres en Desarrollo Dominicana, Inc. (MUDE), Dominican Republic	(Ashburn, Kerrigan, & Sweat, 2008)	Members of women's groups who had received loans from MUDE Gender: 100% female Age: 18-49	Women's groups received loans through MUDE, a Dominican NGO. MUDE offered both loan and health programs to intervention participants.	Cross-sectional study design. Assessments took place only once (N=273). Participants were non-randomly selected.	Negotiation of partner's HIV-related behavior (ever received a loan vs. not): OR: 0.91 (95% CI: 0.56, 1.49) AOR: 1.66 (95% CI: 0.76, 3.64)
IMAGE Study, South Africa	(Kim et al., 2009)	Women aged 18 years or older from “poor” households seeking loans Gender: 100% female Age (median): 45	Participants were either in the control group (received nothing), microfinance (MF) group (loans from an NGO), or the IMAGE group (loans and “Sisters for Life” Program). In microfinance, groups of five women served as guarantors for one another's loans and all five had to repay before any group member was eligible for more credit. The “Sister for Life” program included a 12-15 month training curriculum on gender roles, cultural beliefs, power relations, self-esteem, communication, domestic violence, HIV, and communication and critical thinking skills. Community mobilization among youth and men was also conducted.	Group randomized trial. Assessments took place two years post-intervention (N=1,230). Villages were randomly selected.	Condom use at last sex with all non-spousal partners: MF-Only vs. Control: Risk Ratio (RR): 1.74 (95% CI: 0.37, 8.21); aRR: 1.17 (95% Cl: 0.32, 4.29) IMAGE vs. Control: RR: 2.41 (95% CI: 0.77, 7.54); aRR: 1.83 (95% CI: 0.94, 3.57) IMAGE vs. MF-Only: RR: 1.41 (95% CI: 0.97, 2.04); aRR: 1.41 (95% CI: 0.97, 2.04) Household communication about sex and HIV: MF-Only vs. Control: RR: 1.15 (95% CI: 0.76, 1.72); aRR: 1.17 (95% CI: 0.76, 1.80) IMAGE vs. Control: RR: 1.60 (95% CI: 1.25, 2.05); aRR: 1.57 (95% CI: 1.20, 2.05) IMAGE vs. MF-Only: RR: 1.37 (95% CI: 0.98, 1.93); aRR: 1.32 (95% CI: 0.90, 1.93)
	(Pronyk et al., 2006)	Low income rural women Gender: 100% female Age: median (range): Intervention: 41(34-49) Comparison: 42(33-49)	See above.	Group randomized trial. Assessments took place at baseline (N=843) and 2.1 years later (N=750). Participants were non-randomly selected.	Communication with intimate partner about sexual matters in past 12 months (Int. vs. Control): RR: 1.14 (95% CI: 0.87, 1.50); aRR: 1.14 (95% CI: 0.90, 1.44) Experience of intimate-partner violence in past 12 months (Int. vs. Control): RR: 0.50 (95% CI: 0.28, 0.89); aRR: 0.45 (95% CI: 0.23, 0.91)
	(Pronyk et al., 2008)	Women between the ages of 14-35 Gender: 100% female Age (mean): Intervention: 29.9 Control: 29.2	See above.	Group randomized trial. Assessments took place at baseline (N=220) and at 2 years post-intervention (N=187). Participants were randomly selected.	Gone for voluntary counseling and testing (Int. vs. Control): RR: 1.65 (95% CI: 1.26, 2.16); aRR: 1.64 (95% CI: 1.06, 2.56) More than one sexual partner in the last 12 months (Int. vs. Control): RR: 1.20 (95% CI: 0.55, 2.63); aRR: 0.95 (95% CI: 0.40, 2.27) Unprotected sex at last intercourse with a non-spousal partner in the last 12 months (Int. vs. Control): RR: 0.70 (95% CI: 0.59, 0.84); aRR: 0.76 (95% CI: 0.60, 0.96)
Microfinance with health education (5 studies)
IMAGE Study, South Africa	See above
Project HOPE's Village Health Bank (VHB) program Guatemala, Malawi, and Thailand (3 studies)	(Sherer, Bronson, Teter, & Wykoff, 2004)	Individuals living in impoverished communities in countries with high HIV prevalence Gender: 100% female Age: NR	The VHB program provided small loans to groups of women to generate income through agriculture, sales of goods and services, and other activities. Health education included sessions on early recognition of serious childhood illnesses, sexual health (e.g. STIs and HIV/AIDS), and promotion of personal health. Groups of 18-25 women received small loans as well as biweekly, 1 hour health education sessions. The life of each loan was about 4 months or 8 education sessions.	Before/after study design. Assessments took place at baseline (N=208 in Guatemala, 227 in Malawi, and 68 in Thailand) and at the end of the fourth loan cycle (about one year) (N=208 in Guatemala, 227 in Malawi, and 68 in Thailand). Participant selection approach was not reported.	HIV prevention knowledge (% improvement): Guatemala: 12%; Malawi: 10%; Thailand: +13%; all p<0.05 Knowledge of 3 STI signs (% improvement): Guatemala: 14%; Malawi: 8%; Thailand: +13%; all p<0.05 Self-efficacy (% improvement): Guatemala: 27%; Malawi: 17%; Thailand: +14%; all p<0.05 Access primary care for child health (% improvement): Guatemala: 29%; Malawi: 12%; Thailand: +11%; all p<0.05
