The development of the landscaping survey was an iterative process incorporating feedback from multiple stakeholders. This included regional partners in Africa, Asia and LAC, originally brought together through a partnership award led by The Global Health Network (TGHN) and involving HDR UK’s Global programme team. Feedback was also provided by a working group of global health researchers involved in the Global Health Data Science Hub (26). This collaboration ensured excellent representation from, and understanding of the health research ecosystems in LMICs and the three regions. During this phase, the HDR Global team evaluated various survey delivery platforms (Survey Monkey, MS Forms, Jisc), along with quantitative and qualitative analysis tools, to select the most suitable options. Key considerations in the development process included adherence to HDR UK’s Good Research Practice Guide (27), regular consultations with HDR UK’s data protection team, and iterative feedback from global partners, TGHN and the working group. The team also reviewed relevant surveys and engaged experts in crowd consensus approaches to refine the survey content. The finalised survey was piloted internally with HDR UK’s Global programme team using dummy data, followed by further piloting with global partners in Africa, Asia and LAC to ensure that the survey questions were accessible, easy to interpret and understandable in the four languages in which it was distributed: Spanish, Portuguese, French and English. Translation was carried out by data scientists who are bi-lingual, native speakers. Similarly, the survey was piloted to ensure that responses—particularly qualitative responses—could be interpreted and accurately categorised into themes. The survey was delivered online in December 2023 and January 2024 in all four languages.

Data science—and how it is used—is truly interdisciplinary and not just relevant to one kind of researcher or practitioner. In this context, a convenience sampling approach was taken; we chose not to target specific stakeholder groups, as our focus was a wider group of individual stakeholders working in health research and health services. We wanted to seek diverse views and to be as inclusive as possible, and used our global partners’ existing broad regional, country and wider networks to reach individuals across Africa, Asia and LAC working in a wide range of health research related roles and institution types.

The overall methodological framework is described in Figure 1.

The survey consisted of three sections designed to capture comprehensive data from respondents. Respondents were required to give ‘Consent to proceed’ before beginning the survey.

The survey was distributed widely in four languages—Spanish, Portuguese, French and English—via TGHN, its global partners and HDR UK’s global networks across Africa, Asia and LAC, and remained open for 2 months, between Dec 2023–Jan 2024.

Data were collected through an online survey hosted on Microsoft Forms. To ensure maximum reach, the survey was widely disseminated by global partners through their regional networks as well as through the TGHN newsletter and via the Global Health Data Science digital hub (26). The team undertaking this observational study discussed ethical considerations with HDR UK’s Legal, Trust and Ethics team, and concluded that approval from an Institutional Research Board would not be required. While the survey asked for views and opinions, it did not generate any sensitive data and did not ask for provision of any data owned by institutions or institutional views, but for individual views. Ethical considerations included obtaining informed consent, ensuring anonymity and confidentiality, and securing data storage on HDR UK’s cloud platform, Box, with access restricted to HDR UK’s Global programme team and regional partners. De-identified data were prepared for analysis, with non-LMIC responses excluded and data stratified by regional groupings. To ensure validity of analysis, particularly accurate interpretation and appropriate categorisation of qualitative responses, translation was checked by data scientists based at the regional partner institutions, who are data scientists, native speakers and bi-lingual. Validation of categorisation was continuous and carried out by the team.

Data analysis followed a predefined plan to maintain consistency and validity. Quantitative data were analysed using R version 4.3.3, with descriptive statistics and visualisations employed to identify trends. Free text responses were thematically grouped through workshops using Miro, an online whiteboarding and workshop tool, and analysed at both global and regional levels. The analysis approach was tailored to the question type: drop-down menus and tick box lists allowed for quantitative analysis, while free text responses underwent qualitative analysis. For qualitative analysis, we opted to use Miro, an online collaborative tool, rather than software such as NVivo or MAXQDA. This decision was guided by the relatively small number of free-text responses, which made Miro’s visual and interactive features more suitable for our needs. Miro allowed for efficient organisation and collaborative coding in real-time, facilitating the engagement of multiple team members in the thematic grouping process. To ensure reproducibility, records of each stage of the analysis were maintained, creating a clear audit trail.

Specifically, for free text ‘Other’ responses in Section 2, a colour-coding system was implemented in Miro as a quality control measure: stickers were categorised by colour based on their relevance, with distinctions made between valid and relevant responses, irrelevant responses, and those with multiple skills that required splitting. The assessment of these stickers was conducted collaboratively by three members of HDR UK’s Global programme team. The primary goal for Section 2 was to ensure that new skills, beyond those listed in the multiple-choice options, were identified across all stages of the health data research project lifecycle, including: research planning, data access and management, data analysis, producing outputs and achieving impact, and community and stakeholder engagement. Stickers corresponding to existing essential data science skills were grouped under the relevant skills. Continuous review and validation were carried out by the analysis team, along with HDR UK’s wider Global programme team, to ensure accuracy in the thematic grouping process. Once all open-ended responses had been thematically categorised, a final review was conducted by the Global programme team to reach consensus on the newly identified skills.

For Section 3 responses (gaps, barriers, solutions), a similar approach was employed, with responses further sub-categorised by region. Key statements providing a comprehensive perspective on the challenges faced in different regions were marked for inclusion in the results section. A schematic diagram of this process is depicted in Figure 1.

