Family Social Support Mediates the Relationship Between the COVID-19 Pandemic and Psychosocial well-being in a Cohort of Young South Africans Newly Diagnosed with HIV

Abstract

Poor psychosocial well-being, including depression, anxiety, low self-esteem, and high anticipated stigma, complicates young South Africans’ engagement with HIV care. During the COVID-19 pandemic, the psychosocial well-being of young South Africans with HIV may have been impacted by changing levels of social support. This analysis sought to examine whether social support mediates the relationship between the pandemic and psychosocial well-being in young South Africans with HIV. This secondary analysis compared baseline data from two cohorts of young people ages 18–24 who tested HIV positive either before or during South Africa’s COVID-19 State of Disaster. Baseline sociodemographic, social support-related, and psychosocial data were analyzed using linear regression and mediation analyses. We found that self-esteem was higher (χ² = 9.955, p < 0.01) and anticipated stigma (χ² = 22.756, p < 0.001) was lower in the cohort recruited during the pandemic. Perceived family social support was higher in the cohort recruited during the COVID-19 pandemic (χ² = 38.69, p < 0.001). Family social support partially mediated the relationship between study cohort and self-esteem (Sobel z=−3.04, p = 0.002), family- (Sobel z=−4.06, p < 0.001) and community-type (Sobel z =−3.44, p < 0.001) anticipated stigma, and depressive symptoms (Sobel z =−2.80, p = 0.005). Overall, compared to young people diagnosed with HIV before the pandemic, young people diagnosed during the pandemic reported higher self-esteem and lower anticipated stigma, an effect mediated by higher levels of family social support. Our findings add to the literature examining young people’s psychosocial well-being during the COVID-19 pandemic and suggests that improvements in family support may have broadly positive effects on multiple indicators of psychosocial well-being.

Introduction

With 8.45 million people living with HIV (PLWH) and 20% of the world’s new infections, South Africa remains the epicenter of the global HIV epidemic [1–2]. This burden of disease falls largely on adolescents and young adults (AYA), who are both disproportionately infected with and affected by HIV [3]. Though South Africa runs the world’s largest antiretroviral therapy (ART) program, only 52% of AYA with HIV are on ART, and less than half of PLWH between the ages of 15 and 24 are virally suppressed [4–5]. In addition to the numerous developmental [6–7], social [8–9], and structural [10–11] barriers that complicate HIV initiation, adherence, and viral suppression amongst AYA, indicators of poor psychosocial well-being, such as anxiety, depression, low self-esteem, and anticipated stigma, have been shown to be associated with poorer biomedical outcomes in this population [12–16].

The psychosocial well-being of AYAs living with HIV may have in turn been negatively impacted by the effects of the COVID-19 pandemic and resultant social dislocations [17–19]. Research suggests that the COVID-19 pandemic led to disruptions in HIV care globally [20–21]; however, the social and psychological effects of the pandemic on AYA with HIV have been less clearly delineated. It is theorized that COVID-19 related social control measures, including quarantines, school closures, and physical and social distancing, may have worsened psychosocial outcomes amongst AYAs living with HIV potentially by disrupting health-enabling social relationships outside of the home [22–24]. This may be especially true in the context of South Africa’s strict COVID-19 State of Disaster, which began in March 2020 with a nationwide lockdown characterized by strict home confinement, school closures, and cessation of non-essential business activity, all of which only gradually eased across the subsequent few years [25–26].

Declining levels of social support may be one mechanism by which these societal disruptions manifested in the psychosocial well-being of young adults newly diagnosed with HIV during the COVID-19 pandemic. Nevertheless, while social support has been reliably related to lower overall mortality [27–28], morbidity [29], and improved psychosocial well-being [30–32], including in patients with HIV [33–34], significant disparities in perceived social support during the pandemic have been reported [35–37]. On one hand, several studies investigating the relationship between the pandemic and psychosocial well-being have highlighted a relationship between pandemic-era declines in social support and higher levels of depression [38–40] and anxiety [41–42].

On the other hand, a few studies have suggested improved levels of family cohesion, cooperation, and social support [43–44] amidst COVID-19-related lockdowns. None of these studies included individuals coping with a new diagnosis of HIV, and to our knowledge, there have been no studies investigating the relationship between the pandemic, social support, and psychosocial outcomes amongst young people newly diagnosed with HIV. Thus, given both the mixed results of studies evaluating social support during the COVID-19 pandemic and the lack of data from AYAs living with HIV, the way in which psychosocial well-being was affected by the COVID-19 pandemic through changing levels of social support remains unclear.

To elucidate the relationship between the COVID-19 pandemic, social support, and psychological well-being in AYA with HIV, our present research had two general goals: (1) to compare the psychological well-being and levels of social support between a pre- and intra-pandemic cohort of AYA newly diagnosed with HIV; and (2) to determine whether levels of family and community social support mediated the relationship between study cohort and psychological well-being. In particular, it was hypothesized that relative to AYA newly diagnosed with HIV pre-pandemic, those diagnosed during the pandemic would exhibit lower family and community social support as well as lower baseline measures of psychological well-being, such as anxiety, depression, anticipated stigma, and self-esteem. In addition, it was hypothesized that family and community social support would significantly mediate the relationship between study cohort and psychosocial well-being in this population.

Methods

Participants

This secondary analysis utilizes data from baseline surveys taken from a sequential prospective cohort study and a subsequent pilot randomized controlled trial, which were conducted as part of a larger study called Standing Tall. The first study was conducted in Cape Town in 2018–2019, and the second study enrolled participants in Cape Town and East London between October 2020 and April 2022. Recruitment proceeded throughout this period without any pauses. Each study recruited 100 participants ages 18–24 who tested positive for HIV after initially presenting to well-established mobile and traditional clinics for HIV testing. Baseline interviews investigating primary and secondary outcomes of interest were conducted at the time of study enrollment in both studies.

To be eligible for participation in the first parent study, individuals had to be ART naïve, speak English and/or isiXhosa, and reside in the Cape Town, South Africa metro area. Women who were currently pregnant, positive with HIV and TB, and persons under the age of 18 were excluded from the parent study. Inclusion and exclusion criteria were the same in the second parent study, except that eligibility expanded to patients living in the catchment areas of three public clinics based in the East London metro area.

Measures

Primary Predictor: Time of Diagnosis with Relation to Pandemic

In this analysis, time of diagnosis with relation to the pandemic was utilized as the primary predictor. Study participants were classified into two groups (pre- or intra-pandemic) based on whether they were part of the 2018–2019 Standing Tall cohort or the 2020–2022 Standing Tall cohort. Note that participants recruited into the intra-COVID cohort were recruited at some point during the 2020–2022 COVID-19 State of Disaster.

Study Outcomes and Mediators

Four psychosocial outcomes were assessed.

1. Self-reported depression. The 9-item Patient Health Questionnaire (PHQ-9) was used to measure the severity of depressive symptoms at the baseline timepoint [45]. The PHQ-9 has validity and reliability when used in PLA in sub-Saharan Africa [46]. The 9 items reflect symptoms assessed in the diagnosis of clinical depression. Responses to each item ranged from 0 (‘not at all’) to 3 (‘nearly every day’). Summed item scores of 5–9 reflect minor depression, whereas a score of 10 or greater correlates significantly with major depression⁴⁷. In analyses, we used both the continuous PHQ-9 total score and a binary variable representing the presence of at least moderate depression (PHQ-9 ≥ 10). Cronbach’s alpha for the 9 PHQ-9 items was α = 0.920, indicating high internal consistency.

2. Self-reported generalized anxiety. The 7-item Generalized Anxiety Disorder (GAD-7) was utilized to identify generalized anxiety disorder (GAD) and assess the severity of symptoms associated with GAD [47]. The GAD-7 has demonstrated construct validity and reliability when used in the South African context [48]. Responses to each item ranged from 0 (‘not at all’) to 3 (‘nearly every day’), with total scores ranging from 0 to 21. Summed GAD-7 scores ≥ 10 have demonstrated specificity and sensitivity in identifying cases of GAD⁴⁹. This cutoff was used to create a binary variable representing the presence of at least moderate generalized anxiety. Continuous GAD-7 scores were also utilized in some analyses. Cronbach’s alpha for the GAD-7 survey was α = 0.886 in our sample.

