The Case for Building a Clinical Information Commons

What if no piece of data or biospecimen that entered a given hospital system went to waste? Instead, the authors of a recent JAMA article proposed, this stunning amount of medical information could enter a so-called “clinical information commons” (CIC). From there, they said, the data could evolve into actionable knowledge—the kind that fuels better diagnoses.

Kenneth D. Mandl, MD, MPH, and Florence T. Bourgeois, MD, MPH—researchers in the Computational Health Informatics Program at Boston Children’s Hospital—called for the establishment of this sort of sweeping real-world database by way of a “new compact between patients and the health system.” Though complex, the arrangement could change diagnoses in a “fundamental” way, incorporating the expanding body of genetic information into the process, the authors wrote. CICs from different institutions could also link up, further fostering valuable insights.

“Rather than making preventable diagnostic errors in the future,” they concluded, “clinicians and researchers should engage in a compact with patients to create CICs with maximum representation across the population so the full ‘normal curve’ can underpin digitally driven genetic diagnosis.”

Mandl and Bourgeois said the need is clear. Now, the pathogenicity of a genetic variant tends to come from small cohort studies, limiting comparisons. Monogenic variants connected to hypertrophic cardiomyopathy, for example, represent normal variants among people with African ancestry, the authors noted. And the problem appears to be more general than that, they said.

Much of the existing data on genotype-phenotype relationships came from researchers, the authors noted. The 3 main types of databases rely on research and volunteers, rather than a clinical framework, meaning that populations are self-selected. “The bias inherent in these populations may distort the accuracy of data-driven genomic diagnosis,” Mandl and Bourgeois wrote.

In short, healthcare providers need data gathered in the real world to advance genomic diagnoses.

The CIC would provide exactly that to the health system at large and individual providers, who in turn would use the information to diagnose and treat patients, the authors said.

But what would that require? First, patients would need written notification of the CIC’s existence. Consent-to-treat or privacy practices documents could do the job, making clear that “residual volumes from blood and tissue samples may be stored, linked to [electronic health records] EHR data, and used in a CIC for a patient’s own medical care or the care of others,” the researchers noted.

From there, hospitals would need to set up informatics and lab processes for the influx of samples and necessary additional testing. Ideally, this system would connect with EHRs, churning out population-based insights as a physician considers a diagnosis, Mandl and Bourgeois wrote.

Of course, hospitals would need to combine resources to ensure patient privacy protections, they added. Even so, a “robust informatics and policy basis” for this sort of endeavor already exists, they said.

But larger institutions could tackle such a challenge, despite the sociological, financial, and technical barriers, the authors wrote.

“Importantly, the institutional CIC would be reflective of the patient population treated at that institution,” Mandl and Bourgeois went on. “Even though an institution-specific CIC is useful, joining de-identified data from CICs across institutions would produce even greater benefit.”

The researchers warned that altering a bedrock component of medicine “warrants caution and careful assessment.”