Participants were asked to identify the top three essential health data science skills for each stage of the health data research project lifecycle (Figure 2), from a pre-defined list of skills (Supplementary Appendix 1). Participants were also invited to include additional skills beyond those listed.

While there was a high degree of consensus across all three regions on the top three essential skills for some of the stages, there was variation in the prioritisation of essential data skills across regions, underscoring the diverse regional needs and contexts.

Results of the analysis of responses for each stage of the health data research project lifecycle are shown below.

To understand the current health data science capacity building landscape in each of the three regions we asked respondents to tell us about any gaps in provision of health data science capacity building initiatives and the barriers to accessing provision.

Rich and varied responses to these questions were received. While there were some differences in regional perspectives, the results revealed a number of common themes in relation to gaps and barriers including: funding challenges; lack of training and mentoring opportunities; limited opportunities for networking and collaboration; challenges with data quality, access and data infrastructure; and lack of a conducive research environment. These themes are described in more detail below.

In addition to highlighting gaps in and barriers to accessing capacity building opportunities, we asked respondents to tell us about any potential solutions to address the gaps and barriers identified. Many practical solutions were suggested, and these constitute a valuable source of information for all stakeholders with an interest in health research and health data science capacity building, including funders and research institutions in LMICs wishing to develop or strengthen use of data science approaches. An analysis of suggested solutions was carried out, and a number of overarching themes were identified which are described in more detail below.

An overview of the insights provided by respondents to the survey in relation to gaps in and barriers to the provision of data science capacity building, and potential solutions, is provided in Table 1.

As part of this landscaping survey, respondents were asked about the freely available health data science training resources or courses they were aware of in their region. Overall, respondents across the three regions identified over 80 health data science training initiatives/resources ranging from: short courses, e-Learning courses, tutorials, manuals, online capacity building communities, MOOCs, capacity building courses offered by international organisations, and formal higher education courses. These resources have been collated and made available through a ‘Resource Pack’ [(29) Global Health Data Science Resource Pack].

Analysis of survey responses highlighted the high degree of consensus across regions on the essential health data science skills required for key stages of the health data research project lifecycle. This supports the case for the development of a structured and standardised global framework for capacity building in these skills. However, regional variations in skills prioritisation indicate the need for a common framework for health data science skills that can be tailored to regional contexts and needs.

Additionally, these findings align with existing global health initiatives, such as the WHO’s Digital Health Strategy (2), the Global Partnership for Sustainable Development Data (GPSDD) (11), and regional programmes led by The Global Health Network (TGHN) (13), which focus on capacity building through training support for data governance and infrastructure investments. Integrating recommendations from this study into these initiatives could enhance coordination and sustainability of health data science capacity building efforts.

Responses to questions about the gaps in data science capacity building provided rich insights into the challenges facing health researchers and practitioners in using data science approaches to address global health priorities. A major challenge identified by respondents included the limited funding and resources for training and research. These findings align with existing research that has highlighted the chronic underfunding of health research in LMICs (30, 31). The scarcity of robust funding mechanisms inhibits the development of both essential and sustainable research infrastructure. Moreover, provision of training and related opportunities for capacity building emerged as critical areas for improvement. The survey also highlighted a lack of access to relevant training resources and mentoring opportunities, echoing findings from similar studies that document inadequate professional development and educational resources in these regions (32–34). The need for localised and accessible training solutions is evident, as current offerings often fail to address the specific needs of health data science professionals in diverse roles.

The survey findings highlighted how data access and governance issues further complicate the landscape. Challenges with data quality, infrastructure and privacy are consistent with other studies that have identified these as major barriers to effective data science practices (35, 38). Furthermore, the variability in data infrastructure across regions highlights the disparity in data management capabilities and emphasises the need for improved data governance frameworks and infrastructure investments.

In response to these challenges, a number of tangible solutions have been identified, such as increasing the number of focused data science training opportunities and resources, and improving access to these, as well as increased funding for making improvements in data quality, access and infrastructure (36). Broader solutions were also identified, including enabling increased multidisciplinary collaboration and networking, and the development of more conducive research environments.

To effectively address the identified challenges, we propose a structured framework that outlines specific actions for key stakeholders:

Stakeholder	Recommended actions
Funders (e.g., government agencies, international organisations, private sector)	Increase long-term investment in health data science training programmes, provide grants for early career researchers, and establish funding streams for digital transformation, as well as approaches to improve data quality, access and infrastructure
Academic institutions	Develop standardised curricula for health data science, expand mentorship programmes, and integrate practical data science training into public health and medical education.
Governments and policy makers	Create national and regional data science hubs, support roll out of standardised health science curricula, establish data governance frameworks, and align digital health policies with global initiatives.
Research institutions and networks	Strengthen multidisciplinary collaborations, provide open access training materials, and develop knowledge-sharing platforms.

To measure the long-term impact of capacity building initiatives, future research should incorporate longitudinal studies to track progress over time. These studies would monitor skills development and its effect on the workforce, evaluate the effectiveness of funding and policy interventions, and examine changes in data governance and infrastructure within LMICs. By providing valuable insights into the sustainability and scalability of health data science programmes, such research would help ensure continuous improvement and adaptation to evolving challenges and opportunities.

The survey had 262 respondents representing 54 LMICs and a broad range of roles and organisations, suggesting that the results capture a wide range of perspectives and offer a comprehensive view of the current landscape. Key recommendations emerging from the survey include: