Cognitive and educational interventions for orphans and vulnerable children affected by HIV/AIDS: A meta-analysis

Abstract

Background:

Worldwide, there are over 13.3 million orphans and vulnerable children affected by HIV/AIDS (HIV OVC), defined as individuals below the age of 18 who have lost one or both parents to HIV/AIDS or have been made vulnerable by HIV/AIDS, and who are at risk for negative psychosocial and cognitive outcomes.

Purpose:

This meta-analysis aimed to examine the scientific literature on available interventions for HIV OVC, with a focus on community-based interventions (CBI).

Methods:

Systematic electronic searches were conducted from 4 databases between October 2016-April 2017 to identify articles investigating the effectiveness of interventions for HIV OVC. Effect sizes were calculated for each article which provided enough data for analyses.

Results:

Seventy-four articles were reviewed, including psychosocial interventions (d=0.30), cognitive interventions (d=0.14), social protection interventions (d=0.36), and community-based interventions (CBI; d=0.36). Study-specific effect sizes varied widely, ranging from −1.09 to 2.26, from having a negative effect to an impressively large one, but the majority of studies registered small to medium effects (the overall effect size for all studies was 0.32, SE=0.03, 95% CI: 0.26–0.37). Social protection interventions had the highest positive outcomes whereas CBI tended to have the fewest significant positive outcomes, with some outcomes worsening instead of improving.

Conclusions:

Overall, interventions provided to OVC have potential for improving cognitive, psychosocial, and risk-behavior outcomes. Social protection interventions and CBI had the highest effect sizes, but CBI had positive effects on fewer outcomes. CBI warrant scrutiny for improvement, as they represent an important form of culturally-embedded services with potentially long-term benefits to OVC, yet appear to be differentially effective. Successful components of other types of intervention were identified, including cash grants, mentorship, and family therapy. In addition, more research is needed that attends to which interventions may be more effective for specific populations, or that studies cost-effectiveness.

Introduction

HIV/AIDS is a global pandemic that continues to have a vast impact on families, with 37.9 million people living with HIV in 2018 (1). Since the start of the epidemic 35.4 million people have died from AIDS-related illnesses (1) with an estimated 17 million children having lost one or both parents due to AIDS (2). In addition, in 2018, there were 1.7 million children under 15 years old living with HIV (1). Orphans and vulnerable children affected by HIV/AIDS (HIV OVC) are defined as individuals below the age of 18 who have lost one or both parents to HIV/AIDS, or have been made vulnerable by HIV/AIDS. This includes children who are HIV-positive, living without adequate adult support (e.g., in a household with chronically ill parents, a household that has experienced a recent death from chronic illness, a household headed by a grandparent and/or a household headed by a child), living outside of family care (e.g., in residential care or on the streets), or is marginalized, stigmatized, or discriminated against (3).

Although these effects may be mitigated or exacerbated, depending upon their environment, HIV OVC are at risk for a host of negative experiences and outcomes. For example, they often experience difficulties due to stigma, trauma and stress, illness, food insecurity, poverty, and lack of access to resources and education (4). Many AIDS-orphaned children develop lasting mental health problems and psychological distress (5–6). Similarly, children living with HIV-infected or AIDS-ill adults have been reported to have vulnerabilities in physical and emotional health (7), and families affected by HIV/AIDS are at greater risk of psychosocial problems. HIV/AIDS has been associated with reduced positive parenting, mediated by poverty, caregiver depression, and child behavior problems (8). Children who are displaced from their homes after losing their parents due to HIV or living with HIV-infected parents may be more likely to have suicide ideation, and to have attempted suicide or destroyed public property (9). Children living with HIV-infected or AIDS-ill adults often lack access to and opportunities to complete schooling, have irregular grade progression and school attendance, and are at higher risk of being hungry and having difficulty concentrating at school (7). Children affected by HIV/AIDS may also show deficits in cognitive ability, potentially due to a variety of environmental factors such as stigma, family loss, poverty, trauma, and mental health problems (10–11).