3. Individual self-esteem. The Rosenberg Self-Esteem Scale (RSES) was used to evaluate individual self-esteem. The 10 items reflect measures of both positive and negative feelings about the self [49] All items are answered using a 4-point Likert scale format with scores ranging from strongly disagree (1 point) to strongly agree (4 points). Five of the items, reflecting measures of negative feelings about the self, are reverse scored and summed with the positive feelings items to create a total score. In our analyses, we used total RSES scores as a continuous outcome variable, with higher scores reflecting lower self-esteem. Cronbach’s alpha for the overall RSES questionnaire was α = 0.724.

4. Anticipated stigma. This study utilized the anticipated stigma subscale from the HIV Stigma Framework to quantify the degree to which anticipated stigma manifests in the lives of participants living with HIV [50]. The anticipated stigma scale consists of 9 items, with responses ranging from 1 (‘very unlikely’) to 5 (‘very likely’), with higher scores reflecting a greater degree of anticipated stigma. Each item corresponds to anticipated stigma occurring at one of three levels: the family, community, or healthcare setting. In this secondary analysis, total anticipated stigma scores and scores on each subscale (family, community, and healthcare setting) were considered as continuous variables. Cronbach’s alpha for the total anticipated stigma scale was α = 0.913. A dichotomous variable representing high and low total anticipated stigma was also used in our analysis of baseline characteristics, with scores > 27 (on average believing that anticipated stigma was just as likely as it was unlikely) reflecting high anticipated stigma and scores ≤ 27 reflecting low anticipated stigma.

Perceived social support was utilized as an outcome and mediator variable in this analysis. Perceived social support was assessed utilizing a 10-item perceived social support scale which quantified the degree to which respondents perceived two distinct domains of social support: family and community support. The items used to assess the family support domain of perceived social support were adapted from the Integrated Questionnaire for the Measurement of Social Capital (SC-IQ) [51] whereas items from the community support domain were adapted from the Multidimensional Scale of Perceived Social Support [52]. All items are answered using a 4-point Likert scale format with scores ranging from 1 (‘never’) to 4 (‘often’); one item, reflecting how often one experiences quarrels with family members, was reverse scored. The 10-items, inclusive of the reverse scored item, were summed into total scores ranging from 10 to 40, with higher scores reflecting a greater degree of social support. In our analyses, we primarily divided the complete scale into two community and family support subscales, reflecting total scores from 5 items each (each with a maximum score of 20). Cronbach’s alpha for the community and family subscales were α =0.879 and α = 0.782, respectively. In our analysis of demographic and socioeconomic characteristics, each type of social support (family and community) was binned into dichotomous variables, with scores ≥ 15 reflecting high degrees of social support (on average receiving social support at least ‘sometimes’) and scores ≤ 15 reflecting lower levels of social support.

Covariates

A directed acyclic graph (DAG) analysis was conducted to display assumptions about the relationship between our primary predictor (time of diagnosis) and our outcome variables. Our DAG analysis identified several potential covariates, including demographic characteristics measured in this study, such as gender (female, male), school level (completed high school, did not complete high school), age, and disclosure of HIV status at baseline (yes, no).

Statistical Analyses

First, we examined study participants’ psychosocial and demographic characteristics descriptively. Next, associations between study cohort (pre- vs. intra-COVID) and sociodemographic were assessed. Specifically, we used Chi-square analyses to examine the association between our categorical demographic and social variables (employment status, high/low family support, and community social support) and study cohort.

We then performed Chi-square analyses to examine the association of our dichotomous predictor variable (pre-COVID [reference] vs. intra-COVID) with dichotomous levels (high/low) of our psychosocial outcomes (GAD-7 scores, PHQ-9 scores, HIV Anticipated Stigma Scores). In these analyses, dichotomous outcomes variables were used understand clinically relevant differences in psychosocial outcomes. Subsequently, we performed simple linear regression analyses to examine the relationship of our continuous psychosocial outcome variables and our study cohort predictor variable. Continuous variables were used in these analyses to assess for any changes in psychosocial outcomes.

Finally, a mediation analysis was conducted to test the mediation effect of social support variables on the association between study cohort and select psychosocial outcomes. Our mediation analysis utilized the multi-stage approach elucidated by Baron and Kenny (see Fig. 1). After determining if study cohort (predictor variable) significantly predicted our psychosocial outcomes (outcome variable) as part of our previous regression analysis (path c), step 2 determined if study cohort significantly predicted our mediator of interest, either family or community social support (path a). In step 3, we determined whether these hypothesized mediators predicted our psychosocial outcomes (path b), and finally, in step 4 (path c’), we determined if study cohort remained a significant predictor of our psychosocial outcomes after controlling for our mediator of interest. Whereas full mediation occurs when the association between outcome and predictor is no longer statistically significant once the mediator is included in our analysis, partial mediation occurs if the strength of the association is reduced but still statistically significant [53]. The significance of all mediation effects was tested using the Sobel Test [54].

