The Boloka data repository will leverage and collate available and diverse data from various sources across places, times, and populations (Multimedia Appendix 1 shows the list of potential or current data partners and data types). The inclusion and exclusion criteria (Textbox 1), as well as the Boloka data indicator tool (Multimedia Appendix 2), will guide the eligibility of the data source. The inclusion of HIV-related data from the period 2000 and onwards will enable the capturing of the developments that took place over time in the HIV field. The use of reliable and granular data collected over time is essential for improving population health, setting targets as well as developing strategies for targeted HIV response.

The team reviewed the Joint United Nations Programme on HIV/AIDS global monitoring tool and the National Department of Health (NDoH) national indicator dataset form used for global, regional, and national HIV reporting to identify primary and secondary indicators in line with global, regional, and national priorities. The team also reviewed existing validated questionnaires or instruments from various data partners to further refine priority indicators; from this, broader categories of HIV-related interest areas were formed (Figure 2). Accordingly, each potential data partner is required to indicate the primary and secondary indicators that they have collected data on the Boloka data indicator tool (Multimedia Appendix 2).

Where possible, data will be disaggregated at individual, facility, district, or sub-district levels to enable advanced analyses that are not possible with higher-order aggregated data. This will help us develop an in-depth understanding of various subsets of the populations within the larger datasets. Where data are aggregated, we will seek data disaggregated by sex, gender, age groups, socioeconomic status, geo-location, facility type, and temporal factors to understand a range of HIV indicators, including heterogeneity of HIV risk and burden, engagement in HIV services (treatment cascade, pre-exposure prophylaxis uptake and continuation), and population size estimates, among others. The process for requesting and accessing data varies based on the data type and institution. As such, the process of acquiring and collating the data for the different data types is discussed below.

For open access data, which is typically aggregated or deidentified individual data, we will adhere to procedures set by the respective institutions to obtain access in line with the POPI Act. This may require the submission of designated data access request forms before gaining direct access to the dataset. For data that are only available upon request, we will contact the data partner to understand their specific procedures for accessing data. This typically requires the completion of a data access request form and the sharing of relevant information, such as the project proposal, ethics approval, and timelines for analysis. For institutions such as the NDoH, district health information data is made available for public use upon request and completion of their data user agreement (DUA) forms. The data or indicators requested should be aligned with the approved National indicator dataset, which outlines the data elements and indicators collected by the NDoH. In such cases, the request and DUA forms will be completed to access data. For institutions in possession of data that are not typically available for public access (eg, implementing partners), we will seek to develop formal partnership agreements and additionally use the University of Johannesburg DPA to ensure adherence to the Protection of Personal Information Act in terms of data privacy and security before any data sharing. Once agreed upon, a formalized partnership agreement will be approved by the legal team of both parties. These agreements will be reviewed in accordance with the policies of the respective organizations.

For institutions in possession of data that is not typically available for public access, we will seek to develop formal partnership agreements and additionally use the University of Johannesburg DPA to ensure adherence to the POPI Act in terms of data privacy and security before any data sharing. Once agreed upon, a formalized partnership agreement approved by the legal team of both parties will be signed. These agreements will be reviewed in accordance with the policies of the respective organizations.

Once data are acquired and determined to be eligible according to the inclusion and exclusion criteria (Textbox 1), they will go through a process of screening and assessment for accuracy, relevance, quality, completeness, and consistency before inclusion in the Boloka data repository. Due to limited information and validated tools on how to specifically check for relevance and quality for some of the data types [30], we will adopt the Framework for Data Quality developed by the Federal Committee on Statistical Methodology (FCSM) [31] and the Information Quality Assessment Framework [32]. Below are further details on the data screening process by the different data types.

For research data, a multistep quality assessment process will be carried out using the Global HIV Quality Assessment Tool (Multimedia Appendix 3) [33]. This tool will be used to review and verify the suitability and quality of research data sources in terms of study design and implementation, along with criteria for HIV indicators, specifically prevalence, incidence, engagement in the HIV care continuum, and population size estimates [33]. There will be close supervision and checks to minimize errors by using an honest data broker mechanism to manage and maintain datasets. In addition, 2 individuals (research assistants) will perform the initial quality assessment. Any discrepancies identified will be addressed by a third assessor (Project Manager and Principal Investigator [PI]). Exploratory analyses will be conducted to understand the nature of the data and identify outliers and missing data that will be cleaned for data quality and accuracy. This will be an iterative process to ensure that we leverage and assemble the best available data.