As they reach puberty, HIV OVC tend to have a higher prevalence of sexual risky behaviors in comparison to typical adolescents. Research has shown that adolescent boys and girls living in households where at least one adult is HIV-positive have about 25% higher odds of having initiated sexual activity compared to adolescents with similar characteristics where no adult is HIV-positive (12). A large proportion of perinatally HIV-infected and perinatally HIV-exposed uninfected adolescents engage in risky sexual behavior and substance use, with more than half of those who are sexually active reporting unprotected sex (13). Criminal justice involvement, unstable housing, condomless sex, and suboptimal antiretroviral (ARV) therapy have been associated with increased risk of substance use behaviors in HIV-infected adolescents (14). This increase in risky behavior can lead to increased negative health outcomes, social problems, and premature death.

Although HIV OVC are at risk for increased mental health problems, lack of access to resources, cognitive deficits, and greater risk behaviors, little is understood about how effectively interventions address these difficulties. Existing interventions can be generally divided into four categories, psychosocial/behavioral interventions focused on improving mental health or behavior through therapeutic methods; social protection interventions providing monetary support to improve outcomes; cognitive interventions specifically aiming to improve cognition, and community-based interventions (CBI), which are embedded in the community and usually target multiple areas of difficulty for the child and provide resources to the family. However, little is known about how well these interventions address risk areas. Therefore, we hope to characterize the extent to which interventions address both the psychosocial, cognitive, and behavioral deficits experienced by HIV OVC. In addition, interventions are variable in terms of the country they are implemented in, the type of sample they use, and the type of methodology they utilize. However, interventions may have differential effects depending on the country because of differences in HIV populations between countries; for example, in Sub-Saharan Africa, young women are more likely to be affected than young men (1). In additions, interventions may be more effective for a certain age or population, and it is important to understand if there are fewer effective interventions for a certain age group. We plan to examine effect sizes for subgroups based on these variables to describe the effectiveness of the interventions in different regions or for a particular sample or age group, for greater understanding of contexts where interventions may need improvement. Finally, we will examine the types of study methodology used and how this may impact results. This systematic review and meta-analysis aims to (1) summarize the current research on available interventions for HIV/AIDS OVC; (2) identify outcomes addressed by various types of interventions and compare the effects of interventions on these outcomes, (3) determining effectiveness of interventions depending on country of implementation, sample definition (type of OVC), age of participants, and study design; and (4) identify future directions for research.

Methods:

Articles were identified through systematic electronic searches conducted on PSYCinfo, Pubmed, ERIC, and Web of Science, from October 2016-April 2017 (see Figure 1). The searches used combinations of the keywords HIV/AIDS, treatment, intervention, therapy, adolescents, teenagers, children, youth, young adults, OVC, orphans, and children affected by HIV/AIDS. In addition, references from relevant articles were reviewed for additional studies. The search period was January 2001 through April 2017.

Figure 1.

Flow diagram of search strategy

Types of Studies:

Studies examining the effectiveness of interventions for HIV OVC were included in this review; the age was expanded from 18 in the PEPFAR OGAC (2016) definition to 25, in order to include studies with participants in early adulthood. Studies were included that had samples of children orphaned by AIDS, children with a parent with HIV/AIDS, or HIV-infected children. Studies without a strict definition of OVC, but with samples of assumed OVC drawn from participants in community-based organizations in endemic areas, were included as well. Studies published in peer-reviewed journals and “unpublished” studies by program evaluators performed in any country were included. Randomized controlled trials, quasi-experimental trials, and qualitative studies were included in the review. Qualitative studies were included in the systematic review for thoroughness (see Table S1), but were not included in the effect size analyses presented below. Review articles, dissertations, descriptive studies, opinion pieces, and case studies were excluded.

Types of Interventions:

Only interventions targeting behavioral, psychosocial, or cognitive outcomes of HIV OVC were included in the review; interventions consisting only of medication were excluded. The review included one or more interventions in the following categories, based on the focus and goals of the intervention:

