Sankofa pediatric HIV disclosure intervention did not worsen depression scores in children living with HIV and their caregivers in Ghana

The recruitment and organization of the parent study, ‘Sankofa,’ has been previously described [14]. Study enrollment for the ‘Sankofa’ study lasted from January 2013 till June 2016, and the study ended in 2017. It had a control arm at Korle Bu Teaching Hospital in Accra, Ghana, and an intervention arm at Komfo Anokye Teaching Hospital (KATH) in Kumasi, Ghana. All caregiver-child dyads who received the intervention at KATH were eligible for inclusion in this cross-sectional study which enrolled participants from June to August 2019. Dyads were recruited at KATH’s pediatric HIV clinic during routine clinic appointments. Written informed consent was obtained from the caregivers of the children living with HIV, and consent was obtained from caregivers on behalf of children less than 18 years of age with an assent from the child participant. The study protocol was reviewed and approved by the Institutional Review Boards of Kwame Nkrumah University of Science Technology/KATH (Protocol #: CHRPE/AP/146/12) and Yale University (Protocol #: 1205010310).

The 10-item Child Depression Inventory (CDI) and the 21-item Beck Depression Inventory (BDI) were administered to caregiver-child dyads in the parent Sankofa study every 48 weeks till the end of the study [25]. Other instruments such as Brief Illness Perceptions Questionnaire (Brief IPQ), Social Provisions Scale, brief HIV Knowledge questionnaire (HIV-KQ-18), and HIV Stigma Scale were administered to caregivers during the same timepoints in the parent study [26‐29]. In the current study, we administered only the CDI and BDI to the children and caregivers, respectively. To avoid accidental disclosure, questionnaires were administered to children in the presence of their caregivers whereas caregivers were interviewed alone. The interviews took place in a clinic setting, and the majority were completed in English. For participants who only spoke Twi, the major regional language, native-speaking research staff translated accordingly. The original research staff from the Sankofa study administered the questionnaires with the help of authors AS, QM, and CR.

Demographic data for the caregiver-child dyads were collected upon enrollment in the Sankofa study. The 10-item CDI was used to screen for depression symptoms in children. Each of the 10 items is scored on a four-point scale (0–3) to yield a maximum score of 30. This abridged form of the CDI has yet to be validated in our population; however, a version of the 10-item CDI has been studied in African adolescents without HIV [30]. Cutoff scores are not available, and a higher CDI score indicates greater severity of depression symptoms. The 21-item BDI was used to assess depression symptoms in caregivers. The BDI has been previously validated in sub-Saharan Africa [31], and it rates each of its 21 items on a four-point scale (0–3). A total score is assigned, and its results are divided into four categories depending on the severity of depression symptoms: minimal (0–9); mild (10–18); moderate (19–29); and severe (30–63). For our population, the BDI has been reported to have a three-factor structure (affective, cognitive, and somatic) and high internal reliability (0.90) [31].

Baseline demographic data were summarized with descriptive statistics. Means and standard deviations were determined for continuous variables whereas counts and percentages were determined for categorical variables. Differences in demographic data between disclosed and non-disclosed groups were assessed using t-test or Wilcoxon rank sum test for continuous variables and chi-square or Fisher’s exact tests for categorical variables, as appropriate. The same methods were used to assess differences between dyads who received the intervention and were enrolled in the current study versus those who were not enrolled. The mean changes of CDI and BDI scores between baseline and the current study were examined using paired t-tests. The correlation of CDI and BDI scores was examined using Pearson’s correlation coefficients. The changes in BDI categories, defined by standard cutoff scores, were also evaluated using McNemar’s test. Similar analysis for CDI data was not included due to lack of validated cutoff scores for categorizing symptom severity. Effect of disclosure status on CDI and BDI scores over major assessment time points was evaluated separately using repeated measures analysis with disclosure status entered as time varying covariate [32]. Baseline CDI/BDI score, time, and interaction between time and disclosure status were also included as fixed effects for both models. Two more covariates, baseline HIV Stigma Scale score and Brief IPQ score, were included in the model for BDI score. P value less than 0.05 was considered statistically significant. All analyses were conducted using SAS/STAT® software, Version 9.4 of the SAS System for Windows (Cary, NC, USA).