SUUBI Uganda	(Ssewamala et al., 2010)	Adolescents that have lost 1 or both of their parents due to HIV/AIDS Gender: Control Group: Boys n= 70, Girls n= 78; Treatment Group: Boys n= 55, Girls n = 83 Age: (mean): 13.7	All participants received usual care for orphaned children, consisting of counseling, educational supplies, and national school-based health education (including HIV). Intervention participants also received the SUUBI economic empowerment intervention: (1) twelve 1-2 hour workshops over 10 months focused on financial planning and asset-building strategies, including saving, education, and small business development; (2) a monthly mentorship program for adolescents with peer mentors on future planning and life options; and (3) a matched child savings account for post-primary school.	Group randomized trial. Assessments took place at baseline (N=286) and at 10 months (N=277) post-intervention. Participants were randomly selected.	Attitudes toward engaging in sexual risk–taking behavior (score): Girls: Int: BL: 8.28; FU: 8.25; Control: BL: 7.73; FU: 9.19 Boys: Int: BL: 12.44; FU: 10.29; Control: BL: 8.9; FU: 13.43 Time × Treatment × Gender interaction: F [1,266], 16.1, p<.001
Microfinance, health education, and business development/vocational training (3 studies)
SHAZ! Zimbabwe	(Dunbar et al., 2010)	Adolescent female orphans Gender: 100% female Age (mean): 17.5	Participants received a 10-module life-skills course that covered HIV and reproductive health knowledge and skills, gender, culture, and physical and sexual violence. They also received a 5-day business training session in order to develop business plans, a 4-day skills building workshop, and mentorship from a local businesswoman. Finally, participants received microcredit loans from a local microfinance organization.	Before-after study design. Assessments took place at baseline (n=49) and 6 month post-intervention (n=37). Participants were non-randomly selected.	HIV knowledge (number of correct answers out of 13): 7-8: BL: 22%, FU: 0%; 9-10: BL: 20%, FU: 3%; 11-12: BL: 41%, FU: 60%; 13: BL: 16%, FU: 38%, p<0.001 Sexually active: BL: 18%, FU: 22%, p=0.79 High power in primary sexual relationship: BL: 11%, FU: 50%, p=0.16 High power in nonsexual relationship: BL: 5%, FU: 38%, p=0.04 Use of condoms with primary partner: BL: 67%, FU: 38%, p=0.35
Strengthening STD/HIV Control Project in Kenya (SHCP) Kenya	(Odek et al., 2009)	Female sex workers in Kenya Gender: 100% female Age (mean): 41.09	Participants received peer education, which included safer sex negotiation, STI/HIV education and counseling, promotion of consistent condom use, and education messages to address the importance of reducing number of sexual partners, avoiding unprotected vaginal and anal sex, and increasing use of non-penetrative sex and avoiding sex during menses. They were then given microfinance loans following group-based lending structures. After receiving the loans, they received one-on-one business counseling, group business training and field-based mentorship. They also attended weekly meetings where they contributed to mandatory savings accounts and repaid loans.	Before-after study design. Assessments took place at baseline (N=307) and at 27 months post-intervention (N=227). Analysis included only the 227 participants who completed both surveys. Participants were non-randomly selected.	Number of sexual partners in past week, Mean (SD): All partners: BL: 3.26 (2.45); FU: 1.84 (2.15), p<0.001 Casual partners: BL: 1.42 (2.55); FU: 1.12 (1.53), p=0.098 Regular partners: BL: 1.96 (1.86); FU: 0.73 (0.98), p<0.001 Condom use with casual partners: Always: BL: 93.8%, FU: 85.4%; Sometimes: BL: 4.2%; FU: 4.6%; Never: BL: 2%, FU: 0%; p=0.727 Condom use with regular partners: Always: BL: 78.9%, FU: 93.5%; Sometimes: BL: 9.6%; FU: 2.8%; Never: BL: 11.4%, FU: 3.7%; p=0.031