Fig. 1.

Diagram of proposed mediation model

All analyses were conducted using SPSS 29.0.1.1.

Compliance with ethical standards

This study was approved by the Partners Healthcare Institutional Review Board at Brigham and Women’s Hospital and Massachusetts General Hospital, Harvard Medical School, Boston, MA and the Human Research Ethics Committee at the University of Cape Town, South Africa. Informed consent was obtained from all participants engaging in both studies.

Results

In both of cohorts, all 200 participants identified as Black African. All identified either English or isiXhosa as their preferred interview language.

Table 1 compares the sociodemographic characteristics of young adults included in our two cohorts. As shown in Table 1, the mean age of participants in each study cohort was similar (21.17 years in our pre-COVID cohort versus 20.68 years in our intra-COVID cohort; t = 1.773, P > 0.05). Ages ranged from 18 to 24 in both cohorts. In addition, the gender composition of our cohort was similar (χ² = 0.207, p > 0.05), with 90% identifying as female in our pre-COVID cohort and 84% identifying as female in our intra-COVID cohort. In both studies, most of the participants reported being currently unemployed, and chi-square analyses produced no significant association between study cohort and education level, living status, and employment status. While there was no significant relationship between study cohort and reported levels of community-based social support (χ² = 38.69, p > 0.05), participants in the intra-COVID cohort reported significantly higher levels of family-based social support (χ² = 38.69, p < 0.001). Indeed, whereas only 39% of participants in the pre-COVID cohort reported high levels of family support, 82% of participants in the intra-COVID cohort reported high levels.

Table 1.

Sociodemographic characteristics of young adults newly diagnosed with HIV in two cohorts (pre- and intra-pandemic) (N = 200)

Characteristics	Pre-COVID cohort	Intra-COVID cohort	Test statistic, p-value
Total	100	100
Age group
18–19 years old	26 (26%)	33 (33%)
20–21 years old	23 (23%)	28 (28%)
22–23 years old	42 (42%)	30 (30%)
≥ 24 years old	9 (9%)	9 (9%)
Median age (years):	22	21	t = 1.773, P = 0.078
Mean age (years, SD):	21.17 +/− 1.949	20.68 +/− 1.959
Gender
Male	10 (10%)	16 (16%)	χ2 = 1.592, p = 0.207
Female	90 (90%)	84 (84%)
Education Level
Did not complete high school	44 (44%)	48 (48%)	χ2 = 0.322, p = 0.570
Completed high school	56 (56%)	52 (52%)
Living Status
Living alone	15 (15%)	7 (7%)	χ2 = 3.269, p = 0.071
Living with others	85 (85%)	93 (93%)
Employment Status
Unemployed	81 (81%)	87 (87%)	χ2 = 1.339, p = 0.246
Employed	19 (19%)	13 (13%)
Perceived Social Support
Family Support
Low	61 (61%)	18 (18%)	χ2 = 38.69, p < 0.001
High	39 (39%)	82 (82%)
Community Support
Low	63 (63%)	73 (73%)	χ2 = 2.298, p = 0.130
High	37 (37%)	27 (27%)

Results from our Chi-square analyses evaluating the association between dichotomous levels of each of our psychosocial outcomes and study cohort are indicated in Table 2. Being part of the intra-COVID cohort was associated with a significantly increased odds of having PHQ-9 (χ² = 5.531, p = 0.019) and GAD-7 scores (χ² = 7.779, p < 0.01) indicative of moderate-severe depression and generalized anxiety, respectively. AYAs living with HIV in the intra-COVID cohort also reported significantly higher self-esteem (χ² = 9.955, p < 0.01) and lower levels of anticipated stigma (χ² = 22.756, p < 0.001).

Table 2.