1. Psychosocial/Behavioral interventions

2. Cognitive interventions

3. Social protection interventions

4. Community-based interventions

Psychosocial and behavioral interventions were included as one category because of the substantial overlap in their goals, methods and outcomes addressed. Psychosocial interventions aimed to address psychological distress, improve family relationships, reduce social isolation and stigma, address HIV risk factors and decrease risky behaviors, promote healthy behaviors, and promote education and provide school support, and provide advanced care planning. Cognitive interventions, aiming to improve motor, cognitive, and language development, consisted of computerized neurocognitive training (e.g., Computerized Cognitive Rehabilitation Therapy; CCRT), individual home programs, and training caregivers to promote cognitive stimulation in their children (e.g., Medial Interventions for Sensitizing Caregivers; MISC). Social protection interventions included financial components to them with the goals of promoting monetary savings, social support, and economic empowerment. These interventions promoted savings through cash transfer (which could be conditional or unconditional), grants, or a matched savings account, and often included a financial management workshop or mentorship as well. CBI were included as a separate category due to their broader focus and goals and embedment in the community. CBI had goals of improving psychosocial well-being, access to services, empowerment, medication adherence, school support, HIV education, family support, legal protection, and economic strengthening. Most of these interventions provided multiple services, in the form of emotional support, education, tangible support (e.g., assistance with grant applications), and referral to services. CBI provided services through home visits, where either volunteers or paid paraprofessionals visited the home to provide counseling and support, or kids clubs, to reduce social isolation and stigma.

Statistical Analyses:

Effect sizes were computed or converted for relevant outcomes when sufficient information was provided. Effect sizes were reported using standardized mean differences (Cohen’s d), which is the difference in outcome progression (post-test mean scores minus pre-test mean scores in the HIV OVC sample) divided by the pooled standard deviation. For RCT and other studies providing data on HIV OVC and control samples, Cohen’s d was calculated as the standardized difference between the control and HIV OVC groups at post-test, which may be an overestimation of effect size when compared to studies providing data on HIV OVC samples only (18). Thus, we conducted follow-up analyses to differentiate between studies with within-subjects and between-subjects comparisons. However, the goal of the review was to compute effect size estimates that were comparable across within and between-participant designs and across multiple measures (18).

Cohen’s d was calculated using data or statistics directly provided in the articles for the effect of the intervention. The effect size was calculated from post-intervention (or follow-up means as described below), standard deviations (SD), and sample size. When such information was not provided, Cohen’s d was converted from a variety of statistics: t-values, F-tests, chi-square, eta-squared, or regression coefficients. The Practical Meta-Analysis Effect Size Calculator (19–20) was used to convert the statistics and values provided in the articles into Cohen’s d. For dichotomous outcomes, odds ratios were calculated and then converted to Cohen’s d. When Cohen’s d was calculated from odds ratios or correlation coefficients, which are not available in the Practical Meta-Analysis Effect Size Calculator, or when Cohen’s d, but not variance, was reported in the article, the variance of Cohen’s d was calculated based on Cohen’s d and the sample size. Cohen’s d for correlation coefficients and odds ratios were estimated using relevant equations (20–21). Because of the variability in populations, interventions, and outcome measures between studies, a random effects model was used.

Statistical analyses were carried out using the Comprehensive Meta Analysis package (22). Because effect sizes are likely non-independent, when more than one effect size was reported in a single study (23), outcomes were averaged for each study so that each study contributed a single data-point to the analysis. This approach was taken due to the variability in the number and types of outcomes reported by each study, rendering a three- and four-level meta-analysis of effect sizes nested within studies underpowered (23). For studies with multiple time points, the time point at the end of the intervention was used to ensure consistency across studies. Analyses were conducted to examine effect sizes for subgroups that were divided based on variables such as type of intervention, type of HIV OVC, country, study design, and age of participants on intervention outcomes, in order to determine both the types of interventions that have been effective, but also where studies have been most effective and for what population. The heterogeneity of effects was quantified with the I² statistic (24).

The systematic search yielded 74 articles dating from 2001 to 2017. The interventions were conducted in 14 countries overall, including China, Haiti, Kenya, Malawi, Myanmar, Rwanda, South Africa, Uganda, the United States, and Zimbabwe. Studies were also conducted in Canada, France, India, and Tanzania, but these studies did not have enough information to be included in the meta-analysis. Figure 2 provides a map summarizing the sample sizes by cities or sub-regions. For purposes of illustration only, when studies did not specify specific regions, we assumed the region of data collection coincided with the city of the authors’ primary institution or with the capital city. Forty-three articles presented randomized controlled trials (RCT), while the rest presented quasi-experimental or cross sectional data. Sixty articles reported some form of pre-post data, while the remaining 14 presented cross sectional data. Only one study (25) discussed cost-effectiveness of interventions.

Figure 2.

Sample sizes by cities or sub-urban regions (total n = 32,454).

Results:

Studies investigated were psychosocial/behavioral interventions (k=44), cognitive (k=6), social protection (k=14), and community-based (k=9) interventions. Effect sizes were calculated from the data presented in the article if there was enough information (Table S1). The overall effect size for all studies was 0.33 (SE=0.03, 95% CI: 0.27–0.38, Figure 3). I-squared was equal to 75.87, indicating that there was significant heterogeneity across studies. In the next sections, we first describe the primary studies included in the review (the effect sizes corresponding to this section can be found in Table S1), and then present results of the meta-analysis.

Figure 3.

Effect sizes by intervention category

Summary of the current research on available interventions for HIV/AIDS OVC

Among psychosocial interventions, family therapy was most effective, but if the goals of the intervention are focused specifically on improving risk behaviors, motivational interviewing, more specific to risk behaviors, may be more targeted and effective. While semi-structured family interviews aided with family congruence of advanced care planning, they were associated with increases in negative psychosocial outcomes such as decreased emotional quality of life, lower faith and meaning/purpose. For cognitive interventions, effects were variable on specific measures of cognitive ability, such as attention (26), with some positive effects on visual reception (27–28), receptive language (28–29), and memory (27, 29).

Social protection interventions had positive effects on a variety of outcomes (Table 2), including savings, educational outcomes such as the taking of primary leaving examinations (PLE; a standardized test in Uganda) and scores, educational plans and aspirations, school attendance, and school grades (30–32), cognitive outcomes (33), psychosocial outcomes such as hopelessness, self-concept, and depression (31–32, 34–36), self-esteem (35) and HIV health related behaviors and sexual risk behaviors (30, 37–40). A qualitative study focusing on the integration of mentorship with financial support found that mentorship supports the youth’s ability to develop future educational plans (38). Karimli and Ssewamala (31) demonstrated a possible mediation effect of adolescent savings on self-concept. Therefore, through financial support, in combination with mentorship and financial workshops, these interventions were able to lead to improvements on many areas of vulnerability for HIV OVC.

Table 2.

Summary of Outcomes and Corresponding Effect Sizes

Intervention Type	Number of Studies	Outcomes Measured	Number with Reported Significant Results	Range of Effect Sizes
Psychosocial	44	Psychosocial	32/75	−0.27−1.70
		Health Behavior	19/45	−0.52−1.51
		Sexual Risk Behavior	7/26	−0.14−0.55
		Substance Use	7/17	−0.15−0.81
		Family	4/13	−0.08−1.05
		Social	11/17	0.05−1.31
		Adaptive	1/5	0.03−0.38
		Positive	7/25	−0.26−1.05
		Disclosure	0/6	−0.02−0.38
		Stigma	0/6	−0.11−0.11
		Religious	2/2	0.68−1.09
Cognitive	6	Psychosocial	0/6	−0.26−0.17
		Cognitive	18/64	−0.36−1.18
		Executive Functioning	1/14	−0.36−0.16
		Motor	1/8	−0.16−0.22
		Social Protection	2/2	0.30−1.35
Social Protection	16	Psychosocial	7/7	0.25−0.78
		Health Behavior	3/3	0.39−0.42
		Sexual Risk Behavior	6/9	0.01−0.77
		Positive	3/4	0.11−0.35
		Cognitive	6/6	0.15−0.93
		Education	11/13	0.05−1.09
		Social Protection	8/8	0.17−0.81
		Illness	0/4	−0.13−0.09
Community-based	9	Psychosocial	1/8	−0.06−1.18
		Health Behavior	3/3	0.24−0.50
		Family	2/2	0.19−0.48
		Positive	2/6	0.00−0.39
		Social Protection	4/7	0.01−0.48
		Home Visits	5/5	0.18−1.13

Variables such as depression, anxiety, and externalizing symptoms were classified as psychosocial. Variables including medication adherence and HIV knowledge were coded as HIV Health behaviors. Outcomes related to income, resources, and obtaining birth registration were coded as social protection outcomes. Social outcomes included variables related to perceptions of social support, social behavior, or number of friends. Positive outcomes were variables such as positive coping, perseverance, emotional intelligence, or happiness. Adaptive outcomes included daily living skills and functional impairment. Home visits described whether home visits were provided as part of the intervention.

