UK Public Support for Health Data Sharing in AI Research: BMJ Focus Groups Reveal Willingness Amid Concerns

The Growing Role of AI in UK Health Research

Artificial Intelligence (AI) is transforming healthcare research across the United Kingdom, promising faster diagnoses, personalized treatments, and more efficient patient care. From advanced imaging tools that detect cancers earlier to natural language processing that sifts through electronic health records, AI relies heavily on vast datasets of anonymized health information. Universities like the University of Oxford are at the forefront, developing algorithms that could revolutionize orthopaedics, rheumatology, and beyond. However, this progress hinges on public willingness to share sensitive health data, raising questions about privacy, security, and trust. Recent research highlights a nuanced public stance: supportive yet cautious.

Landmark BMJ Study Explores Public Perceptions

A groundbreaking study published on February 9, 2026, in BMJ Digital Health & AI delves into UK public views on sharing health data for AI research. Titled "Public perceptions of health data sharing for artificial intelligence research: a qualitative focus group study in the UK," it was led by Rachel Kuo from the University of Oxford's Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences (NDORMS), in collaboration with Oxford University Hospitals NHS Foundation Trust. The paper, co-authored with patient and public involvement (PPI) contributors, analyzes discussions from eight online focus groups involving 41 diverse participants. It identifies factors shaping conditional support for data sharing, offering vital insights for researchers and policymakers.

Study Design and Participant Diversity

Conducted between May and July 2024, the research used purposive sampling to ensure representation across age (23-78 years, mean 56.7), gender (22 female), ethnicity (32 White, others including Asian, Black, mixed), income (£50,000-£69,000 median household), education (25 with degrees), health status (30 with chronic conditions), and geography. Participants were recruited via the NIHR Biomedical Research Centre Patient Research Registry and social media, receiving £25 reimbursement. Each 90-minute Microsoft Teams session explored three realistic scenarios: university-led cancer research, broad mental health databases, and international/commercial child data projects. Transcripts underwent inductive thematic analysis using NVivo, guided by the SRQR reporting checklist for rigor.

Perceived Risks: The Core Concerns

Participants universally acknowledged risks in health data sharing, starting with the limits of anonymization. While seen as essential, many doubted its infallibility, especially for rare conditions or linkable datasets. "How anonymous, anonymous data is, is really hard to say," one noted. Data sensitivity—mental health, STIs, family history—amplified fears of emotional harm. Governance and security were pivotal; demands included regular audits and breach accountability. Custodianship mattered: universities and NHS earned higher trust for altruistic motives, while commercial entities faced skepticism unless tied to patient benefits. These perceptions influenced comfort levels, highest for targeted university studies and lowest for broad or profit-driven uses.

Risk-Benefit Weighing: Altruism Meets Caution

Individuals assessed sharing through a personal lens, balancing harms like discrimination, insurance denial, or future misuse against benefits such as altruism and improved care. Those with chronic conditions often prioritized the "greater good," viewing data as a tool to "save lives" or speed treatments. Yet, fears of unintended consequences, especially for children (no future opt-out), tempered enthusiasm. This theme underscores a pragmatic calculus: willingness rises with demonstrable public value, like faster diagnostics, but plummets without safeguards against personal repercussions.

Informed Consent: Foundation of Public Trust

Consent emerged as trust's bedrock, requiring clear, concise, scenario-specific information on AI's purpose, data scope, and uses. Participants favored opt-in processes via trusted clinicians, avoiding high-stress moments like hospital visits, with cooling-off periods and easy withdrawal. "Clear, concise," was a refrain, emphasizing accessible language over jargon. Tailored communication—explaining step-by-step how data trains AI models—could bridge knowledge gaps and foster equity, particularly for underrepresented groups.

Voices from Participants and Researchers

Raw quotes paint vivid pictures: "Illness is money for them" captured commercial distrust, countered by "mutual benefit" for regulated partnerships. Lead author Rachel Kuo stated, "Public trust can't be taken for granted... people are willing under clear conditions like transparency and governance." PPI contributor Judi Smith highlighted dual mindsets: reservations as a "person" versus eagerness as a "patient." These insights, co-produced with public input, ensure the study resonates authentically.
For more on pioneering AI health studies at UK universities, check our research jobs listings.

Implications for UK Higher Education and AI Research

As hubs of innovation, UK universities like Oxford drive AI health advancements but must prioritize ethical data practices. The study bolsters calls for frameworks like the FUTURE-AI principles (fairness, universality, traceability, usability, robustness, explainability). Institutions can leverage trusted status to lead federated learning—sharing model insights without raw data—enhancing equity. Faculty and postdocs in AI-health intersections will find opportunities in grants emphasizing public engagement. Explore postdoc positions or academic CV tips to join this field.

UK's Broader Health Data Ecosystem

The findings align with national efforts like the January 2026 National Data Library update, promoting AI-ready data sharing across silos. Initiatives such as Our Future Health, Genomics England, and UK Biobank exemplify progress, while the Goldacre Review advocates trusted research environments (TREs). Public attitudes trackers show steady support (e.g., 51% positive on tech's care impact), but underscore commercial wariness. Read the full BMJ study or Oxford's press release.

Persistent Challenges in Data Sharing

• Bias and Equity: Underrepresented data risks amplifying disparities, especially for ethnic minorities.
• Commercial Pressures: Profit motives erode trust without oversight.
• Technical Hurdles: Data quality, interoperability across NHS systems.
• Regulatory Gaps: Aligning UK rules with EU Health Data Space (EHDS).

These issues demand multidisciplinary solutions from academia.

Solutions and Best Practices Emerging

Recommendations include PPI in study design, dynamic consent platforms, and transparent reporting. Universities can pioneer secure data enclaves, as in NHS TREs, and educate publics via workshops. Step-by-step: 1) Assess data needs; 2) Engage communities; 3) Implement governance; 4) Monitor outcomes. Aspiring researchers, bolster your profile with free resume templates tailored for higher ed roles in AI.

Future Outlook for AI-Driven Health Discoveries

With conditional public backing, UK AI health research could accelerate, from predictive orthopaedics models to pandemic forecasting. Expect expanded collaborations, bolstered by 2026 policies. Universities remain pivotal, fostering talent amid rising demand—explore UK university jobs. This study signals optimism: informed publics propel ethical innovation.

Photo by Sasun Bughdaryan on Unsplash

In summary, the BMJ study illuminates a path forward for health data sharing in AI research, emphasizing trust through transparency and involvement. For academics eyeing this dynamic field, visit Rate My Professor, Higher Ed Jobs, Career Advice, University Jobs, or post a job to connect with top talent.