Fonkoze Haiti	(Rosenberg, Seavey, Jules, & Kershaw, 2010)	Women participating in microfinance Gender: 100% female Age: Mean: 36.1; Range: 18-49	Groups of five women organized together to take out and repay loans. All women began with an initial loan of US $75 with a 3 month repayment period. The program policy was to provide at least two of the five possible training modules to each client within the first year of membership. The modules included trainings in basic literacy, business skills, children's rights, environment, and health education. Health education used a book illustrating pertinent health issues for Haitian women such as hygiene, reproductive health, and STIs (including HIV).	Cross-sectional study design. Assessments took place one time only (N=192). Participants were non-randomly selected.	Condom use in last year (Intervention participant greater than 12 months vs. less than 12 months): OR: 0.63 (95% CI: 0.26, 1.54) Condom use among those with an unfaithful partner (Intervention participant greater than 12 months vs. less than 12 months): OR: 3.95 (95% CI: 0.93, 16.85) Ever tested for HIV (Intervention participant greater than 12 months vs. less than 12 months): OR: 0.78 (0.41, 1.49)
Vocational training with health education (3 studies)
SiRCHESI Hotel Apprenticeship Program (HAP) Cambodia	(Lee et al., 2010)	Women in high-risk professions in Cambodia Gender: 100% female Age: Mean: 24.93; Range: 19-31	Participants received a 24-month training. The first 8 months consisted of morning classes that focused on Khmer literacy, conversational English, health education including HIV, reproductive and sexual health knowledge, and life and social skills education. In the afternoons, participants had mentored internships at a 3-or 5-star hotel. For the remaining 16 months, participants had full time hotel apprenticeships.	Time series study design. Assessments took place at baseline with a matched pair group (N=14) and at various points throughout the intervention (N varied throughout follow up). Participants were non-randomly selected.	HIV knowledge (out of 5), Mean (SD): Int: 4.79 (0.39); Control: 4.79 (0.39); p=1.00 Condom suggestion at last sex: Int: 66.7%; Control: 87.5%; p=0.278 Condom use at last sex: Int: 60%; Control: 100%; p=0.050
Street Smart Uganda	(Rotheram-Borus, Lightfoot, Kasirye, & Desmond, 2012)	High-risk urban youth Gender: Intervention Group: 60% Female 40% Males Delayed Control Group: 38% Female 62% Male Age Range: 13 to 23 years	All participants received the Street Smart intervention, a 10-week, 10-session HIV prevention intervention. Participants were also randomized to immediate or delayed vocational training. Vocational training consisted of apprenticeships with local artisans in hairdressing, catering, tailoring, mechanics, electronics, carpentry, cell phone repair, and welding. Youth attended classes regularly for 4–8 hours, 5 days a week. Artisans received a 5-day training on how to talk to youth (generally and about HIV), conflict resolution, HIV prevention, and coping with unprofessional behavior.	Individual randomized trial. Assessments took place at baseline (N=100), 4 months (N=85) and at 24 months (N=74) post-intervention. Participants were non-randomly selected.	Engaged in sex: Int: 58%; FU: 59% ; Control: BL: BL: 80%; FU: 91% Number of partners, Mean (SD): Int: BL: 2.10 (3.33); FU: 0.88 (0.90); Control: BL: 1.82 (1.51); FU: 1.36 (0.81) Abstinent or 100% condom use: Int: BL: 54%; FU: 95%; Control: BL: 29%; FU: 64% Intervention × Time interactio: F [1, 82], 14.15, p=0.0003
Pi Bags India	(Sherman et al., 2010)	Female sex workers Gender: 100% female Age (median): 35	The month-long intervention included 8 hours of HIV prevention education taught by health educators plus 100 hours of tailoring training taught by master tailors. The HIV prevention sessions aimed to both educate participants about HIV risk and to help women develop the skills necessary to reduce sexual risk when exchanging sex. The training goal was for women to independently make large cotton totes.	Individual randomized trial. Assessments took place at baseline (N=100) and 6 months (N=99). Participants were non-randomly selected.	Number of sex partners, Mean (SD): Int: BL: 10.9 (22); FU: 5 (5); Control: BL: 10.3 (14); FU: 11.9 (10); Int vs. Control at FU: p<0.001 Number of sex exchange partners, Mean (SD): Int: BL: 6.4 (6); FU: 3.1 (2); Control: BL: 6.9 (9); FU: 5.1 (3); Int vs. Control at FU: p<0.001 Condom use at last sex exchange: Int: BL: 86%; FU: 98%; Control: BL: 90%; FU: 100%; Int vs. Control at FU: p=0.32 Always regular use of condoms with clients in past 6 months: Int: BL: 82%; FU: 96%; Control: BL: 80%; FU: 96%; Int vs. Control at FU: p=1.00 Adjusted intervention effect: Beta: −1.8 (95% Cl: −2.9, - 0.8)

