Experts call for action to fulfil the potential of real-world data in research

A recent perspective by former FDA commissioner Robert Califf published in the New England Journal of Medicine calls for smarter data capture, better access, and regulatory innovation to unlock the full potential of RWD in clinical research.

There is no doubt that real-world data (RWD) have long been regarded as a transformative resource for clinical research and health policy. By capturing insights from routine healthcare delivery, RWD offers the promise of faster evidence generation, more informed decision-making, and improved population health outcomes. However, as underscored in a recent New England Journal of Medicine perspective by Ali B Abbasi, Lesley H Curtis, and Robert M Califf, realizing this vision remains a challenge. Structural, technical, and regulatory hurdles continue to limit the utility of RWD in generating high-quality, actionable evidence.

Despite 96% of US hospitals adopting electronic health records (EHRs), the authors note these systems are, “designed to document individual episodes of care and facilitate the highest possible payment for that care, not to support clinical research.” Data that fall outside reimbursement requirements are often incomplete or poor quality, undermining efforts to use RWD in observational studies and regulatory submissions.

Observational studies using RWD can provide important insights, but establishing causality is challenging. Confounding, inadequate documentation of risk factors, and incomplete capture of outcomes like mortality remain persistent issues. For interventional trials, the situation is even more complex. As the authors explain, these studies are still, “dependent on error-prone, expensive, and labor-intensive manual data collection.”

Three priorities for progress

The authors draw on their experiences at the US FDA to propose three key approaches to overcome these challenges:

1. Strengthen data quality at the point of care

Without clear incentives for clinicians, high-quality structured data are rarely collected in EHRs. The authors argue that insufficient attention has been paid to rewarding clinicians and health systems for higher-quality RWD collection. They propose strengthening initiatives like the CMS Coverage with Evidence Development program, linking reimbursement to participation in evidence generation.

Emerging technologies could also play a role. Artificial intelligence (AI) tools, including large language models (LLMs), offer opportunities to streamline documentation and reduce burdens on clinicians. But the authors caution: “Ambient AI scribes… cannot record information on topics that weren’t discussed and may create fictitious data.” Careful investment and workflow redesign are needed to ensure AI enhances, rather than undermines, data integrity.

2. Expand access to comprehensive, longitudinal data

Patients often receive care across multiple sites, resulting in fragmented datasets that limit the utility of EHRs for research. As the authors observe, “EHR data from individual institutions – no matter how well curated they may be – are of limited utility in isolation.” To address this, they propose modifying the Trusted Exchange Framework and Common Agreement (TEFCA) to permit health data exchange for research purposes. They also recommend establishing a national all-payer claims database to help close gaps in claims data and investing federally to reduce the costs and delays associated with accessing mortality data from the National Death Index.

Importantly, the authors highlight that whole-person, longitudinal health data should encompass not only clinical information but also factors such as nutrition and social determinants of health (SDOH), which play a critical role in health outcomes. “Collecting whole-person, longitudinal health data will require strengthening linkages among data sources and expanding the types of data that are available for linkage,” they stress. For example, data from financial services could help researchers examine how shopping and dining habits affect nutrition-related outcomes, with appropriate participant consent. Organizations such as PCORnet and professional registries could act as hubs for coordinating and linking these diverse datasets.

3. Modernize regulatory frameworks for RWD use

Finally, the authors argue for modernizing regulatory oversight to better accommodate the use of RWD in clinical research. They emphasize the need for additional training for FDA reviewers and inspectors on evaluating RWD and call for greater investment in regulatory science, particularly to support the safe and effective use of AI tools in evidence generation. As they point out, “the FDA currently lacks the expertise and funding to proactively develop safe uses of AI tools, including large language models, in this rapidly evolving area.”

They also stress the importance of updating ethical frameworks to reflect the realities of using RWD in research. Current regulations define research risks in absolute terms, which, they argue, can impede the conduct of comparative-effectiveness trials. Instead, they propose that risks should be assessed relative to the standard of care patients would normally receive. This shift could help streamline consent processes and reduce barriers to conducting pragmatic studies using RWD.

The authors conclude that realizing the full potential of RWD will depend on coordinated action across policy, technology, and regulatory domains. As they note:

“RWD have enormous potential to help researchers and policymakers develop better ways to prevent and cure disease, to enhance quality of care, and to empower patients to improve their health.”