Of the 240 caregiver-child dyads who received the Sankofa intervention at KATH during the parent study, 65% (N = 157) of them were successfully enrolled in the current study. Of the remaining 83 caregiver-child dyads, 51 dyads were nonresponses—they did not attend clinic during the study period and/or did not respond to phone calls to invite participation. The remaining dyads (N = 32) were excluded for the following reasons: child passed (N = 18); caregiver passed (N = 11); both child and caregiver passed (N = 3). Table 1 summarizes the children’s demographics; 50.3% were female, and the median age at enrollment was 10 years (IQR 8–12). The majority attended school (93%), and mother to child transmission was the most common mode of HIV transmission (79.6%). The disclosed group was significantly older (p < 0.001), had been on ART for a longer period of time (p = 0.014), and had a longer mean time between enrollment and the current study (p < 0.001). No significant differences were noted between children who were or were not enrolled in this study.

The demographics of the caregivers are presented in Table 2. Caregivers’ mean age was 42.22 ± 9.69 years, and 82.8% were female. Fifty-six percent of caregivers were married, and most were living with HIV (61.2%). Caregivers in the disclosed group were significantly older (p = 0.04) and differed in their employment statuses (p = 0.020). Compared to caregivers who were enrolled in the current study, those who were not had significant differences in marital status (p = 0.016), monthly household income (p = 0.036), and baseline HIV-KQ-18 score (p = 0.018). Compared to the caregivers not enrolled in the current study, the caregivers enrolled lived significantly closer to KATH (p = 0.046).

In the current study, participants’ depression scores were collected a median of 5.19 years (IQR 4.16–6.07 years) after participants’ respective dates of enrollment in the parent study. The CDI scores of the children were 5.19 ± 3.77 (mean ± SD) and 3.35 ± 3.50 (mean ± SD) at ‘Sankofa’ baseline and the current study, respectively. There was a significant mean reduction of 1.82 (95% CI: 1.01–2.63) between baseline and the current study’s measurements (p < 0.0001). Interestingly, children who were disclosed to had a greater reduction in CDI scores compared to those who were not disclosed to, although, this did not reach statistical significance. After adjusting for time since enrollment, we still found no significant difference in CDI reduction between disclosed and non-disclosed children (p = 0.19). CDI scores of the children in the current study were significantly correlated with BDI scores of the caregivers in the current study (r = 0.19, p = 0.019).

BDI scores at ‘Sankofa’ baseline and the current study for caregivers were 6.55 ± 6.02 and 3.94 ± 4.49, respectively. We observed a statistically significant reduction in BDI scores between baseline and the current study, (2.65 (95% CI: 1.76–3.55), p < 0.0001). However, we did not observe any significant difference in mean reduction between caregivers of disclosed children compared to caregivers of non-disclosed children. When adjusting for time since enrollment for caregivers, we still did not find a significant difference in BDI reduction between caregivers of disclosed and non-disclosed children (p = 0.41). BDI scores from baseline and the current study were found to be highly correlated (r = 0.45, p < 0.0001). When BDI scores were analyzed as a categorical variable, we observed improvement of scores between baseline and the current study: 6 (86%) caregivers with mild depression changed to minimal depression; 5 (100%) caregivers with moderate depression changed to mild or minimal depression, and 2 (100%) caregivers with severe depression changed to moderate or minimal depression over time. Only 2 (1.5%) caregivers changed to mild or moderate depression from minimal depression. These changes did not reach statistical significance (Table 3). However, when the categories were combined into minimal versus mild, moderate, and severe depression, the changes reached statistical significance (p = 0.02). We repeated this analysis by disclosure status and found borderline significant improvement in the caregivers of disclosed children (p = 0.06) but not in the caregivers of undisclosed children (p = 0.16).

Overall, we did not find a statistically significant association between disclosure status and CDI score of children (p = 0.48). The adjusted mean CDI score of children who were disclosed to was lower than that of children who were not disclosed to across all major assessment time points except week 144. Only at week 48 was the mean difference statistically significant (Fig. 1, Table 4).

A separate mixed model was created for BDI scores of caregivers with baseline HIV Stigma Scale score and Brief IPQ score added as covariates. We did not find a statistically significant association between disclosure status of children and BDI score (p = 0.11). Compared to caregivers of non-disclosed children, the adjusted mean BDI score of caregivers of disclosed children was lower at all time points during the parent study and the current study; however, this was not statistically significant except at week 48 (Fig. 2, Table 4).