Introduction

In 2019, the United Nations estimated that approximately 1.7 million children around the world were living with HIV, of which 88% reside in the region of Sub-Saharan Africa, SSA [1]. Enormous progress towards reducing morbidity and mortality among youth living with HIV has shifted the focus of treatment to chronic complications and comorbidities, including mental health problems [2, 3]. Despite this, existing studies in SSA are few and have produced mixed results regarding the prevalence of psychological impairment amongst HIV-infected (HIV+) youth, with estimates ranging between 17 and 80% across the SSA region [4, 5]. Further, most of the globally published data focus predominantly on behaviorally-infected youth, with the mental health of perinatally HIV-infected (PHIV+) youth being poorly characterized [4, 6]. PHIV+ youth may face more significant challenges relative to behaviorally-infected youth due to early exposure to risk factors throughout crucial stages in development [7, 8].

Perceived HIV-related stigma has been identified as a key challenge amongst behaviorally-infected youth [9], with negative effects on psychosocial and treatment outcomes [10,11,12]. However, the effects of HIV-related stigma among PHIV+ youth have not been well studied in SSA, with existing studies providing contrasting results [4, 10, 13,14,15]. Associations between demographic or clinical characteristics and psychological outcomes similarly differ across studies. Studies from the United States (US) and South Africa have found that younger age and female sex are associated with worse psychological outcomes in PHIV+ youth, whereas no associations were observed between these factors and psychological outcomes in other studies [5, 16,17,18]. Similarly, low CD4 cell count has been found to be associated with depressive and conduct-related symptoms in high- and low-income settings, with other studies from the same contexts providing contradictory findings [3, 19,20,21,22]. Given this, identifying the context-specific socio-demographic and clinical predictors of poor psychological adjustment in high-burden settings is crucial to be able to identify high-risk adolescents and ensure that the most vulnerable youth receive appropriate support [5, 6].

In addition to HIV-related stigma, there is some evidence to suggest that poor mental health itself may lead to adverse treatment outcomes in HIV+ youth, with rates of suboptimal antiretroviral therapy (ART) adherence ranging anywhere from 18 to 50% in those with documented mental health symptoms [10, 18, 23]. Adolescents demonstrating suboptimal adherence in the US, for example, have been observed to have higher depression scores relative to their adherent counterparts [18, 24, 25]. However, other literature from the US has suggested that poor treatment outcomes may be more dependent on structural socio-environmental barriers (for example, problems with medical insurance, or increased school or family responsibility) than on psychological health in PHIV+ youth [26]. With an estimated 50% of lifelong mental illness beginning in adolescence and a dearth of existing literature in South Africa, there is a crucial need to further explore the relationship between psychological health and treatment outcomes in PHIV+ youth as a way of ensuring positive health outcomes throughout the lifespan [24, 25, 27]. We thus examined factors associated with psychological adjustment, including the role of HIV-related stigma, and explored the relationship between psychological adjustment and treatment outcomes in PHIV+ youth in Cape Town, South Africa.

Methods

Study Design

This analysis draws on data from a larger cohort of 515 PHIV+ youth aged 9–14 years at enrollment and followed as part of the Cape Town Adolescent Antiretroviral Cohort (CTAAC). Youth were recruited from 7 ART services throughout Cape Town and were eligible for enrollment if they had been on ART for at least 6 months and knew their HIV-status. Follow-up visits were conducted every 6 months after enrollment, through 48 months. All study measurement visits were conducted separately from routine HIV care. Measures for this analysis were gathered from the 3rd (12-month after enrollment), 4th (18-month after enrollment), and 5th (24-month after enrollment) follow-up visits as described in the sections that follow. Informed consent and assent were obtained from caregivers and adolescents prior to enrollment. Approval for the CTAAC study was obtained from the Faculty of Health Sciences Human Research Ethics Committee of the University of Cape Town (HREC Ref. 215/2013).

Measures

All adolescent- and caregiver-reported measures were administered by trained interviewers in the participant’s preferred language. Demographic data, including age, gender and educational attainment, were collected at enrollment in the CTAAC study. Socioeconomic status (SES) was assessed based on housing type, household assets, and caregiver employment status using a composite poverty score that was developed following the approach used in the South African Stress and Health Study [28]. SES was categorized into Most, Moderately, and Least Disadvantaged groups using distribution-based cut-offs to facilitate comparisons within the study sample. Clinical information was collected via medical record abstraction from routine ART sites at enrollment and included age at ART initiation and ART regimen history.