BL: Baseline; FU: Follow-up

Overall study rigor was moderate (Table 2). There were two group-randomized trials (Kim et al., 2009; Pronyk et al., 2006; Pronyk et al., 2008; Ssewamala et al., 2010) and two individual-randomized trials (Rotheram-Borus et al., 2012; Sherman et al., 2010). The remaining studies used before/after, time series, or cross-sectional designs. The rigor of the randomized trials was generally high, with participants randomly assigned to interventions (by group or individually) and follow-up rates above 80%.

Table 2.

Quality assessment of included articles

Study name and country	Citation	Cohort	Control or comparison group	Pre/post intervention data	Random assignment of participants to the intervention	Random selection of participants for assessment	Follow-up rate of 80% or more	Comparison groups equivalent on socio-demographics	Comparison groups equivalent at baseline on outcome measure
Microfinance alone (2 studies)
MUDE, Dominican Republic	(Ashburn, et al., 2008)	No	Yes	No	No	No	NA	NR	NR
IMAGE, South Africa	(Kim, et al., 2009)	Yes	Yes	No	Yes	Yes	Yes	Yes	NA
	(Pronyk, et al., 2006)	Yes	Yes	Yes	Yes	No	Yes	Yes	No
	(Pronyk, et al., 2008)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Microfinance with health education (5 studies)
IMAGE Study, South Africa	(Kim, et al., 2009)	Yes	Yes	No	Yes	Yes	Yes	Yes	NA
	(Pronyk, et al., 2006)	Yes	Yes	Yes	Yes	No	Yes	Yes	NA
	(Pronyk, et al., 2008)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	NA
Project HOPE VHB, Guatemal a, Malawi, and Thailand (3 studies)	(Sherer, et al., 2004)	Yes	No	Yes	NA	No	NA	NA	NA
SUUBI, Uganda	(Ssewamal a, et al., 2010)	Yes	Yes	Yes	Yes	Yes	Yes	NR	NR
Microfinance, health education, and business development/vocational training (3 studies)
SHAZ!, Zimbabwe	(Dunbar, et al., 2010)	Yes	No	Yes	NA	No	NR	NA	NA
SHCP, Kenya	(Odek, et al., 2009)	Yes	No	Yes	NA	No	No	NA	NA
Fonkoze, Haiti	(Rosenberg , et al., 2010)	No	Yes	No	No	No	NA	NA	NA
Vocational training with health education (3 studies)
SiRCHESI HAP, Cambodia	(Lee, et al., 2010)	No	No	No*	NA	No	No	NA	NA
Street Smart, Uganda	(Rotheram-Borus, et al., 2012)	Yes	Yes	Yes	Yes	No	No	No	NA
Pi Bags, India	(Sherman, et al., 2010)	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes

^*Baseline measures were among a group of 14 women matched to program participants; no baseline measurements were taken when the program began.

All interventions except three included microfinance, either with or without additional components such as vocational training and health education; the remaining three interventions provided vocational training combined with health education. We present results by these intervention combinations: microfinance alone, microfinance with health education, microfinance with health education and vocational training, and vocational training with health education. Figure 2 presents the number of studies in each category.

Figure 2.

Venn diagram showing number of studies with each primary intervention component

Microfinance alone

Two studies examined the effect of microfinance alone on HIV-related behaviors: one group-randomized trial (Kim et al., 2009; Pronyk et al., 2006; Pronyk et al., 2008) and one cross-sectional study (Ashburn et al., 2008). In South Africa, the IMAGE project randomized groups to either microfinance alone, microfinance with health education, or control (Kim et al., 2009; Pronyk et al., 2006; Pronyk et al., 2008). Compared to the control, participants receiving microfinance alone did not report significantly different rates of condom use at last sex with all non-spousal partners (Table 1) (Kim et al., 2009). Across a variety of additional outcomes, there was no consistent pattern of directionality of association when comparing participants receiving microfinance alone to the control group (Kim et al., 2009). The second study examined microcredit for women's group members in the Dominican Republic (Ashburn et al., 2008). Receiving a loan was not significantly associated with HIV-related negotiation with a current partner.

Microfinance with health education

Five studies examined microfinance combined with health education (Kim et al., 2009; Pronyk et al., 2006; Pronyk et al., 2008; Sherer et al., 2004; Ssewamala et al., 2010), including both group-randomized trials..

The IMAGE study's primary intervention was group-based microfinance plus a 12-15 month “Sisters for Life” gender and HIV training curriculum (Kim et al., 2009; Pronyk et al., 2006; Pronyk et al., 2008). Participants reported higher rates of condom use at last sex with all non-spousal partners compared to control and microfinance-only participants, but these differences were not significant in multivariate analysis (Kim et al., 2009). In another analysis, compared to control participants, IMAGE participants were more likely to have gone for voluntary HIV counseling and testing and less likely to report unprotected sex during last sex with a non-spousal partner, but there was no difference in frequency of multiple partnerships (Pronyk et al., 2008). Across additional outcomes related to knowledge, attitudes, communication, violence, gender roles, collective action, and social engagement, there was a general trend towards improved outcomes comparing IMAGE to control participants, although many differences were not statistically significant. However, IMAGE participants reported a 55% reduction in intimate partner violence over two years (Pronyk et al., 2006).

The second study was a group-randomized trial among AIDS-orphaned adolescents in Uganda (Ssewamala et al., 2010). Intervention and control groups received counseling, educational supplies, and school-based health education; intervention participants also received workshops on asset-building, financial planning, and business development; mentoring; and microfinance in the form of matched child savings accounts. The only primary outcome was attitudes towards engaging in risk behaviors; gender-stratified analyses showed attitude improvements among boys, but not girls.

The remaining 3 studies, presented in one article, evaluated group loans and health education provided through village health banks for women in Guatemala, Malawi, and Thailand (Sherer et al., 2004). After approximately one year, participating women showed moderate but significant improvements ranging from 8 to 29% in HIV and sexually transmitted infection (STI) knowledge, self-efficacy, and accessing primary care for child health.

