This was a cross-sectional study conducted through telephone interviews between 25 September 2020 and 8 January 2021, prior to the availability of COVID-19 vaccines in South Africa.
We contacted a subset of participants enrolled in a prospective observational cohort study evaluating a decentralized ART delivery program in KwaZulu-Natal province, South Africa. The parent study recruited participants meeting eligibility criteria for the national Chronic Medicine Dispending and Distribution (CCMDD) program (i.e., not pregnant, on ART for ³1 year, and virologically suppressed as per program guidelines) in nine public sector clinics offering the CCMDD program in the urban township of Umlazi . At enrollment into the parent study, participants completed a baseline questionnaire assessing demographic characteristics, HIV care history, barriers to HIV care, competing needs, mental health, and social support, and agreed to be contacted later by telephone. Competing needs assessed whether in the preceding 6 months, the participant had gone without healthcare because they needed the money for basic needs, such as food, clothing, or housing, or if they had gone without basic needs because they needed the money for healthcare . Baseline data were collected between October 2018 and March 2020.
For this study, we contacted a random subsample of the 2,220 participants enrolled in the parent study . 900 participants were chosen using random sampling, stratified to include equal numbers of participants enrolled in the decentralized ART delivery program for 0-6 months, 6-12 months, and >12 months, and equal proportions from each clinic site. Questions pertaining to COVID-19 vaccination were added partway through the process of telephone interviews for this randomly selected subsample and completed for all subsequent respondents. We did not calculate an optimal sample size; the number of participants included in the study was dictated by staff availability and response rate. Three trained, bilingual research assistants telephoned participants at the number provided at parent study enrolment. Participants who provided verbal informed consent were administered a semi-structured questionnaire by the research assistant in their preferred language (isiZulu or English), with each interview lasting 25-30 minutes.
Willingness to accept COVID-19 vaccination and overall vaccine confidence
To evaluate levels of acceptance of a future COVID-19 vaccine, we asked participants, “do you intend to accept future COVID-19 vaccination for yourself?” We asked participants to list all concerns regarding a future COVID-19 vaccine. For further context, we assessed overall vaccine confidence, defined as the “belief that vaccination serves the best health interests of the public and its constituents”  along with trust in vaccines, vaccine providers, and vaccine decision-makers . To assess vaccine confidence, we asked for level of agreement or disagreement with four statements evaluating importance, safety, effectiveness, and religious compatibility of vaccination, adapted from a recent worldwide study of vaccine confidence . To ascertain underlying local rates of adult vaccination in this population, we asked whether participants had been vaccinated against seasonal influenza in 2019, and reasons for not doing so for those who did not. In order to assess the internal consistency of our question on willingness to accept vaccination, we created a summary COVID-19 vaccine confidence measure, assigning one point for those reporting no concerns regarding COVID-19 vaccination, and one additional point for disagreement with each of the following statements: “if a vaccine were available to prevent COVID-19 in the future, I: ‘would not want to get it’, ‘would not trust it’, or ‘am worried that it could be harmful’”. These items were adapted from a previous study of COVID-19 medical mistrust and vaccine hesitancy among PLWH . The measure had a possible score range of 0-4, with higher scores indicating greater COVID-19 vaccine confidence.
HIV care history, reactions to COVID-19, stigma, and medical mistrust
Demographic data, HIV care history, and measures of healthcare access were obtained from the baseline questionnaire of the parent study. In the COVID-19 telephone interview, we asked participants about their sources of information on COVID-19 and changes in daily activities due to the pandemic (all that apply). We defined ‘recommended’ changes in daily activities as those falling within the recommendations of the National Department of Health of South Africa . The activities include physical distancing (avoiding large gatherings, not hugging other people, avoiding people who present with symptoms, avoiding public transport, not going outside), mask wearing, and hygiene (washing hands, not touching face, using hand sanitizer frequently). We assessed stigma related to COVID-19 using six questions adapted from previously published stigma scales for HIV and chronic illness [20–22] and described in detail previously . We assessed medical mistrust related to COVID-19 using seven questions; two adapted from a published scale assessing conspiracy theories around HIV  and the remainder developed for the current study, as described previously . We defined medical mistrust as distrust in healthcare systems and medical providers with the belief that they are acting against one’s best interest [24–26]. Responses to stigma and medical mistrust questions were on a 5-point Likert scale with scores ranging 0-4, with higher scores indicating higher stigma or medical mistrust. We calculated summary scores for overall COVID-19 stigma and medical mistrust by adding the scores for each individual question.
We used descriptive statistics (median, interquartile range [IQR], frequency) to report baseline and COVID-era participant characteristics, sources of information on COVID-19, levels of COVID-19 stigma and medical mistrust, and responses to questions on COVID-19 vaccination and general vaccine confidence. For modelling measures of medical mistrust and stigma related to COVID-19, we categorized data into above and below the median to use the two variables consistently in the analysis. This also allows for easy interpretation of the effect of the variables. We determined internal consistency among the questions on COVID-19 vaccine confidence using Cronbach’s alpha. A Cochran-Armitage test of trend was used to assess the relationship of willingness to accept vaccination with the summary COVID-19 vaccine confidence measure. We used univariate and multivariable logistic regression to assess predictors of willingness to accept COVID-19 vaccine (as defined by an answer of “yes” to “do you intend to accept future COVID-19 vaccination for yourself?”). Factors with p<0.05 in univariate logistic regression models were included in a multivariable model in addition to age and gender, which were pre-specified. All reported p-values were two-tailed, and p<0.05 was considered statistically significant. Analyses were conducted using SAS software (version 9.4, SAS Institute, Cary, NC).
The study protocol was approved by the Biomedical Research Ethics Council of the University of KwaZulu-Natal (Protocol BE092/18) and by the Partners Healthcare Institutional Review Board (Protocol 2017P001690).