For CBI studies, some variables measured were simply whether a service was received, not whether there were significant improvements in the functioning of the youth, and none of the studies were randomized controlled trials. Even though paraprofessional programs tended to provide more visits, more child interaction, and more tangible support to caregivers than volunteer programs (41), both paraprofessional and volunteer programs did not have positive effects on outcomes. Depression for boys, behavioral problems, and family functioning worsened over the course of the paraprofessional intervention (42). However, we do not suggest that worsened outcomes are related to potentially deleterious effects of paraprofessional interventions. While stigma was found to be a significant risk factor for psychosocial vulnerabilities (44), studies investigating the impact of intervention on stigma registered inconsistent results (25, 44), and the measures used were generally created or adapted for the studies. However, programs using a paraprofessional model were found to be more effective at linking families to social grants for children, associated with greater household food security and obtainment of resources (d=0.22–0.48) (45). One study comparing 4 CBI found improved food security for two of the interventions, but no effect for the other two (25). Therefore, results were overall mixed for CBI, and they tended to focus on providing basic resources without effecting significant improvements in psychosocial outcomes for HIV OVC.

Gender.

There were few studies that included analyses of gender effects, but those that did include such analyses reported significant gender differences for certain outcomes. For example, Eloff and colleagues (46) found that boys gained greater benefit from family group therapy than girls for externalizing behaviors, internalizing behaviors, and depression, indicating that gender may have an effect. In addition, two studies of peer group therapy found gender specific effects where stigma and negative self-image decreased for men, but were non-significant or actually increased for women (47–48). Therefore, studies that did investigate the effects of gender found that effects tended to be stronger for boys regarding psychosocial outcomes. With regard to risk behaviors, Thurman and colleagues (49) registered gender effects for sexual risk behaviors following a group therapy intervention, with a significant increase in condom use for girls and a decrease in risky sexual partnerships for boys, but no effect on sexual debut. Similarly, there were some gender specific effects related to sexual risk taking behavior in response to social protection interventions as well. While Goodman and colleagues (50) demonstrated that females had fewer sex partners (d=0.27) and were more likely to use a condom (d=0.30), these differences were not found among males. Conversely, another study revealed that approval of risky sexual behavior significantly decreased for boys but not girls (d=0.77; 51). Overall, these analyses demonstrate that interventions may differential impact on health or sexual risk behaviors for different genders.

Results of the Meta-Analysis

The most effective categories of intervention were the CBI, with an overall effect size of 0.36 (SE=0.05, 95% CI: 0.26–0.46, k=7; Table 1), and social protection interventions (d = 0.36, SE = 0.06, 95% CI: 0.25–0.47, k=13), followed by psychosocial interventions (d = 0.30, SE=0.05, 95% CI: 0.20–0.40, k=32). Of the psychosocial interventions, therapy involving the entire family had the largest effects, including individual family therapy and group family therapy (Table 1). Cognitive interventions had the lowest overall effect size (d = 0.14, SE=0.13, 95% CI: −0.13–0.41, k=5).

Table 1.

Effect Sizes by Intervention Type

Intervention Type	Effect Size
Psychosocial Interventions	Cohen’s d = 0.30 (SE=0.05, 95% CI: 0.20–0.40, k=32)
Education Sessions	Cohen’s d = 0.42 (SE=0.20, 95% CI: 0.02–0.81, k=1)
Group Family Therapy	Cohen’s d = 0.37 (SE=0.10, 95% CI: 0.18–0.56, k=10)
Individual Family Therapy	Cohen’s d = 0.53 (SE=0.13, 95% CI: 0.28–0.79, k=2)
Peer Group Therapy	Cohen’s d = 0.31 (SE=0.12, 95% CI: 0.07–0.55, k=6)
Home Based Nursing Intervention	Cohen’s d = 0.58 (SE=0.38; 95% CI: −0.16–1.32, k=1)
Motivational Interviewing	Cohen’s d = 0.38 (SE=0.07; 95% CI: 0.25–0.51, k=8)
Peer Mentorship	Cohen’s d = 0.07 (SE=0.06, 95% CI: −0.04–0.18, k=1)
Semi-Structured Family Interviews	Cohen’s d = 0.12 (SE=0.46, 95% CI: −1.03–0.79, k=3)
Cognitive Interventions	Cohen’s d = 0.14 (SE=0.13, 95% CI: −0.13–0.41, k=5)
Social Protection Interventions	Cohen’s d = 0.36 (SE = 0.06, 95% CI: 0.25–0.47, k=13)
Community-Based Interventions	Cohen’s d = 0.36 (SE=0.05, 95% CI: 0.26–0.46, k=7)

Table 2 presents a summary of outcomes and corresponding effect sizes. Psychosocial/behavioral interventions had moderate effects, while cognitive interventions had variable effects depending on the specific cognitive outcome measured. Paradoxically, although CBI and social protection interventions had the largest effects sizes, CBI had the weakest outcomes, while social protection interventions were most effective overall. This finding could be partly due to the fewer number of studies on CBI as well as fewer outcomes reported in these studies.