Psychosocial characteristics of young adults newly diagnosed with HIV in two cohorts (pre- and intra-COVID) (N = 200)

Psychosocial Characteristic	Pre-COVID cohort	Intra-COVID cohort	χ2-value, p-value
Total	100	100
PHQ-9 scores
Minimal-Mild (0–9)	93 (93%)	82 (82%)	χ2 = 5.531, p = 0.019
Moderate-Severe (≥ 10)	7 (7%)	18 (18%)
GAD-7 scores
Minimal-Mild (0–9)	97 (97%)	86 (86%)	χ2 = 7.779, P < 0.01
Moderate-Severe (≥ 10)	3 (3%)	14 (14%)
Rosenberg Self-Esteem Scores (RSES)
Low Self-Esteem (≥ 15)	99 (99%)	88(88%)	χ2 = 9.955, P < 0.01
High-Self Esteem (< 15)	1 (1%)	12 (12%)
Anticipated Stigma
Low anticipated stigma (Scores < 27)	69 (69%)	93 (93%)	χ2 = 22.756, P < 0.001
High anticipated stigma (scores ≥ 27)	27 (27%)a	3 (3%)a

^aMissing participants (n = 4 from each cohort) were excluded from the analysis

Results from the regression analyses evaluating the association between study cohort and each of our continuous psychosocial outcomes are presented in Table 3. Being in the intra-COVID time cohort was associated with significant lower RSES scores and therefore higher self-esteem (t=−10.95, p < 0.001). In addition, intra-pandemic time period was associated with lower total anticipated stigma scores (t=−8.25, p < 0.001), as well as lower scores on the family-type (t=−3.42, p < 0.001), community-type (t=−10.07, p < 0.001), and health worker-type anticipated stigma subscales (t=−9.94, p < 0.001).

Table 3.

Linear regression results for association between study cohort (pre- vs. intra-COVID) and psychosocial outcomes at baseline (N = 200)

Outcome Variable	β for study cohort (pre-COVID vs. intra)	95% CI	t-statistic	P-value
GAD-7 summed scores	0.325	(−0.852, 1.503)	t = 0.545	0.586
PHQ-9 summed scores	0.214	(−1.095, 1.522)	t = 0.322	0.748
RSES scores	−5.409	(−6.383, −4.435)	t=−10.95	P < 0.001
Anticipated stigma scores	−7.630	(−9.454, −5.807)	t=−8.253	P < 0.001
Anticipated stigma: family type	−1.510	(−2.382, −0.638)	t=−3.416	P < 0.001
Anticipated stigma: community type	−2.856	(−3.415, −2.297)	t=−10.07	P < 0.001
Anticipated stigma: health-worker type	−2.926	(−3.507, −2.346)	t=−9.94	P < 0.001

Results from the analyses examining the mediation effect of social support on the relationship between our predictor variable, study cohort and psychosocial outcomes, self-esteem (Table 4), total anticipated stigma and anticipated stigma subscales (Table 5), depression (Table 6), and anxiety (Table 7) are shown below. Partial mediation of the relationship between study cohort and self-esteem (RSES scores) was observed when family social support (Sobel z = −3.04, p = 0.0019), but not community social support, was included in the regression model. Only family social support demonstrated a statistically significant partial mediation effect of the relationships between study cohort and total anticipated stigma (Sobel z = 1.57, p = 0.005), family-type anticipated stigma (Sobel z = −4.06, p < 0.001), community-type anticipated stigma (Sobel z= −3.44, p < 0.001), and PHQ-9 scores (Sobel z=−2.80, p = 0.0052).

Table 4.

Regression analysis for mediation of social support between study cohort (pre- vs. intra-COVID) and levels of self-esteem (RSES scores)

Mediator	β for path a (SE)	β for path b (SE)	β for total effect c (SE)	β for direct c’ (SE)	Sobel z	p-value of indirect effect (ab)
Social support: family	1.860 (0.343)***	−0.364 (0.099)***	−5.409 (0.494)***	−4.732 (0.513)***	Z=−3.04	P = 0.002
Social support: community	0.440 (0.577)***	−0.005 (0.061)	−5.409 (0.494)***	−5.407 (0.407)***	Z=−0.08	P = 0.935

^*Significant at the p < 0.05 level

^**Significant at the p < 0.01 level

^***Significant at the p < 0.001 level

Table 5.

