As we’ve discussed before, clinical research really comes down to a numbers game. But getting enough patients for a study, particularly for rare diseases, can be a daunting challenge. Similarly, it can be hard to tell whether observations made in just two or three patients, say a possible new medication interaction or a new diagnostic presentation, are part of a trend – one that’s worth grant money to study.
The Shared Research Information Network (or SHRINE) could help solve these vexing problems by mining data across hospitals. Developed through Harvard Catalyst by a team led by Isaac “Zak” Kohane, director of the Children’s Hospital Informatics Program, SHRINE links the clinical databases of participating Harvard-affiliated hospitals – currently Children’s, Beth Israel Deaconess Medical Center, Brigham and Women’s Hospital, Dana-Farber Cancer Institute and Massachusetts General Hospital – letting researchers see the numbers of patients seen at those hospitals who meet selected criteria.
Why is this important? Because by gathering aggregate patient numbers, SHRINE essentially turns clinical data into population data. With every query, the system sifts through the records of about 6 million patients, searching demographics, diagnoses, medications and lab results. It then presents the aggregate number of matching patient records, as well as the number per institution. This population-level perspective can help researchers see if those limited observations in the clinic are part of a larger phenomenon meriting study, or get an initial sense of whether, between the hospitals in the system, there might be enough patients to conduct a therapeutic trial.
So, to use some completely fabricated examples, a diabetes researcher could see whether there really is a rising trend of boys with type 1 diabetes under the age of 10 experiencing headaches after taking a new medication. Or a dermatologist could find out how many young adults using a topical treatment for eczema are suffering allergic reactions. Those numbers can be enough for an investigator to decide whether to write a grant or design a study to delve into the question more deeply.
“We can answer questions in seconds that would otherwise take years,” Kohane says in a recent interview in HMS Focus. “Or that you could never even ask.” He tells the story of a multicenter study of links between diabetes and a certain combination of statins and antidepressants, bringing together researchers from three different medical centers from across the US. “That took months and a big Rolodex,” he notes. “SHRINE takes seconds and a good idea.”
“We can answer questions in seconds that would otherwise take years,” Kohane says. “Or that you could never even ask.”
“Wait!” you shout. “What about patient privacy? What about HIPAA?”
SHRINE addresses privacy concerns by scrubbing away any identifiable information and providing only aggregate numbers. Also, SHRINE is what is called a federated system: Each participating hospital maintains complete control over the information in the databases that SHRINE peeks into. No actual data is ever transferred past the hospitals’ respective firewalls, only counts of patients. (Actually accessing patient data or identifying possible study recruits still requires applying for IRB approval at each relevant hospital.)
At the heart of SHRINE is a technology called Informatics for Integrating Biology and the Bedside (i2b2). This NIH-funded informatics framework – developed under Kohane’s watch – provides a common language and standards for securely exchanging clinical information across hospitals, universities and other institutions specifically for research, chiefly of the genomic variety.
i2b2 can also help bridge between clinical and research information systems within an institution. For instance, Children’s Gene Partnership will use i2b2 to exchange data between the partnership’s data warehouse, this hospital’s clinical databases, and Indivo (Children’s personally controlled heath record system).
But back to SHRINE, and the numbers. Say you run a query and the results are inconclusive: there are some patients at the five Harvard hospitals who meet the criteria of your search, but their numbers are too small to confirm a trend or justify next steps. The SHRINE team is piloting an expanded version of the network that adds six more medical and clinical research centers in California, Michigan, North Carolina, Ohio, Texas and Washington State. That expanded network will have access to millions more patient records, giving population research on nearly any question imaginable an even bigger boost…and helping improve patient care in the process.