It is important to highlight that program data present unique data quality challenges. Program data are typically aggregated to administrative units, and information on individuals is generally not available. Hence, it is not always feasible to link program exposure directly to an outcome [34]. In terms of screening the program data for inclusion into the Boloka data repository, the World Health Organization’s “Data quality review: a toolkit for facility data quality assessment—Module 1: Framework and metrics” will be adapted for use [35]. This framework is made up of 4 dimensions. For this study, dimensions 1 and 2 will be used: completeness and timeliness of the data (dimension 1) and internal consistency of reported data (dimension 2). Dimension 1 focuses on the data elements and indicators collected. For example, the NDoH and implementing partners’ data elements and indicators will be assessed and checked for completeness [35]. Dimension 2, that is, the internal consistency of the data, relates to the coherence of the evaluated data. In this regard, 4 metrics of internal consistency will be used [35]: (1) presence of outliers, (2) consistency over time, (3) consistency between indicators, and (4) consistency of reported data.

In this initial screening, articles, reports, and datasets will be excluded if their scope does not adhere to the inclusion criteria mentioned in Textbox 1. Data that focuses on the general population will be included in addition to data with a focus on KPs, as the general population provides important KP contextualization. Furthermore, data that have been empirically collected will be prioritized. Estimates and modeled data will be included but carefully reviewed to ensure they meet quality standards and are deduplicated within the datasets.

A report detailing the results of the quality assessment will be written. The report will include the recommendations for data quality improvement plans considering the context and existing constraints. A feedback loop will be used to engage with stakeholders on the potential data quality issues and anomalies, such as completeness of data, discrepancies, and inaccuracies. In addition, the data quality results will be available and stored alongside the data in the repository as part of the metadata.

Data will be extracted from data sources and put into a staging area before being transformed and loaded into the data repository (Figure 3). We will use the Data Intensive Research Initiative of South Africa (DIRISA) platform, which will enable the provision of much-needed timely information at a level and scale that will improve understanding of HIV heterogeneities. Structured data will be preprocessed and normalized into a standardized format using the Boloka Data Harmonization Tool (Multimedia Appendix 4). The repository will be flexible and updatable for structured data by design [36]. Unstructured data will be stored in the NoSQL data store. Preprocessing will entail data cleaning, transformation, and integration to make the data complete for analysis. Data management will involve assembling multiple big, existing HIV data sources and datasets, capturing, preprocessing, cleaning, integrating, and building them into the Boloka data repository.

For disaggregated data, personal identifiable information will be encrypted and hashed to ensure anonymity and protect privacy. This role is also performed by the appointed honest data broker. Ideally, the Boloka data repository will not store any identifiable personal information, but a procedure in accordance with the POPI Act will be put in place for handling this issue if such information comes with the data. The honest data broker mechanism will be set up to screen the data for compliance in terms of the latter before it passes to the Boloka data repository. The resulting data repository will be responsive to the size and scope of the collected data. It will be dynamic and flexible to accommodate diverse data that has a high number of dimensions.

The access to the data repository will be secured with login credentials (username and password) in conjunction with multifactor authentication (MFA) to ensure compliance with the POPI Act and avoid any confidentiality breaches as well as contravention of human rights. Access to the repository will be made to the PI and Project Manager through a formal and written request. Login credentials (password) will be updated on a 3-month basis to enhance layers of security and reduce the risk of unauthorized access.

The stakeholder engagement is ongoing using the data partner tracking tool (Multimedia Appendix 7) listing potential and current stakeholders and their level of engagement with data (eg, district, provincial, and national) is in place.

The research team secured open access data from 5 (2002, 2005, 2008, 2012, and 2017) South Africa National HIV Prevalence, Incidence, Behavior and Communication Surveys from the Human Sciences Research Council (HSRC) on January 5, 2023. Since the data is open access, no signing of DPA was required to access and secure the data. Engagement with stakeholders is expected to conclude in December 2026 in terms of finalizing data sharing agreements, and an example of the University of Johannesburg DPA has been shared for consideration by the potential data partners. An engagement with NDoH was done in 2022, and HIV-related routine data was secured (April 2024).

