Study population and procedures
Data for analyses were obtained from a mental health study of YPLHIV and YPAHIV. Participants were recruited from the antiretroviral therapy (ART) center affiliated to Byramjee Jeejeebhoy Government Medical College and Sassoon General Hospitals (BJGMC & SGH), a publicly funded tertiary health care center in Pune, a city located in the state of Maharashtra, western India. In 2019, the state had the highest number of PLHIV (n = 396,000) in India . The ART center caters to approximately 350 YPLHIV. Recruitments were done between August 2018 and June 2021.
Two study counsellors approached all YPLHIV between 15 and 25 years of age attending the ART center for HIV care. To enroll YPAHIV, PLHIV attending the ART center with children between 15–25 years of age were approached. To prevent accidental disclosure, only YPLHIV and YPAHIV aware of their own or their parent’s HIV status, respectively, were enrolled. Informed assent and parental/guardian consent were required for participants < 18 years of age or informed consent for particiapants ≥ 18 years. Institutionalized young people were excluded from the study.
All study scales were self-administered on handheld devices. Marathi (the locally spoken language) was used in all study proceedings, including study scales. Participants were first required to undergo a reading/comprehension test, using a paragraph from an eighth grade Marathi textbook used in publicly funded schools (In India, the average age to reach eighth grade literacy is between 12–14 years). Participants were excluded (YPLHIV: n = 6, YPAHIV: n = 2) if they were unable to read/comprehend the paragraph. After successfully completing the test, participants were provided handheld devices to complete the study scales. Study counselors were present in the study room if participants required scale items, but study responses were hidden from them.
The Ethics Committee of Byramjee Jeejeebhoy Government Medical College and the Johns Hopkins Institutional Review Board approved this study.
The HSS was used to assess HIV-related stigma among YPLHIV and YPAHIV. The HSS is scored on a 4-point Likert scale (1-strongly agree to 4 - strongly disagree). Total scores range between 40–160, with higher scores indicating greater HIV-related stigma. The scale is further divided into four subscales, with each subscale having a different number of items. Items can load on to more than one subscale. The subscales measure: personalized stigma (18 questions, score range: 18–72); disclosure concerns (10 questions, score range: 10–40); concern about public attitudes (13 questions, score range: 13–52); and negative self-image (20 questions, score range: 20–80).
The Center for Epidemiological Studies Depression (CES-D) scale was used to assess depressive symptoms in those ≥ 18 years of age, and its modification, the Center for Epidemiological Studies Depression Scale for Children (CES-DC) for those < 18 years [29, 30]. Both scales are worded similarly and ask participants to rate how often they experienced depressive symptoms over the past week, using 20 items scored on a 4-point Likert scale (0 – rarely or none of the time to 3 – most or all the time). Total scores range between 0–60, with higher scores indicating greater depressive symptoms. The scales had good internal consistency for YPLHIV (Cronbach’s α: 0.86 for those < 18 years and 0.88 for those ≥ 18 years) and YPAHIV (Cronbach’s α: 0.91 for those < 18 years and 0.80 for those ≥ 18 years).
The UCLA Loneliness Scale version 3 was used to measure participants’ subjective feelings of loneliness and social isolation, using 20 questions scored on a 4-point Likert scale (1 – Never to 4 – Often). Total scores range between 20–80, with higher scores indicating greater perceived loneliness and social isolation . The scale had good internal consistency for YPLHIV and YPAHIV (Cronbach’s α: 0.85 and 0.79, respectively).
Scale Adaptation And Modification For Ypahiv
The HSS was first translated from English to Marathi. The translated scale items were verified for consistency, cultural relevancy, and comprehensibility by a review committee. The review committee included three study counsellors (graduates in social work, each with ≥ 5 years of conducting quantitative or qualitative research), and two study investigators (SN & IM) trained in mental health, instrument development and psychometrics. The scale approved by the review committee was then back translated into English to assess for original item equivalence, by two individuals unrelated to the study and not familiar with the HSS, proficient in both Marathi and English. Following this, the translated scale was re-tested for participant comprehensibility, demographic and cultural relevancy using cognitive interviews with 33 YPLHIV.
Scale items were modified according to cognitive interview findings. For example, the question, “People I care about stopped calling after learning I have HIV”, was modified as, “People close to me have stopped calling me on the telephone, coming to my house, after learning I have HIV.”
For YPAHIV, the same procedures as described for YPLHIV were followed. However, HSS items were reworded to reflect the HIV status of the participant’s parents. For example, the question, “Telling people I have HIV is risky” was modified as, “Telling people my parents are/were living with HIV is risky”. Cognitive interviews were conducted with 20 YPAHIV and modifications to scale items were made accordingly. For example, the question, “I feel guilty because my parents have/had HIV” was reworded as “I feel ashamed because my parents have/had HIV”.
Participants that took part in cognitive interviews were invited to be part of the study only after three months had passed. Similar procedures were followed to adapt the CES-D, CES-DC, and UCLA Loneliness Scales into Marathi.
We identified eight published shortened adaptations of the HSS. Two of these adapted scales were in English [20, 32], three in Swedish [33–35], one in Spanish , one in Thai , and one in Tamil (a linguistically unrelated Indian language) . The number of questions in these adapted scales range between 10–39.
Internal consistency was assessed using Cronbach’s alpha for the overall scale and the subscales. We tested the four-factor structure (modeled after the HSS subscales) for the eight adapted scales and the HSS among YPLHIV using confirmatory factor analysis (CFA). These are described as primary models. As responses in the HSS and adapted scales are ordinal, a weighted least squares estimator with a diagonal weight matrix and robust standard errors, and a mean- and variance-adjusted chi-square (χ2) statistic was used. The four factors i.e., personalized stigma, disclosure concerns, negative self-image and public attitudes concern were modelled as latent variables. Correlation between latent variables was allowed, but we did not allow inter-error correlation. Factor loadings between items and latent variables were standardized. Model fit was evaluated using χ2 test, Root Mean Square Error of Approximation (RMSEA), Tucker-Lewis Index (TLI), Comparative Fit Index (CFI) and the Standardized Root Mean Square Residual (SRMR). Good model fit was indicated by a χ2 associated p-value > 0.05, RMSEA < 0.08, TLI and CFI ≥ 0.90 and SRMR < 0.08 .
Secondary models were constructed by a) replacing a subscale(s) with low Cronbach’s alpha in models with good fit indices, with a subscale(s) from the eight primary adapted scales with a higher Cronbach’s alpha; b) combining different subscales with the highest Cronbach’s alpha values from the eight adapted scales. CFA was performed on all secondary models.
The final identified model was chosen based on four criteria, a) good model fit indices; b) consistency of items with the original four factor structure loadings i.e., factors that loaded on to the latent variables in the abridged model were a subset of the factors that loaded on the corresponding latent variables in the HSS; c) absolute magnitude of factor loadings > 0.40 ; d) demographic and cultural relevance, as judged by findings from cognitive interviews. The final identified model was then tested on the dataset of YPAHIV using CFA.
We used Bonferroni-corrected Spearman’s correlation coefficients to assess for correlations between the HSS and the scale identified. Internal construct validity was evaluated using correlations between subscales, and convergent validity using correlations between the HSS, the identified scale and with CES-D(C) and UCLA Loneliness scales. We hypothesized that the subscales would be positively correlated with each other, and, with depressive and loneliness scores, as reported in previous studies [25, 35, 36]. All evaluations were conducted separately for YPLHIV and YPAHIV.
All analyses were performed in R version 4.1.2 and Stata 17.0.