Abstract
Informed citizens suggested a transparent culture of research for the public benefit, and continuous improvement of technology to protect patient privacy, to mitigate their concerns regarding privacy risks of using patient text data.
Introduction: The Value of the Narrative
Electronic Health Records (EHRs) are a cornerstone of modern medical research, yet they are fundamentally divided into two categories: structured and unstructured data. Structured data consists of standardized codes for diagnoses and medications, while unstructured data—often referred to as "free-text"—comprises the narrative notes, letters, and reports written by clinicians.
While structured data is easier to de-identify and share, it often lacks the nuance and context found in clinical narratives. However, data custodians are frequently hesitant to share free-text due to the difficulty of ensuring 100% de-identification. This study investigates whether an informed public supports the use of this more revealing text data for research purposes.
Methodology: The Citizens' Jury
To capture informed public opinion, researchers employed a "Citizens’ Jury." Unlike standard opinion polls, this deliberative method involves recruiting a demographically representative group to engage deeply with a complex topic.
The jury consisted of 18 individuals selected to reflect the UK's demographic mix in terms of age, gender, ethnicity, and education level. Over three days, these jurors heard evidence from expert witnesses—including clinicians, computer scientists, bioethicists, and privacy campaigners—before deliberating and voting on recommendations.
Key Insights: Pragmatism and the "Social License"
The jury process revealed several critical insights into public attitudes:
- Contextual Importance: Jurors recognized that free-text provides the "flesh" on the "bones" of coded data, offering essential context for high-quality research.
- Acceptance of Risk: Participants generally accepted that 100% de-identification is currently unattainable. They viewed a "slight risk" to privacy as an acceptable trade-off for improved medical treatments and saved lives.
- The Power of Information: Deliberation significantly shifted perspectives; 12 out of 18 jurors changed their minds during the process, with the majority becoming more supportive after understanding the safeguards and benefits.
- Hybrid De-identification: There was a clear preference for a combination of human and automated (NLP) processing to scrub identifiers, rather than relying on a single method.
Results and Recommendations
By the conclusion of the study, 100% of the jurors expressed support for using free-text data in health research. Key recommendations included:
- Opt-out Model: A majority (12/18) preferred an "opt-out" system to prevent the data bias inherent in "opt-in" models.
- Universal Application: Jurors were equally willing to share data regarding mental health as they were for physical health, provided safeguards remained consistent.
- Mandatory Transparency: The primary concern was not the data use itself, but the current lack of public awareness. Jurors demanded clear, accessible communication (e.g., posters and leaflets) explaining how patient data is utilized.
Conclusion and Future Directions
The study underscores that public trust relies on a "social license"—a commitment to transparency and patient agency. While the small sample size is a limitation, the findings suggest that an informed public is highly supportive of sharing unstructured EHR data for the greater good. Future work should focus on scaling these deliberative methods and developing better mathematical risk estimates for re-identification to further inform public and policy decisions.
Summary updated