Psychosocial Measures

Psychosocial measures were assessed at the 4th follow-up visit (18 months after enrollment).

HIV-related stigma among adolescents was assessed using a 10-item scale from a family-based intervention developed and piloted in South Africa [29]. The scale quantifies perceived HIV-related stigma by exploring themes centered around self-esteem and the perceived consequences of disclosure. Each item was assessed on a scale ranging from 1 to 3 based on the corresponding response options of “Not at all”, “Sometimes”, and “All the Time”. The mean response was calculated across items, where higher scores indicated higher levels of perceived HIV-related stigma [Cronbach’s alpha (α) = 0.87].

The Strengths and Difficulties Questionnaire (SDQ) was administered separately to adolescents and caregivers as a measure of psychological adjustment. The SDQ is a 25-item instrument that is divided into five subscales measuring emotional symptoms (Cronbach α for adolescent/caregiver report, respectively [α] = 0.65/0.69), conduct problems (α = 0.46/0.57), hyperactivity inattention (α = 0.49/0.56), peer problems (α = 0.36/0.56), and prosocial behavior (α = 0.55/0.56); difficulty subscales can be summed into a total difficulties score (α = 0.77/0.81). The inventory is widely used in child mental health research and has been validated in several cross-cultural settings, including in the Sotho-speaking population in South Africa, thus serving as a suitable measure for gauging the emotional and behavioral well-being of PHIV+ youth [30, 31]. Response options for each item are “Not True”, “Somewhat True”, and “Certainly True”, with each item scored between 0 and 2 [32]. Items on each subscale can be summed, with higher scores indicating greater difficulties on the emotional symptoms, conduct problems, hyperactivity-inattention, peer problems, and total difficulties subscales, and higher scores indicating greater strengths on the prosocial behavior subscale [32].

Information on the relationship between adolescents and their healthcare providers was collected using a 17-item patient–provider relationship scale which gauged the adolescent’s experience in the clinical setting and their consistency of contact with the same health personnel. This self-reported questionnaire was developed and validated in the South West Tshwane sub-district of Pretoria, South Africa using similar tools from the US as a model [33]. The scale correlates with global measures that indicate an association between patient characteristics, increased adherence to ART, and having an undetectable viral load [33]. Response items were selected based on the relevance of each item to CTAAC study objectives and were scored from 1 to 4, ranging from “Always” to “Never” [33]. A mean score was then calculated across items, where higher scores indicated a better patient–provider relationship.

Adherence and HIV Viral Load

At enrollment and then annually, adolescent-reported adherence was assessed using structured questionnaires, and HIV viral load testing (Abbott RealTime HIV-1) was conducted by the South African National Health Laboratory Services. For the purposes of this analysis, viral load and adolescent-reported adherence measures were collapsed into a single variable from data collected at the 3rd (12 months after enrollment) and 5th (24 months after enrollment) study measurement visit, as these measures were not assessed at the 4th study measurement visit. Given the dynamic nature of adherence behaviors over time, we compared participants who reported suboptimal adherence or had an elevated viral load at either visit with those who reported optimal adherence or were virally suppressed at both visits. This was used to give a general sense of adherence and viral suppression at the 4th study measurement visit.

Data Analysis

Data were analyzed using STATA Version 14 (Stata Corporation, College Station, Texas, USA). Summary statistics in the form of medians and proportions were used to describe socio-demographic and clinical characteristics. For the purposes of this analysis, each subscale of the SDQ scale, including the total difficulties subscale, was assessed separately. Multivariable linear regression was used to examine associations between HIV-related stigma and SDQ scores, adjusting for socio-demographic and clinical characteristics. Multivariable logistic regression models were used to examine the impact of SDQ scores and HIV-related stigma on elevated viral load and adolescent-reported suboptimal adherence to ART. Elevated viral load was defined as viral load ≥ 50 copies/mL at either the 3rd or 5th study measurement visit, and suboptimal adherence as reporting missing ART dose(s) on ≥ 1 day(s) during the last 30 days at either of these visits. Sensitivity analyses were conducted to assess the consistency of results when elevated viral load was defined as ≥ 1000 copies/mL and suboptimal adherence as reporting missing ART dose(s) on ≥ 2 days during the last 30 days.