Microfinance, health education, and business development/vocational training

Three studies examined microfinance, health education, and business development or vocational training combined: one cross-sectional (Rosenberg et al., 2010) and two before/after studies (Dunbar et al., 2010; Odek et al., 2009).

One study conducted among female sex workers in Kenya's urban slums involved peer education on HIV/STIs, condoms, group-based microfinance loans, and business counseling and mentorship (Odek et al., 2009). Post-intervention women reported fewer total and regular sex partners, but no change in casual sex partners. Women also reported increased consistent condom use with regular partners, but no change in already high condom use rates with casual partners.

The second study, SHAZ!, provided life-skills based HIV education, business training and mentorship, and access to microcredit loans to adolescent female orphans in Zimbabwe (Dunbar et al., 2010).

Participants also received workshops in soap, tie-dye or candle making. Post-intervention the percent of participants reporting sexually activity and condom use with primary partners did not change; however, as few adolescents were sexually active, the study likely lacked statistical power to detect these differences. Further, there were difficulties implementing the intervention, with low rates of loan repayment and business success; The third study provided microfinance loans, HIV education, and 3 months of business training to Haitian women (Rosenberg et al., 2010). In a cross-sectional design comparing women with less than or greater than 12 months of program experience, there were no significant differences in number of lifetime sexual partners, past-year condom use, or condom use among women who reported unfaithful partners; however, women in the program for longer were less likely to report having unfaithful partners.

Vocational training with health education

Three studies evaluated vocational training and health education without a microfinance component (Lee et al., 2010; Rotheram-Borus et al., 2012; Sherman et al., 2010). Two studies involved sex workers or bar workers; provided training in specific job skills rather than general business development (Lee et al., 2010; Sherman et al., 2010). The third study took place among youth (Rotheram-Borus et al., 2012).

One individual-randomized controlled study in Chennai, India, taught street-based sex workers tailoring skills, specifically for making tote bags (Sherman et al., 2010). Intervention participants received 8 hours of HIV prevention education and 100 hours of tailoring training, while control participants received only HIV prevention education. At 6-month follow-up, intervention participants reported significantly fewer sex partners and sex exchange partners than control participants. However, there were no significant differences in condom use at last sexual exchange or consistent condom use with clients.

In Cambodia, female bar workers were placed in hotel internships (Lee et al., 2010). Participants received 8-month language training, health education, life and social skills education, and internships followed by 16-month apprenticeships in 3 or 5-star hotels. The study met almost none of our quality criteria, and, with just 14 participants, was likely significantly underpowered. Nevertheless, condom use at last sex was reported as not significantly different between baseline and cumulative follow-up among intervention participants. Counter to intervention goals, intervention participants actually reported lower rates of condom use at last sex than program drop-outs (60% vs. 100%, p=0.05).

The final study placed Ugandan youth in apprenticeships in hairdressing, catering, tailoring, mechanics, electronics, carpentry, cell phone repair, and welding (Rotheram-Borus et al., 2012). Youth also received a 10-week, 10-session HIV prevention intervention. An individual-randomized controlled design compared outcomes between intervention (apprenticeships/health education) and comparison groups (health education only). After 4 months, both groups showed significant decreases in number of sexual partners and increases in abstinence and condom use, but no change in sexually activity.

Discussion

We identified twelve studies from low- and middle-income countries evaluating the impact of income generation interventions on HIV prevention. Studies were diverse in terms of locations, interventions, target populations, study designs, and rigor. Often studies showed no change in HIV-related outcomes, although some may have lacked statistical power to see intervention effects. Some studies showed positive effects, although none measured HIV incidence or prevalence among intervention participants. Importantly, no rigorous studies included in this review showed negative or harmful outcomes. Nevertheless, there is inconclusive evidence that microfinance and vocational skills interventions are effective at changing HIV-related sexual risk behaviors. These interventions may have important effects on outcomes beyond HIV prevention, however. For example, one rigorous study included in this review showed a positive effect on intimate partner violence (Pronyk et al., 2006), and several studies showed improvements in financial outcomes.