Sub-group Analyses

Country.

Effect sizes were descriptively examined by country (Table 3). Studies conducted in Myanmar and China had the largest overall effect sizes; however, there were few studies conducted in these countries overall. Most of the studies were conducted in South Africa (d=0.29, SE=0.03, 95% CI: 0.23–0.36, k=11), Uganda (d=0.37, SE=0.07, 95% CI: 0.24–0.50, k=16), and the United States (d=0.25, SE=0.08, 95% CI: 0.10–0.40, k=20). Overall, effect sizes were comparable across countries.

Table 3.

Countries Included and Corresponding Effect Sizes

Country	Number of Studies	Effect Size
China	3	0.61
Haiti	1	0.46
Kenya	2	0.16
Myanmar	1	0.90
Rwanda	2	0.53
South Africa	11	0.29
Uganda	16	0.37
United States	20	0.25
Zimbabwe	1	0.25

OVC type.

Effect sizes were descriptively examined by OVC type. OVC samples included HIV positive youth, youth living with a caregiver with HIV, children orphaned by AIDS, children orphaned or living with a caregiver with HIV, or a broader definition of children affected by HIV (Table 4). Studies using samples defined broadly as children affected by HIV often recruited participants from areas where HIV was common, or from children already participating in CBI, without defining specific inclusion criteria. Overall, effect sizes were largest for children orphaned by HIV, but they were comparable across OVC sample categories (Table 4).

Table 4.

Effect Sizes by Sample

OVC Type	Number of Studies	Effect Size
Children orphaned by HIV	14	0.43
Children whose caregivers have HIV	10	0.38
Orphaned by HIV or caregiver HIV	8	0.24
HIV positive youth	22	0.28
Broadly defined OVC	3	0.26

Age.

Table 5 shows how age ranges were broadly categorized by group; below 5, elementary (5–9), adolescence (10–18), and young adults (19 and over). Many studies had samples that spanned some of these defined age ranges (e.g. adolescence to young adult). Most of the studies focused on adolescence, the adolescence to young adult range, or the elementary to adolescence range (Table 4). Studies focused specifically on adolescence had the largest effect sizes (Table 4). Only three studies had a sample of children below the age of 5, with an overall effect size of 0.28 (SE=0.12, 95% CI: 0.04–0.51, k=3). Overall, studies tend to be focused on adolescents or elementary-aged children, and the greatest effects were found in studies focused specifically on adolescents.

Table 5.

Effect Sizes by Age Category

Age Category	Number of Studies	Effect Size
Below 5	3	0.12
Below 5 to Adolescence	1	0.38
Below 5 to Young Adult	1	0.58
Elementary to Adolescence	11	0.36
Elementary to Young Adult	1	−0.01
Adolescence	22	0.38
Adolescence to Young Adult	17	0.22

Study design.

Studies included were RCT, quasi-experimental longitudinal or cross sectional studies, randomized cross sectional studies, or pre-post studies with no control (Table 6). The largest overall effect size was registered for quasi-experimental longitudinal studies (d = 0.40, SE=0.05, 95% CI: 0.29–0.51, k=5), followed by pre-post studies (d=0.36, SE=0.09, 95% CI: 0.19–0.54, k=9), RCT (d=0.32, SE=0.04, 95% CI: 0.24–0.39, k=40), and quasi-experimental, cross-sectional studies (d=0.16, SE=0.09, 95% CI: −0.01–0.32, k=2). Cross-sectional, randomized studies had an effect size of 0.30 (SE=0.06, 95% CI: 0.19–0.42), but this was based on one study.

Table 6.