Results

Socio-demographic and Clinical Characteristics

A total of 515 PHIV+ youth were enrolled in the overarching CTAAC study between July 2013 and March 2015. Of this sample, 92% (n = 473) attended the 18-month follow-up visit and completed the adolescent-reported SDQ; 97% (n = 461) of caregivers completed the SDQ. No appreciable differences in socio-demographic or clinical characteristics were observed between youth who did and did not attend the 18-month follow-up visit. The median age at 18-month follow-up was 13.6 years, and 50% of adolescents were male (Table 1). At the time of the 18-month follow-up visit, the median duration on ART was 7.8 years.

Table 1 Demographic and clinical characteristics for perinatally HIV-infected (PHIV+) youth
Full size table

HIV-Related Stigma

Of youth with completed adolescent-reported SDQ’s, 99.8% (n = 472) also completed the HIV-related stigma questionnaire. Overall, participants did not frequently report experiencing HIV-related stigma. In most cases, participants did not report perceived discrimination, fear from others, or a perceived doubt of their integrity. However, 10% did report having been teased [“Sometimes” n = 33 (7%), “All the Time” n = 12 (3%)], 9% experienced feeling different and alone [ “Sometimes” n = 30 (6%); “All the Time” n = 13 (3%)], and 8% stated that they avoided making new friends [“Sometimes” n = 21 (4%); “All the Time” n = 20 (4%)] or that people have gossiped behind their backs [“Sometimes” n = 33 (7%); “All the Time” n = 6 (1%)]. The mean score for HIV-related stigma responses was 0.1 [Standard Deviation (SD) = 0.2] based on the 1–3 scoring scale described earlier (Table 2).

Table 2 Adolescent report of HIV-related stigma
Full size table

Patient–Provider Relationship

Adolescents generally indicated positive experiences when interacting with their healthcare providers for regular ART follow-up visits. Participants reported a median score of 3.5 [Interquartile Range (IQR): 3.5–3.58] on the patient–provider relationship scale, indicating an overall satisfaction with their relationships to their regular healthcare providers (Table 1). The vast majority of participants reported feeling listened to (90%) or cared about (95%) by their healthcare providers, even though a similar proportion felt some fear when speaking to their providers about their problems (95%).

Stigma and SDQ Scores

All adolescent and caregiver reported inventories were significantly correlated (p < 0.001) with a moderate linear trend for each subscale (Spearman’s ρ 0.28–0.56). No significant differences were found between median reported total difficulties for PHIV+ youth amongst adolescent (5.86; IQR 2–9) and caregiver (6.06; IQR 2–9) reported scores. Unadjusted linear regression revealed significant associations between adolescent reports of perceived HIV-related stigma and each of the adolescent and caregiver-reported difficulty subscales. After adjustment for socio-demographic factors, the associations between perceived HIV-related stigma persisted for all adolescent (Table 3) and caregiver (Table 4) reported difficulty subscales.

Table 3 Linear regression models comparing the relationship of perceived HIV-related stigma with adolescent-reported strengths and difficulties scores among perinatally HIV-infected (PHIV+)
Full size table
Table 4 Linear regression models comparing the relationship of perceived HIV-related stigma with caregiver-reported strengths and difficulties scores among perinatally HIV-infected (PHIV+)
Full size table

On average, higher perceived HIV-related stigma was significantly associated with an 8-point increase on the adolescent-reported total difficulties subscale [Adjusted β Coefficient (Adj.) 8.81, 95% confidence interval (CI) 7; 10.63, p < 0.001] and a 7-point increase on the caregiver-reported total difficulties subscale (Adj. 7.85, 95% CI 5.83; 9.87, p < 0.001). Similarly, adjusted models for the prosocial strengths subscale revealed that higher perceived HIV-related stigma was significantly associated with lower scores on the adolescent-reported subscale (Adj. − 0.72, 95% CI − 1.26; − 0.17, p < 0.05). These associations persisted when stratified by age and sex.

