MAPPING THE FUTURE OF HEALTH DATA SCIENCE: INSIGHTS FROM AN EXPERT SURVEY
Author(s)
Olivia Oliveira1, Alberto Freitas, PhD2, Rute Almeida, PhD1, Hernani Gonçalves, PhD1, Mariana Lobo, PhD1.
1Faculty of Medicine University of Porto, Porto, Portugal, 2Faculty of Medicine of University of Porto, Porto, Portugal.
1Faculty of Medicine University of Porto, Porto, Portugal, 2Faculty of Medicine of University of Porto, Porto, Portugal.
OBJECTIVES: To map expert perspectives on emerging trends, key challenges and underutilized data sources shaping Health Data Science, and to synthesize actionable research and implementation priorities relevant to digital health, real-world data governance, and health outcomes research communities.
METHODS: Cross-sectional online expert survey (n=110) using purposive sampling via professional networks (LinkedIn). Respondents reported role, years of experience, and views on emerging trends, challenges, and underutilized data sources. Analyses included descriptive statistics, chi-square tests with Cramér's V for role- and experience-stratified comparisons, co-occurrence analysis, and tripartite network analysis to map thematic clusters across all three item sets. Study reported per CROSS checklist.
RESULTS: AI and Large Language Models (LLMs) were the most endorsed emerging trend (84.5%), followed by precision medicine (61.8%) and predictive analytics (49.1%). Leading barriers were data quality/standardization/integration (77.3%) and interoperability (58.2%). EHRs/clinical notes (64.5%) and wearables/remote monitoring (50.0%) were the most frequently identified underutilized high-potential resources. Stakeholder priorities varied by role: consulting respondents prioritized federated learning and predictive analytics; industry favored precision medicine; whereas academic respondents prioritized explainability. Co-occurrence network analysis identified four coherent thematic clusters: (1) explainable AI linked to adoption/trust; (2) precision medicine/omics constrained by interoperability and regulatory hurdles; (3) telehealth and predictive analytics anchored to wearables and IoT sensors; and (4) a multimodal AI/LLM-centred super-cluster bridging heterogeneous data streams, with scalability and adoption flagged as critical socio-technical bottlenecks.
CONCLUSIONS: Experts anticipate an AI-forward future constrained by data infrastructure and implementation realities. Three expert-aligned priorities emerge: (1) trustworthy clinical AI; (2) interoperable real-world data infrastructure; and (3) longitudinal multimodal patient modelling. These provide an evidence-based agenda with direct relevance to HEOR research prioritisation, digital health assessment, and real-world data governance.
METHODS: Cross-sectional online expert survey (n=110) using purposive sampling via professional networks (LinkedIn). Respondents reported role, years of experience, and views on emerging trends, challenges, and underutilized data sources. Analyses included descriptive statistics, chi-square tests with Cramér's V for role- and experience-stratified comparisons, co-occurrence analysis, and tripartite network analysis to map thematic clusters across all three item sets. Study reported per CROSS checklist.
RESULTS: AI and Large Language Models (LLMs) were the most endorsed emerging trend (84.5%), followed by precision medicine (61.8%) and predictive analytics (49.1%). Leading barriers were data quality/standardization/integration (77.3%) and interoperability (58.2%). EHRs/clinical notes (64.5%) and wearables/remote monitoring (50.0%) were the most frequently identified underutilized high-potential resources. Stakeholder priorities varied by role: consulting respondents prioritized federated learning and predictive analytics; industry favored precision medicine; whereas academic respondents prioritized explainability. Co-occurrence network analysis identified four coherent thematic clusters: (1) explainable AI linked to adoption/trust; (2) precision medicine/omics constrained by interoperability and regulatory hurdles; (3) telehealth and predictive analytics anchored to wearables and IoT sensors; and (4) a multimodal AI/LLM-centred super-cluster bridging heterogeneous data streams, with scalability and adoption flagged as critical socio-technical bottlenecks.
CONCLUSIONS: Experts anticipate an AI-forward future constrained by data infrastructure and implementation realities. Three expert-aligned priorities emerge: (1) trustworthy clinical AI; (2) interoperable real-world data infrastructure; and (3) longitudinal multimodal patient modelling. These provide an evidence-based agenda with direct relevance to HEOR research prioritisation, digital health assessment, and real-world data governance.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MSR228
Topic
Methodological & Statistical Research, Real World Data & Information Systems, Study Approaches
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas