Which HIV-infected youth are at risk of developing depression and what treatments help? A systematic review focusing on Southern Africa

Study characteristics and quality

Twelve studies (nine independent samples) were included, comprising HIV-infected youth aged 9–19 years. Of these, three papers (two samples) comprised both eligible and non-eligible participants, with data relevant to this review provided by authors upon request [28], [29], [30]. Results are therefore based on 3573 HIV-infected youth from nine samples.

Sample sizes ranged from n = 31 (eligible participants from Målqvist et al.’s [30] larger dataset) to n = 1060 [31], [32]. Most studies recruited from specialist HIV clinics using convenience sampling [28], [29], [33], [34], [35], [36], [37]. While this provides a feasible recruitment method for researchers, and is an effective way of verifying the HIV status of participants, the risk of bias is high [38]; recruiting from HIV clinics is likely to exclude youth who have not undergone testing, choose not to use HIV services, or are unable to do so. Målqvist et al. [30] used the more robust method of consecutive sampling, but was community-based and relied on self-report of HIV status. The small number of participants eligible for inclusion in this review (n = 31) reduces the power of the analysis and generalisability of the results. Two samples were recruited through community tracing in addition to directly through HIV clinics, enabling inclusion of participants not attending specialist services [31], [32], [39]; thus, providing a less-biased and more representative sample of HIV-infected youth.

Most studies recruited from South Africa [28], [29], [31], [32], [33], [37], [39], [40]; with some representation from Malawi [34], [35], Zambia [36] and Swaziland [30].

Only one sample was assessed for depression via clinical interview – a method deemed more rigorous than screening tools [34]. However, it is worth noting that despite excellent psychometrics internationally, the lack of child mental health research and public sector resources in Southern Africa means the Child Depression Rating Scale-Revised (CDRS-R) has been culturally adapted but not validated in these countries [41]. This sample was also assessed using the Beck Depression Inventory-II (BDI-II) screening tool. The remaining studies used screening tools only: one sample used the BDI [28], [29], one used the Edinburgh Postnatal Depression Scale (EPDS) [30], one used the Center for Epidemiological Studies Depression scale (CESD) [36] and six samples used the Child Depression Inventory Short Form (CDI-S) [31], [32], [33], [37], [39], [40]. The BDI-II, EPDS and CESD have all been validated for use in Southern Africa, with psychometrics reported [41], [42], [43]. The CDI-S has been used in Southern Africa, but not adequately validated; whilst Snider and Dawes [44] addressed the wording of items, no psychometrics were provided for Southern African populations. The BDI, which measures different constructs to the BDI-II has not been validated in Southern Africa. Some studies used overall or mean scores on depression measures to create continuous variables [32], [33], [34], [35], [37], [40] while others used cut-offs [28], [29], [30], [36]. In addition to using overall score, Woollett et al. [37] conducted a separate analysis using presence of >50% of symptoms (defined by the authors as scoring 1 or 2, on five or more items), to indicate ‘increased risk of depression’. One study did not describe how depression was used as a variable in their analysis [39].

Quality of included studies varied, with quality assessment scores ranging from 3 to 12 (possible range 0–16). Kim et al. [35] and Sherr et al. [40] were rated highest. Sherr et al. [40] recruited a large sample using robust methods, with clearly described and reported analyses. However, Kim et al. [35] used the methodologically superior clinical interview to determine depression. See Table 2 for quality assessment and Table 3 for summary of findings.

Gender

Seven studies assessed the association between gender and depression. Four studies found no association [31], [32], [36], [40]. The remaining three found depression to be associated with female gender using scores on screening tools [28], [34], [37]. No significant association was found when using clinical interview [34] or ‘increased risk of depression’ [37], although this approached significance (p = 0.06).

Age

Six studies assessed the association between age and depression; three found depression to be associated with older age [31], [34], [40]. Kim et al. [34] found this association when using clinical interview to determine depression. When using the BDI-II, there was no direct effect of age; however, there were significant interactions – older participants who were more dissatisfied with their appearance, or who had a lower height-for-age z-scores, had higher depressive symptoms. The remaining three studies found no association between age and depression [28], [36], [37]; studies that found an effect of age recruited younger participants than the studies that found no effect.

Education

Three studies measured schooling variables. Kim et al. [34] found fewer years of schooling was associated with depression using both clinical interview and the BDI-II score. Failing a school term/class was also associated with a higher BDI-II score. However, the other two studies found no relationship between depression and education [36], [37].

Caregiver demographics

Two studies considered depression in relation to caregiver variables. Lower education of the primary caregiver was associated with higher depressive symptoms [33]. Non-significant variables were primary caregiver type (i.e. single parent; both parents; aunt/uncle; grandparent/other); maternal employment status; and changes in caregiver [34]; caregiver HIV status; caregiver employment, caregiver depression (measured using screening tools) [33].

Other household variables

Other household variables were assessed by four studies. Increased depression was associated with higher household density, although this did not remain significant when controlling for age, gender and study indicator [33]. Surprisingly, Woollett et al. [37] found that youth looking after a sick person in the home was associated with lower CDI-S score. It is possible that this role provides youths with a sense of purpose, acting as a protective factor. However, the paper found no association between looking after younger children in the family and depression. As a standalone finding in a single paper, this warrants further investigation.

Analysis of Målqvist et al.’s [30] dataset demonstrated significantly lower depression in those renting, rather than being a homeowner/living with parents; whilst this replicated findings from the overall dataset, the eligible sample was very small (n = 22), with only one participant owning their home. No other variables from this dataset were significant. Other variables not found to be associated with depression include whether the caregiver is living in their own home [33]; location of home; travel time between home and clinic [34]; household composition; formal or informal housing; looking after younger siblings at home [37] and rural vs. urban location [31].

Household income

Five studies measured household income, with no association found with depression [31], [33], [34], [37], [40].

Food security

Three studies looked at measures of food security. Two found an association between less food security and higher depression scores [33], [37], although this did not remain significant in Bhana et al.’s [33] adjusted analysis. This may be due to differing levels of food insecurity between the two included samples [pilot and main randomised controlled trials (RCT)]. Analysis of eligible participants from Målqvist et al.’s [30] dataset (n = 28) found that, in contrast to the overall dataset, there was no association between depression and their three measures of food security.

Disclosure

Three studies looked at disclosure in relation to depression. Kim et al. [34] found that participants who were aware of their HIV status and had disclosed it to others had lower depression than those who had not disclosed it to others, or who were not aware of their status. While Woollett et al. [37] found that participants who knew their HIV status had significantly lower depression scores than those who did not know, they found no difference between those who had disclosed their status to others (27% of the sample) and those who had not. Boyes et al. [31] found no difference in depression between those who knew about their HIV status before or after 12 years of age.

ART

Two studies assessed ART adherence, with limited evidence of a relationship with depression. Although Kim et al.’s [35] univariate analyses found poorer ART adherence was associated with depression using the BDI-II and clinical interview, this did not remain significant in multivariate analyses. Okawa et al. [36] found no relationship between depression and ART non-adherence.

Two studies examined the role of other ART-related variables. The presence of ART side-effects was associated with higher depression [31], while ART use (93.6% of the sample); an efavirenz-based regimen; and second-line ART [34] were not.

Health

Three studies assessed associations with other health variables. Boyes et al. [31] found worse overall health and negative clinic interactions predicted higher depression. More severe immunosuppression classification (based on CD4) predicted depression using the BDI-II, but not clinical interviews [34]. No other bio-clinical parameters assessed by Kim et al. [34] were associated with depression, including: history of tuberculosis; most recent CD4 count; nutritional status; alcohol use and hospital admission in the past year. Woollett et al. [37] found no association between depression and history of tuberculosis or being hospitalised.

Mode of transmission

Two studies assessed the relationship between depression and mode of HIV infection. Sherr et al. [40] found higher depression scores in those who had acquired HIV behaviourally compared with perinatally, while Boyes et al. [31] found no differences. However, it is worth noting that Boyes et al. [31] used an age cut-off and algorithm in the absence of clinic notes regarding the mode of infection.

Other HIV factors

Pantelic et al. [32] found HIV-related disability to be associated with depression via higher abuse victimisation and more enacted HIV stigma.