Findings also revealed some associations between SDQ scores and demographic characteristics. Older age at follow-up was associated with lower adolescent-reported hyperactivity scores (Adj. − 0.11, 95% CI − 0.19; − 0.03, p < 0.05) and higher adolescent-reported prosocial scores (Adj. 0.11, 95% CI 0.04; 0.19, p < 0.05). Higher SES was associated with lower scores on the adolescent (Adj. − 0.5, 95% CI − 0.84; − 0.16, p < 0.05) and caregiver (Adj. − 0.64, 95% CI − 1.07; − 0.21, p < 0.05) reported hyperactivity scales and the caregiver reported conduct (Adj. − 0.54, 95% CI − 0.92; − 0.17, p < 0.05] and total difficulties subscales (Adj. − 1.61, 95% CI − 2.77; − 0.46, p < 0.05). Male sex was also associated with lower scores on the caregiver reported conduct (Adj. − 0.36, 95% CI − 0.63; − 0.04, p < 0.05) and hyperactivity (Adj. − 0.49, 95% CI − 0.84; − 0.15, p < 0.05) subscales.

SDQ Scores, Stigma and Treatment Outcomes

The relationship between SDQ scores and treatment outcomes was examined in a sub-analysis of 428 youth with complete viral load and adherence data (Supplementary Tables V, VI, VII, VIII), with no significant differences in socio-demographic or clinical characteristics observed relative to youth who did not have full adherence and viral load data. After adjustment for age at follow-up, sex, SES, and patient–provider relationship, it was found that a one-point increase in the caregiver reported conduct subscale was associated with a 20% increased likelihood [Adjusted Odds Ratio (Adj. OR) 1.2, 95% CI 1.06; 1.4, p < 0.05] of elevated viral load defined as ≥ 50 copies/mL (Table V). Analysis of adolescent-reported adherence in models adjusted for age, sex, and SES revealed that a one-point increase in adolescent-reported prosocial strengths was associated with a 15% decreased likelihood (Adj. OR 0.85; 95% CI 0.74; 0.99, p < 0.05) of reporting a missing ART dose(s) for ≥ 1 day(s) in the last 30 days (Table VI). No other associations between SDQ and viral load or SDQ and suboptimal adherence were found among adjusted models for either adolescent- or caregiver-reported inventories, even after stratification for age and sex. Similarly, an analysis of the relationship between perceived HIV-related stigma and treatment outcomes found no associations. Sensitivity analyses were conducted to assess the consistency of results when elevated viral load was defined as ≥ 1000 copies/mL and sub-optimal adherence as reporting missing ART dose(s) on ≥ 2 days during the last 30 days. Findings remained consistent between adolescent-reported prosocial strengths and adolescent-reported adherence (Table VIII), yet did not persist between elevated viral load and caregiver-reported conduct scores (Table VII).

Discussion

These data are among the first to use the SDQ to assess psychological adjustment in PHIV+ youth in SSA. Findings from this study revealed that HIV-related stigma was significantly associated with all adolescent- and caregiver-reported SDQ difficulty subscales and with adolescent-reported prosocial strengths. Exploration of other possible associations with SDQ scores found significant associations between socio-demographic factors and the adolescent-reported hyperactivity and prosocial scales, as well as the caregiver-reported hyperactivity, conduct, and total difficulties scales. Amongst clinical characteristics, sub-analysis of adolescent- and caregiver-reported SDQ scores and treatment outcomes identified associations between worse scores on the caregiver-reported conduct subscale and elevated viral load (which did not remain consistent after sensitivity analysis), as well as the adolescent-reported prosocial subscale and adolescent report of suboptimal ART adherence (which remained consistent after sensitivity analysis).

The finding that HIV-related stigma is associated with symptoms of psychological distress is consistent with outcomes from other studies conducted among PHIV+ youth in the SSA region [4, 13, 34]. Healthy cognitive and behavioral functioning has been identified to be closely related to a well-regulated self-concept, which can be significantly affected by a high degree of perceived illness-related stigma [4]. Perceived HIV-related stigma prompts internalized fear of gossip, of physical assault, or of discrimination in family and community circles, which can manifest in internalized and externalized symptoms of mental disorders [35]. This is particularly troubling when considering that 1 in 10 adolescents who completed the HIV-stigma questionnaire in this study reported having been teased. Both internal and external awareness of an individual’s HIV-status is also constantly reinforced through daily consumption of medication, frequent medical visits, and witnessing other HIV-related incidences, such as deaths in the household, which are triggers for mental health difficulties [34]. This relationship is exemplified by the consistent and significant effect of HIV-related stigma on SDQ scores across adolescent- and caregiver-reported subscales, despite the relatively infrequent reporting of stigma-related experiences by youth in this study.