Effect Sizes by Study Design

Study Design	Number of Studies	Effect Size
Randomized Controlled Trials	40	0.32
Quasi-experimental, longitudinal	5	0.40
Quasi-experimental, cross-sectional	2	0.16
Pre-post (no control)	9	0.36
Randomized, cross-sectional	1	0.30

Discussion

Summary of Findings

Our meta-analysis of interventions for HIV OVC demonstrated that overall, interventions tended to have small to medium effect sizes. Social protection interventions tended to have the best outcomes when considering both overall effect size and the proportion of outcomes that were significant (Table 2). In addition, social protection interventions had positive effects on a broad range of outcomes, including cognitive, education, health behavior, and psychosocial outcomes. Cognitive, psychosocial, and community interventions also had overall small to medium effect sizes, but a smaller proportion of outcomes had significant effects compared to social protection interventions (Table 2). For psychosocial interventions, family interventions tended to have a greater effect, while motivational interviewing interventions had positive effects specifically for risk behaviors. Cognitive interventions tended to have some positive effects only on a few specific cognitive outcomes, without significant generalization to other areas of cognition or psychosocial functioning. For community-based home-visit and kids’ club interventions, programs using a paraprofessional model were able to provide more tangible support compared to volunteer programs (42–43, 45), but the authors reported that neither paraprofessional nor volunteer programs had significant positive effects on outcomes; in fact, psychosocial, behavioral, and family outcomes often appeared to worsen over the course of time (42).

Areas for Future Research

Based on the results of this meta-analysis, we suggest and discuss three areas of future research below: 1) to examine the heterogeneity of impact to identify vulnerable subgroups, 2) to study the cost-effectiveness of interventions to identify intervention components that can be provided to HIV OVC at low cost, and 3) to identify the most effective intervention characteristics that could be feasibly incorporated into community-led interventions to improve their outcomes;

1) Examine Heterogeneity of Impact to Identify Vulnerable Subgroups

An important area of future research would be to identify the heterogeneity of impact for subgroups, because interventions may affect subgroups (e.g. male versus female, younger children versus older children) differently. Therefore, identifying the impact on these specific subgroups will help to identify which populations are best served by a certain intervention, or to identify vulnerable subgroups who are not benefitting from a particular intervention. This knowledge will allow interventionists to refine and adapt interventions depending on the population of interest.

For example, gender is one area of vulnerability that may affect individuals’ levels of risk or how they respond to interventions. In sub-Saharan Africa, four in five new infections among adolescents aged 15–19 years are in girls, and young women aged 15–24 years are twice as likely to be living with HIV than men (1). Therefore, there are clear gender differences in vulnerability to infection depending on gender. Intervention studies that did examine the effects of gender found that there were differential effects on psychosocial outcomes or sexual risk behavior depending on gender (46–49), but few studies measured these gender effects. Additionally, there was insufficient reporting on the effects of gender on psychosocial outcomes, even though psychosocial outcomes tend to vary by gender (52).

In addition, a critical gap in the literature is the investigation of impact of interventions on younger children. There were very few studies that included younger children, especially pre-school aged children. The vast majority of interventions enrolled adolescent children. However, previous research has demonstrated that psychosocial, behavioral, and cognitive outcomes may be improved if difficulties are addressed earlier (53). For studies that included a broad range of ages, there was no investigation of whether the interventions may have differential effects on younger as compared to older children. However, interventions such as psychotherapy may have different effectiveness for younger compared to older children (54). Therefore, more research is needed to identify and develop interventions that are effective for younger children, and determine which interventions may be more appropriate for children of different ages.

Our study revealed that effect sizes tended to be overall similar for different samples of OVC, including samples limited to a specific subcategory such as children orphaned by AIDS, or more broadly defined and inclusive samples (Table 4). However, studies that used more broadly defined samples did not investigate whether different interventions exerted differential effects. It is possible, however, that a therapeutic intervention may have different effects depending on the type of OVC targeted. For example, Ssewamala and colleagues found that single orphans had higher levels of self-concept than double orphans following intervention (32). These results indicate that various subcategories of OVC may respond differently to intervention, depending on how the sample is defined. It is likely that children who are orphaned by AIDS may have different vulnerabilities compared to those who are living with a caregiver with HIV or those who are HIV positive. Because studies defined their populations differently, it was difficult to fully disentangle differences between subgroups.