Occasionally, income generation interventions may unintentionally place participants in situations of greater risk of HIV, gender-based violence or other harms. In the SHAZ! study, adolescent girls qualitatively reported threats to their personal safety while transporting goods to market, including harassment by men and police, and lacked safe accommodation or secure places to store money against theft (Dunbar et al., 2010). While these potential harms were not measured quantitatively, they demonstrate potential risks of income generation interventions. Other studies have raised concerns that women's empowerment may exacerbate intimate partner violence by challenging gender norms and provoking conflict between partners (Jewkes, 2002; Schuler, Hashemi, & Badal, 1998). Future income generation evaluations should rigorously monitor and attend to potential harms.

Intervention models included different combinations of microfinance, vocational training, and health education components. The two studies examining microfinance alone did not find significant intervention effects. Microfinance combined with health education, with or without additional vocational training, appears more promising, although outcomes remained mixed compared to either microfinance-alone or control groups. Vocational training without microfinance appeared moderately efficacious for female sex workers in just one rigorous study.

The relatively small number of included studies prevents strong recommendations about intervention design. Most microfinance interventions were group-based; only a few provided loans to individuals. Most were financed and managed by non-governmental organizations. Only the IMAGE study directly compared microfinance alone to microfinance plus health education (Kim et al., 2009); no other studies compared interventions with different characteristics. As included studies represented diverse target populations, settings, study designs, and outcomes, we cannot know whether differences in efficacy were due to intervention components or other factors. Furthermore, intervention efficacy may differ by population or by participants’ income. These questions may inform future evaluations.

All studies except two targeted women and girls, so the effect of income generation programs on HIV prevention among men and boys is largely unknown. Mechanisms through which income generation programs affect HIV-related outcomes may differ by gender. Programs for women often describe their relative lack of empowerment in ability to earn income or control financial resources, and argue that economic empowerment will alter gender norms and reduce transactional sex. For men, pathways through which income generation programs might shape HIV risk are less clear. Further, evidence from cash transfer interventions suggests that providing money to men might actually increase sexual risk behavior (Kohler & Thornton, 2012).

Conclusions of this review must be seen in light of several limitations. All studies relied on self-report of sexual risk behaviors, which may be subject to social desirability bias, particularly for intervention participants. No studies measured biological markers of HIV risk, such as STIs, and none measured HIV incidence or prevalence directly among intervention participants; indirectly, the IMAGE study found no effect on HIV incidence among youth living with intervention participants (Pronyk et al., 2006). Finally, rigor of included studies was often weak. Non-randomized studies likely suffered from selection bias, as individuals participating in income generation programs likely differ from non-participants or dropouts. Although there were four randomized trials, follow-up length may have been too short to measure intervention effects. If these interventions gradually increase the ability of individuals to generate income, thus eventually changing their economic circumstances that may affect other life choices and diminish risk, even two years may not be sufficient to assess more distal health outcomes.

In this review, we assessed the impact of income generation interventions on HIV prevention. However, studies have also examined the impact of these interventions on health and quality of life for people living with HIV (Caldas et al., 2010; Pandit et al., 2010). Such programs may not only improve economic self-sufficiency for people living with HIV, but also provide psychosocial support and increase members’ sense of dignity and self-worth. Income generation programs could similarly help HIV-affected households (Mutenje, Nyakudya, Katsinde, & Chikuvire, 2007). We also excluded studies involving conditional or unconditional cash transfers, which offer alternative models for reducing poverty and increasing economic opportunity; recent studies suggest these interventions hold promise for HIV prevention (Baird, Garfein, McIntosh, & Ozler, 2012; de Walque et al., 2012; Pettifor et al., 2012).

Finally, income generation interventions may lead to changes beyond the individual level. If implemented at sufficient scale, these programs could shape patterns of poverty and income inequality. Although the impact of such changes on HIV-related outcomes is unclear, interventions operating through structural mechanisms may have a variety of complex intended and unintended effects, both positive and negative. Ultimately, though, the opportunity to shape risk environments may offer the most sustainable solution to reducing HIV transmission globally.