The research team had a series of meetings with SANAC regarding the Boloka data repository. These meetings have established a strong partnership and system for the coordination of data-sharing partnerships. The partnership with SANAC is essential in ensuring that the Boloka data repository is used to guide national HIV strategy related to KPs. This strong relationship will ensure the complementarity of the Boloka data repository rather than duplication of SANAC’s efforts to develop “The Situation Room,” a central HIV data repository for program data in South Africa. The Situation Room, still in its infancy stages, seeks to enable data checks and balances, dynamic visualization, and sharing of the national, provincial, and district HIV data in the general population to monitor progress toward reaching set targets. The Boloka data repository will be a resource for KP data toward SANAC’s efforts. SANAC introduced the research team to its implementation partners to acquire potential data partners, namely Beyond Zero [37], Networking HIV and AIDS Community of Southern Africa (NACOSA) [38], Sex Worker Education Advocacy Taskforce (SWEAT) [39], Sisonke [40], and AIDS Foundation South Africa (AFSA) [41] as well as Perinatal HIV Research Unit (PHRU) [42].

A series of engagements have been carried out with the respective organizations; the engagements are at different stages. Beyond Zero has been engaged since September 2022, and a final copy of the University of Johannesburg and Beyond Zero data-sharing agreement is under review by the Beyond Zero legal team before it is signed by both parties. A data partnership agreement with PHRU is expected to be finalized by the end of December 2024. For this partnership, the PHRU data-sharing agreement is being developed by their legal department. With AFSA, the agreement is in the early stages of communication. In addition to engagements with SANAC and its implementing partners, the research team has consulted with other community leaders who work or worked with community-based organizations, such as treatment action campaigns.

Significant progress has been made in securing program data, published research data, and technical reports. The security of the acquired data is prioritized during this interim period; thus, the data have been stored in a secure staging area and will later be moved to appropriate data storage software.

Open access data from 5 cycles (2002, 2005, 2008, 2012, and 2017) of the population-based multistage cluster cross-sectional of up to 85,000 randomly selected households in South Africa has been secured from the HSRC (January 5, 2023). The South African HIV Behavioural, Sero-status and Media data is within the restricted access level. The data was accessed by completing a non–web-based internet data request form detailing the name of the project, a brief description of the intended use, and the expected date of project completion. The surveys provide data on HIV incidence, prevalence, antiretroviral therapy (ART), viral load suppression, drug resistance, risk behaviors, and HIV care, among others. The Key Population Implementation Science (KPIS) data on HIV testing and engagement in care was secured at Emory University (May 16, 2022) for the degree purpose of postgraduate students. Further, data on HIV testing proficiency was generated by a postgraduate student from the health care facilities in Eastern Cape Province (May 2023).

Data from the Demographic Health Surveys Program has been secured (August 2022). This required registration and submission of a data access request form along with the justification of the request. The demographic health survey collects indicators such as HIV prevalence, prevention and treatment, stigma, and discrimination, as well as sexual behavior. We also secured HIV-related data from the NDoH, and the data includes the following HIV indicators: HIV testing and HIV prevention; ART (ie, ART initiation, ART rate, viral load & CD4 T lymphocyte count testing, ART type, ART adherence); sexually transmitted infections; maternal and neonatal (antenatal HIV test, ART initiation, ART adherence, live birth, infant HIV test); management of inpatients and management of primary health care facility.

There are plans to use data from the 1173 NDoH High Transmission Area sites across South African provinces based in communities that function like clinics and provide services to KPs. The NDoH’s High Transmission Area program reaches key and vulnerable populations with HIV, tuberculosis, and sexually transmitted infections prevention and management services in key hotspots across the country.

This protocol will also leverage a partnership with the largest HIV service provider for key populations in South Africa, TB HIV Care, a nonprofit organization, building on a decade-long period of dynamic and exciting work; preliminary discussions have already taken place with the Chief Executive Officer. TB HIV Care focuses on preventing, finding, and treating HIV and TB across 7 provinces and 22 districts in South Africa [43]. Its KP program provides services to sex workers and people who use drugs. Furthermore, data-sharing agreements are being drafted and checked for Beyond Zero, NACOSA, and AFSA, as detailed above.

The data received in stage 2 went through screening and assessment for accuracy, completeness, and consistency before its placement in the staging area. The data received was screened and filtered for relevance and inclusion by a member of the research team using the FCSM and the Information Quality Assessment Framework [32] where applicable.

South African HIV Behavioural, Sero-status and Media data has undergone a quality check to assess the completeness of the data received. Using the Needs Assessment Form, the DHIS data received has been checked to assess the relevance of the data to the project aims. The assessment checked that the data included the relevant indicators and if the data was complete. This process has been done by 2 researchers on the team.