2) Cost-Effectiveness

As HIV/AIDS continues to be a global pandemic, significant active funding and research is being directed towards improving this population’s outcomes. The U.S. President’s Emergency Plan for AIDS Relief (PEPFAR), although not the only source of international funding for HIV, represents the majority of U.S. global health funding (62% in FY 2019, or $6.8 billion), with 10% of the budget addressing the needs of OVC (55). However, only one study included a cost-effectiveness analysis, making it impossible to evaluate the practicability and generalizability of most interventions. The study that did include this analysis compared the cost effectiveness of 4 CBI and found that food support and income generation, individual counseling, kids’ clubs, and school-based HIV education were most cost effective, while home visiting and guardian support groups had a higher cost (25). This research, although 9 years old, demonstrates that some components of interventions such as income generation, counseling, and HIV education have the potential to provide benefits to OVC at a low cost. Considering the significant funding being directed towards interventions for HIV OVC, more studies should include analyses of cost-effectiveness to determine which interventions provide the greatest benefits at lower costs, increasing access for this vulnerable population.

3) Community-Based Interventions

CBI are a relevant area of research because they use an existing structure and are carried out by community members themselves. CBI are carried out in the context of the community and adapted to local conditions, empowering individuals in the community to have control and ability to solve their own problems (15). Results from this meta-analysis demonstrate that the existing CBI had positive effects on fewer outcomes than other types of interventions. Therefore, more research should be conducted on CBI to improve the quality of community-based, home-visiting interventions.

Based on the results of this meta-analysis, we can identify effective components of interventions to identify the characteristics that most improved their outcomes. For example, for interventions with a psychosocial component, therapy with the whole family tended to lead to better outcomes, while motivational interviewing was beneficial specifically for improving risky behaviors. Training programs to improve the community’s ability to provide effective therapeutic intervention, and incorporating effective components flexibly depending on the needs of the HIV OVC would likely lead to improvements in areas of vulnerability. In addition, the results of this meta-analysis suggest that unconditional or conditional cash grants tended to have the most significant positive effects on a broad range of areas of vulnerability. These social protection interventions were able to be implemented through partnerships between local families, civil society, the private sector, and the government, carried out within the community itself (32). Therefore, if CBI are able to partner with these entities to provide financial support in combination with mentorship and financial management workshops, it may improve the effectiveness of CBI in all high risk countries and regions. Given the multicomponent nature of CBI, future research may benefit from incorporating some of the effective components of other types of interventions to determine if they could improve outcomes for CBI. In addition, none of the home-visit CBI were randomized-controlled trials, and they tended to define OVC broadly depending on who was already participating in CBI, rather than clearly defining or examining effects on sub-populations of OVC (i.e. children orphaned by AIDS). More rigorous research should be conducted in these communities to better improve CBI for OVC.

Limitations of Existing Studies

There were significant limitations in the study designs or measures in the studies included. Sixteen studies did not include enough information to calculate effect sizes, and therefore were not included in the meta-analysis. In addition, only about two thirds of the studies included long-term follow-up assessments, preventing us from determining if the positive effects from interventions were lasting. With regard to measurement outcome, for many of the studies, most outcomes were measured through self-report or parent-report, which could lead to some bias in measurement; few studies utilized maximum performance assessments. Some well-validated instruments tended to be used often in the studies (e.g. Achenbach Child Behavior Checklist; Achenbach, 1999), but many of the studies used single questions to measure outcomes of interest (e.g. rating confidence in educational plans or rating health as good or excellent) or inconsistent measures. Stigma, a variable found to be highly relevant to OVC (58–60), was not evaluated at all in the majority of the intervention studies. When it was included, stigma was evaluated with variable instruments, preventing the direct comparison of intervention effects on stigma between studies. Overall, the quality of the assessments used was inconsistent, and not enough of the studies used comparable instruments. Lastly, a limitation of the current review is that the literature is three years old, so additional studies may have been published since this time.

Summary

The results of this systematic review and meta-analysis indicate that further research on interventions for HIV OVC is necessary to further develop and improve interventions for HIV OVC. Overall, interventions addressed the areas of difficulty that HIV OVC tend to experience: mental health issues, cognitive deficits, and a higher incidence of risk behavior. However, more studies need to be conducted to determine how the effective components of existing interventions, such as cash grants and mentorship within social protection interventions, can be incorporated into communities to provide long-lasting support for HIV OVC. In addition, OVC intervention research should be expanded to include younger children and to examine the impact on various vulnerable subgroups. Further examination of these variations within studies will provide greater understanding of who certain interventions may be most effective for and why. Lastly, more analysis of cost effectiveness of interventions is necessary to determine practicability of implementation of various interventions.