Relationships

Four studies assessed relationships between youth and their family/caregivers. Okawa et al. [36] found unsatisfactory relationships with family was related to depression. Bhana et al. [33] found lower youth ratings of caregiver supervision was associated with higher depression, as was lower likelihood of youths seeking social support. In contrast, Boyes et al. [31] found no association between depression and youth-rated parental monitoring, nor parent communication. They did, however, find that youth-rated positive parenting was associated with lower depression, as was better self-rated social support. Surprisingly, Woollett et al. [37] found higher depression scores in youth who reported receiving praise at home; getting the same things as other children in the home; and feeling like they belonged in the family they were being raised in.

Two studies examined other relationships. Unsatisfactory relationships with friends and healthcare workers were both associated with depression [36]. Those with a current or past boyfriend/girlfriend had higher BDI-II scores than those who had never had a boyfriend/girlfriend [34].

Community

Two studies looked at youth engagement with the wider community. Having a safe place in the community and accessing a clinic support group were associated with lower depression [31], [37].

Violence and abuse

Six studies assessed various forms of violence and abuse in relation to depression. Although there was heterogeneity in how sexual abuse history was measured, four studies found an association with depression [28], [31], [32], [37]. Four studies found an association between depression and experiences of violence [28], [29], [32], [37]. Measures of violence also varied, including ‘being hit’, interpersonal violence among females, collective violence and peer violence. Having witnessed somebody being stabbed or shot was not associated with depression [37].

Feeling unsafe at home, emotional abuse and cumulative adverse/abuse experiences were also associated with depression in this review [28], [31], [32], [37]. In contrast to the other studies, Kim et al. [34] found no association between depression and experience of forced sex, physical abuse or witnessing physical abuse in the home.

Bereavement

Four studies assessed parental survival status; all found no association with depression [31], [36], [37], [40].

Two studies assessed other bereavement variables, with mixed findings. Kim et al. [34] found a death in the family/household was associated with a higher BDI-II score, whilst Woollett et al. [37] found no association between bereavement and depression.

Bullying

Two studies examined bullying. Boyes et al. [31] found bullying victimisation was associated with higher depressive symptoms. Kim et al. [34] found being bullied for taking medication was associated with depression when measured by both the BDI-II and CDRS-R, but being bullied for appearance was not.

Stigma

Five studies explored the relationship between depression and stigma; all five found an association [31], [32], [33], [36], [39]. Three of these measured enacted, anticipated and internalised stigma separately using the Adolescents Living with HIV Stigma Scale [31], [32], [39]. Only Pantelic et al. [39] found an association between depression and anticipated stigma; two found an association with enacted stigma [32], [39]; and all three found an association with internalised stigma.

Other psychosocial factors

Three studies explored various other psychosocial factors. As to be expected from the overlap between psychosocial factors and characteristic symptoms of depression, many of the psychosocial factors examined were found to be associated with depression. Woollett et al. [37] found that suicidality; feeling unable to control one’s future and not having dreams for one’s future were all predictive of the CDI-S score and ‘increased risk of depression’. A higher depression score was also associated with a higher score on the Revised Child Manifestation of Anxiety Scale. Depression was not associated with score on the Child Post Traumatic Stress Disorder (PTSD) Checklist. Boyes et al. [31] found that lower self-efficacy was related to higher depression. Bhana et al. [33] found lower self-concept (measured using a scale validated for use in Southern Africa) was associated with higher depression. Two of the 10 coping styles measured by the Kidcope were associated with higher depression score: social withdrawal and resignation, although this was only assessed in the pilot-study participants (n = 66).

Study characteristics and quality

Only one identified study examined the psychosocial interventions for depression in HIV-infected youth [45]. Participants (10–13 years) were clinically-verified as being HIV-infected and were randomly allocated to a family-based intervention (n = 33) or wait list control group (n = 32). The intervention used a cartoon-based storyline about a boy with HIV to deliver information and promote discussion within and between families over six sessions. The curriculum covered a range of topics relevant to HIV in youth (listed in Table 5). The intervention was delivered by lay counsellors, supervised by a psychologist. Depression was measured pre- and post-intervention using the CDI-S. The study quality is described in Table 4.

Effect on depression

Bhana et al. [45] used generalised linear models to compare intervention and control groups over time, accounting for the effect of repeated measures. The intervention did not have a significant effect on depression.