The People Living with HIV Stigma Index is the world’s largest social research project developed by and for people living with HIV to measure nuanced changes in stigma and discrimination (37). This global survey tool has been implemented in more than 100 countries and relies heavily on the Greater Involvement of People Living with HIV and AIDS (GIPA) principle (38). We utilize data from the Ontario implementation of the HIV Stigma Index which included additional externally validated scales to measure stigma and other health risks. The present analysis primarily uses data from these externally validated scales.
Trained peer research associates (PRAs) living with HIV were hired to recruit 724 people living with HIV and administer the Ontario HIV Stigma Index. Survey participants are a cross-section of people living with HIV from several regions across Ontario and include priority populations such as gay, bisexual, and other men who have sex with men (GBMSM), African/Caribbean/Black (ACB) individuals, women, youth, Aboriginal peoples, injection drug users, and individuals from rural communities. PRAs administered the HIV Stigma Index in their respective regions through face-to-face interviews lasting approximately two hours between September 2018 and August 2019. All adult participants were considered for enrollment if they were (1) HIV-positive, (2) able to adequately communicate in English or French for the duration of the interview process, and (3) willing and able to complete the interview process and provide informed consent. The study was approved by the Research Ethics Board of St. Michael’s Hospital.
Various demographic data about each participant were collected at the beginning of the survey including age, years since HIV diagnosis, gender, sexual orientation, ethnicity, education, and employment status for use as potential covariates. For multivariable analyses, gender was dichotomized into male vs. non-male, sexual orientation into heterosexual vs. non-heterosexual, ethnicity into Caucasian vs. non-Caucasian, education into high school completion or less vs. greater than high school completion, and employment status into employed vs. not employed.
To measure HIV stigma, we used the 32-item version of the HIV Stigma Scale which measures stigma using four subscales (39, 40). The enacted stigma subscale (formerly personalized stigma) examines personal experiences of rejection, the negative self-image subscale deals with feelings of guilt or shame around being HIV positive, the disclosure concerns subscale refers to the need to conceal information regarding one’s HIV status, and the concern with public attitudes subscale measures what a person with HIV believes other people may think of them because of their HIV status (40). The disclosure concerns and concern with public attitudes subscales were merged to form a single anticipated stigma subscale and the negative self-image subscale was renamed to internalized stigma to match the HIV Stigma Framework (11, 41). Participants were asked to respond to each item using a 4-point Likert scale ranging from “strongly disagree” to “strongly agree”. Factor analysis showed that all items loaded into the same dimensions as described during the development of the scale (40). Subscale scores were calculated by taking the mean of all items in the subscale with higher scores indicating greater stigma. Internal consistency was high for subscales with enacted stigma, internalized stigma, and anticipated stigma having Cronbach’s alphas of 0.940, 0.894, and 0.901 respectively.
We utilized the 9-item Patient Health Questionnaire (PHQ-9) to measure depression. The scale covers the domains of depression as defined in the DSM-IV (which remain the same in the DSM-V) and can provide both a provisional depression diagnosis and a grade of depression symptom severity (42). Participants were asked “over the past 2 weeks, how often have you been bothered by any of the following problems” and presented with a list of 9 depressive symptoms to respond to on a 0-3 scale from “not at all” to “nearly every day”. Total depression score was calculated by taking the sum of all items with higher scores indicating greater depression. Scores from 0-9 are classified as “none” or “mild” depression and scores from 10-27 are classified as “moderate”, “moderately severe”, or “severe” depression (42). Factor analysis demonstrated that items on the PHQ-9 loaded on one dimension and internal consistency was strong with a Cronbach’s alpha of 0.852.
To measure self-rated health, we used the one-item self-report health question: “In general, how would you describe your health at the moment?”. Participants were asked to rate their health on a 1-5 scale from “poor” to “excellent”. This single item measure is often included in other popular validated measures on quality of life including the World Health Organization Quality of Life (WHOQOL) assessment, SF-36, and the QOL10 and allows participants to evaluate their health and life satisfaction as it relates to their own experience (43-46).
All statistical analyses were conducted using IBM SPSS Statistics version 24 (47). Participants with no data for any demographic variables, depression, stigma, or self-rated health were removed from analyses (n=54) leaving a final sample of N=676. First, we conducted multiple regression analyses to examine which variables had a significant impact on self-rated health. Variables were entered in blocks with demographic variables being entered first, then depression was added, followed by dimensions of stigma being added separately, and lastly all dimensions of stigma were added together. Results from the multiple regression guided the mediation analysis. The parallel mediation model was created using the PROCESS macro, a regression-based tool designed for conducting mediation, moderation, and conditional process analysis (48). The antecedent variable was enacted stigma, the mediating variables were internalized stigma and depression, and the outcome variable was self-rated health. Demographic variables were added as covariates to control for any possible confounding effect. Mediation was deemed significant if the bootstrap 95% confidence interval associated with the indirect effect did not include zero. Analysis of the indirect effect was conducted with 5000 bootstrap samples (48). Completely standardized indirect effects were computed to compare the effects of the mediators with each other by removing the scaling of antecedent and outcome variables.