Regression analysis for mediation of social support between study cohort (pre- vs. intra-COVID) and total anticipated stigma and anticipated stigma subscores

Outcome Variable: Total Anticipated Stigma
Mediator	β for path a (SE)	β for path b (SE)	β for total effect c (SE)	β for direct effect c’ (SE)	Sobel z	p-value of indirect effect (ab)
Social support: family	1.860(0.343)***	−0.620 (0.186)***	−7.630 (0.925)***	−6.476 (0.966)***	Z=−2.84	P = 0.005
Social support: community	0.440(0.577)***	−0.057 (0.114)	−7.630 (0.925)***	−7.608 (0.927)***	Z=−0.42	P = 0.67
Outcome Variable: Anticipated Stigma, Family-Type
Social support: family	1.860(0.343)***	−0.491 (0.080)***	−1.510 (0.442)***	−0.681 (0.127)	Z=−4.06	P < 0.001
Social support: community	0.440(0.577)***	−0.046 (0.056)	−1.510 (0.442)***	−1.493 (0.444)***	z=−0.56	P = 0.576
Outcome Variable: Anticipated Stigma, Community-Type
Social support: family	1.860(0.343)***	−0.285 (0.064)***	−2.856 (0.284)***	−2.665 (0.301)***	Z=−3.44	P < 0.001
Social support: community	0.440(0.577)***	−0.056 (0.043)	−2.856 (0.284)***	−2.665 (0.301)***	Z=−0.66	P = 0.510
Outcome Variable: Anticipated Stigma, Health Worker-Type
Social support: family	1.860(0.343)***	−0.286 (0.67)***	−2.926 (0.294)***	−2.747 (0.314)***	Z=−0.43	P = 0.67
Social support: community	0.440(0.577)***	−0.057 (0.44)	−2.926 (0.294)***	2.910 (0.295)***	Z=−0.13	P = 0.898

^*Significant at the p < 0.05 level

^**Significant at the p < 0.01 level

^***Significant at the p < 0.001 level

Table 6.

Regression analysis for mediation of social support between study cohort (pre- vs. intra-COVID) and PHQ-9 scores

Mediator	β for path a (SE)	β for path b (SE)	β for total effect c (SE)	β for direct c’ (SE)	Sobel z	p-value of indirect effect (ab)
Social support: family	1.860(0.343)***	−0.408 (0.125)***	0.214 (0.664)	1.118 (0.690)	Z=−2.80	P = 0.005
Social support: community	0.440(0.577)***	−0.008 (0.082)	−0.214 (0.664)	0.218 (0.666)	Z=−0.10	P = 0.923

^*Significant at the p < 0.05 level

^**Significant at the p < 0.01 level

^***Significant at the p < 0.001 level

Table 7.

Regression analysis for mediation of social support between study cohort (pre- vs. intra-COVID) and GAD-7 scores

Mediator	β for path a (SE)	β for path b (SE)	β for direct effect c (SE)	β for indirect effect c’ (SE)	Sobel z	p-value of indirect effect (ab)
Social support: family	1.860(0.343)***	−0.183 (0.115)	0.325 (0.597)	0.765 (0.636)	Z=−1.53	P = 0.127
Social support: community	0.440(0.577)***	0.011 (0.074)	0.325 (0.597)	0.322 (0.600)	Z = 0.61	P = 0.884

^*Significant at the p < 0.05 level

^**Significant at the p < 0.01 level

^***Significant at the p < 0.001 level

Discussion

The primary goals of this analysis were to compare measures of psychosocial well-being between a pre- and intra-COVID cohort of AYA newly diagnosed with HIV and to determine whether perceived family and/or community social support were mediators of the relationship between study cohort and psychosocial well-being. Our simple linear regression analyses revealed that participants in the intra-COVID cohort had significantly higher self-esteem, lower levels of total anticipated stigma, and lower levels of family- and community- related anticipated stigma. At the same time, our results indicate that the pandemic may not have affected all aspects of psychosocial well-being equally. Indeed, even though there was no significant association between study cohort and levels of depressive and anxiety symptoms, newly diagnosed AYA in the intra-COVID cohort were more likely to report moderate-severe anxiety and depressive symptoms as compared to the pre-COVID cohort. While this is the first study to our knowledge showing an association between the pandemic and improved anticipated stigma and self-esteem, the finding that depressive and anxiety symptoms increased in the intra-pandemic setting is consistent with several other recent studies [55–57]. It is not exactly clear why self-esteem, but not anxiety and depression, was positively impacted by improved family support. However, given the relationship between higher levels of stigma and poor self-esteem, it is plausible that the effects of improved family support on outcomes such as stigma may have led to improvements in one’s individual self-esteem without impacting depressive and anxiety symptoms worsened by the pandemic itself.