We observed few associations between socio-demographic or clinical variables and SDQ scores, consistent with findings from several studies in the US [18, 21, 22, 36]. However, it is important to highlight some of the exceptions identified in this study. For example, the finding that male sex and higher SES were significantly associated with lower scores on the caregiver-reported hyperactivity and conduct subscales contradicts findings by Mellins et al., but is consistent with other literature on adolescent conduct disorders [37]. It is possible that families with more material resources (i.e., better SES) may be more involved in their children’s lives in a supervisory or disciplinary role, such that they have fewer perceived symptoms of hyperactivity or conduct problems [38]. Similarly, the finding that older age was associated with lower adolescent-reported hyperactivity and higher prosocial behavior scores, whereas better SES was associated with lower caregiver-reported total difficulty scores may suggest that older youth may have improved inhibition and that youth with better SES may have fewer perceived total difficulties due to reduced neighborhood-related stressors (i.e., violence, crime, etc.) [17]. Certainly, low-SES individuals may face compounded stressors through contextual and socially regulated factors, such as increased exposure to poverty, orphanhood, or inconsistent guardianship, all of which are frequently found in the South African context [39]. These findings reinforce the argument that psychological health in this population is tied to socio-environmental influences [4].

Although we observed few associations between SDQ scores and treatment outcomes, we observed associations between caregiver reports of conduct difficulties and elevated viral load, as well as poorer adolescent-reported prosocial strengths and suboptimal ART adherence. The finding that increased conduct problems are associated with a greater likelihood of elevated viral load is consistent with evidence from studies in the US and Rwanda, where HIV+ children with noted conduct problems were significantly more likely to report suboptimal adherence to treatment [40, 41]. Similarly, studies in the US and SSA have found aspects of better caregiver–child relationships to be linked with better adherence and prosocial strengths in PHIV+ youth, which may explain the associations found here [4, 42, 43]. Given the inherently low level of self-reliance experienced by many youth and adolescents, it is possible that adherence behaviors may be more heavily effected by youth–caregiver relationships than by the youth’s mental health. It is also possible that the high prevalence of self-reported positive relationships between patients and their providers could play a role in treatment outcomes. Previous studies have found that patients who feel understood and cared for by their providers had higher rates of medication adherence and undetectable viral loads, which may also influence the associations discussed here [44].

Several limitations deserve emphasis. First, the cross-sectional nature of this study limits causal conclusions in the relationships between SDQ scores and socio-demographic factors or treatment outcomes. Second, the internal consistency observed in some of the SDQ subscales used in this population are slightly lower than those found in some western cohorts, although they were similar to those found in other South African cohorts of a similar age range [30, 45, 46]. This may lower the reliability of the SDQ for measuring psychological adjustment in this cohort and skew overall scores, influencing the relationships observed in this study. Moreover, the sample size for the viral load and ART adherence sub-analysis was smaller than that used for the main analysis, bringing into question the power of statistical analysis to make meaningful comparisons between the groups. The adolescent-reported nature of the SDQ and adherence measures may result in errors inherent to self-reported measures, such as the under/over reporting of behaviors or social desirability bias. Finally, this study did not assess caregiver-related factors that have shown to be related to psychological impairment in PHIV+ youth, such as caregiver mental health status, which may play an important role in assessing the relationships between SDQ scores, stigma, and treatment outcomes [5, 16].

Conclusion

Perceived HIV-related stigma may serve as an important trigger for psychological adjustment issues amongst PHIV+ youth in the South African context. These issues may promote behaviors that are not conducive to positive health outcomes for people living with HIV, such as difficulties with attention or difficulties establishing connections with peers. Poor prosocial strengths and conduct problems may serve as potential indicators for suboptimal adherence, highlighting the importance of understanding psychological adjustment as a way of promoting treatment outcomes. These findings outline the need for further investigation of the socio-environmental factors that may impact HIV-affected South African youth, drawing specific attention to the need for interventions designed around reducing illness-related stigma. Ultimately, these findings highlight the importance of psychosocial health in achieving positive health outcomes amongst HIV-affected youth overall, and they add to the argument for more integrated use of psychosocial metrics and interventions aimed at reducing illness-related stigma across HIV care in South Africa.