Our analysis also demonstrated that perceived family social support was a significant mediator of the relationship between study cohort and several of our psychosocial outcomes, including levels of self-esteem, family-type, and community-type anticipated stigma, and depressive symptoms. On the other hand, community social support was not found to exhibit a significant mediation effect on any indicators of psychosocial well-being, and individuals in the intra-COVID cohort exhibited lower levels of perceived community social support compared to family social support. The significant mediation effect of family social support, but not community social support, is likely explained by the differential effect of each type of support on psychosocial functioning. Whereas higher family social support was associated with improved depressive and anxiety symptoms, self-esteem, and anticipated stigma, community support was not associated with any of these outcomes. Taken together, our results add to the literature suggesting that social support may exhibit a significant protective effect on certain aspects of psychosocial well-being [28, 32–33, 58–59], though our results suggest that at least in our sample of AYA living with HIV, it was family support, rather than that emanating from the community, that was particularly beneficial during the pandemic.

Our results further contribute to the extant literature surrounding the effects of the COVID-19 pandemic on psychosocial functioning in AYA with HIV by highlighting potential associational pathways by which different measures of psychosocial well-being may have been impacted by the pandemic. While at least one study has reported an increase in family-level interactions during the early phase of the pandemic [60], our analysis is the first to report a surprising improvement in certain aspects of psychosocial well-being mediated by an increase in family social support during the pandemic. There are several possible explanations for this unexpected finding. On one hand, COVID-era restrictions on social interactions with people outside of the household [25] may have led young South Africans, the majority of whom live with family members [61], to rely more on their relatives for social support than before the pandemic. Thus, like other young people globally [62–64], the young people recruited in our study may have simply had a greater quantity of contact with the relatives with whom they were living, leading to greater social integration into the family unit [65] and an improved sense of psychological well-being. Alternatively, family dynamics themselves may have changed during the pandemic, such that the family unit itself became more effective at providing quality social support than it had been previously. Indeed, there is some evidence to suggest that the pandemic may have provided families with the opportunity to strengthen relationships [66–67], possibly by promoting better communication dynamics, emotional expressiveness, and a collaborative spirit amongst family members [68]. Conceivably, these factors may have contributed to a greater sense of perceived emotional social support, a type of social support that has been consistently linked to improved mental health outcomes [65, 69].

While the exact mechanism by which family social support may have improved during the pandemic requires further exploration, the association between family support and higher self-esteem, fewer depressive symptoms, and lower anticipated stigma highlights potential mechanisms by which family social support may positively impact mental and physical health in AYA with HIV. Previous theoretical models have indeed highlighted the role of social support in promoting resilience to the negative health effects of stigma by improving emotional regulation, fostering self-protective attributions, and serving as an outlet for emotional expression [70–72]. However, our findings also suggest that family social support may contribute more directly to both improved self-esteem and reduced expectations of HIV-related social rejection both within and outside the family unit. Moreover, given that the positive effects of family social support both manifested amidst a highly disruptive global pandemic were associated with indicators of social well-being outside of the family (e.g. community-level stigma), our findings also lend support to the body of literature highlighting family social support as particularly promising target for intervention [73–76] As public health practitioners have increasingly realized, so-called HIV competent households that support HIV discussion and disclosure may foster a “health-enabling environment” that is conducive to both psychosocial well-being and adherence to ART [33–34, 77].

Our findings, which only partially confirmed our hypotheses, further elucidate the diverse effects of the pandemic on the psychosocial well-being of people living with HIV (PLWH). Though at least one recent scoping review has indicated high rates of mental illness during the global COVID-19 pandemic [78], the reported psychosocial effects of the pandemic on PLWH seem to vary significantly across different populations [19, 79, 80]. In fact, even within the South African context, studies have alternatively indicated both a worsening of mental health [81–82] and improvement in individual well-being amongst PLWH [83]. Situated within the context of these studies, our findings underscore the need for both further research investigating the complex pathways affecting the psychosocial well-being of AYA during the pandemic as well as differentiated interventions targeted to the unique needs of different subpopulations of PLWH.

Strengths, Limitations, and Future Directions

This study has several strengths. For one, the timing of our cross-sectional assessments of psychosocial well-being in two demographically similar cohorts of newly diagnosed AYA living with HIV provided the rare opportunity to compare psychosocial well-being across this important moment in time. In addition, our focus on the psychosocial well-being of AYA newly diagnosed with HIV allowed us to evaluate the unique social and psychological effects of the pandemic on a relatively vulnerable population that faces a disproportionate burden of South Africa’s HIV epidemic. In turn, this focus on youth permitted us to explore and identify trends in well-being that may have been obfuscated by larger analyses of pandemic-era changes in psychosocial well-being across the general population.

On the other hand, this study also has several important limitations. For one, though family social support was a significant partial mediator of the relationship between study cohort and several of our indicators of psychosocial well-being, these partial mediation effects were relatively small. Along similar lines, our mediation analyses were conducted using data from two cross-sectional surveys, and therefore, the directionality of our proposed mediator to outcome pathway cannot be directly determined. For example, it is possible that those with more social support were better poised to overcome COVID-19 related barriers to accessing our recruitment sites, such that greater social support increased the likelihood of participating in the intra-pandemic cohort rather than the other way around. Moreover, there are several other unmeasured variables, such as disclosure experiences [84] interactions within the healthcare setting [85] and changes in community norms [86] that could have conceivably mediated the relationships between study cohort and psychosocial outcomes in this study. Additionally, given the correlational nature of our mediation analyses, we are unable to conclude that it was the COVID-19 pandemic itself that directly led to changes in each of the psychosocial variables assessed. It is thus possible that other time-dependent factors, such as the South African government’s shift to universal testing and treatment in 2017 [87], may have had a similar or greater effect on psychosocial outcomes such as stigma, depression, and anxiety, than that of the COVID-19 pandemic alone. Our small sample size also suggests that our results must be interpreted with caution. In particular, our difficulty recruiting men at the mobile clinic and clinic-based recruitment sites means that our results may not be generalizable to young South African men newly diagnosed with HIV. While our difficulty recruiting men may reflect a broader lack of male engagement with the HIV treatment cascade [88–90], our disproportionate sampling of females makes it challenging to determine if the observed associations also apply to young males. In addition, in large part because of this small sample size, our study considered time to be a dichotomous predictor variable, neglecting the fact that the experience of the COVID-19 pandemic certainly changed as South Africa adjusted lockdown restrictions during the National State of Disaster [91]. Along the same lines, the results of our study may not be generalizable to AYA living with HIV in all contexts, even within South Africa. Indeed, both the emerging literature on the psychosocial impacts of the pandemic and our study’s finding of differences in psychosocial outcomes between the two intra-COVID recruitment sites is suggestive of significant heterogeneity in the impacts of the pandemic and resultant lockdowns. Finally, our study did not evaluate the associational pathways between study cohort, psychosocial outcomes, and HIV clinical outcomes.

While our study adds to the literature regarding the effect of the pandemic on AYA newly diagnosed with HIV, larger samples may be needed to disentangle the complex effects of the COVID-19 pandemic on psychosocial outcomes in this vulnerable population. Further research is also necessary to evaluate the likely multifactorial ways in which clinical measures of well-being, such as ART adherence and viral load, were impacted by changing social and psychosocial well-being during pandemic. To evaluate the complex pathways between social mediators and psychosocial outcomes future settings, longitudinal studies measuring proposed mediators prior to outcomes, are likely to be necessary to ascertain the directionality of mediation effects. Finally, translational research evaluating the impact of family-support based interventions on psychosocial and physical well-being in AYA living with HIV, may be helpful in elucidating the degree to which psychosocial and physical health is responsive to improved family support outside of the pandemic setting.

Conclusions

In summary, the results of this analysis indicated that in our sample of South African AYA living with HIV, youth who were diagnosed with HIV during the pandemic had lower anticipated stigma and higher self-esteem. The relationship between study cohort and these psychosocial outcomes was partially mediated by family social support. Our findings expand our understanding of the complex effects of the COVID-19 pandemic on the psychosocial functioning of PLWH and lend support to interventions seeking to improve family social support as a means of improving the well-being of PLWH.