Med school platform puts knowledge in computable format
Health information technology is enabling healthcare organizations to analyze the data they generate during the process of taking care of patients to create new “local” evidence to improve outcomes.
Combining this locally generated evidence with evidence from peer-reviewed medical literature is critical to creating a learning health system that can continuously study and improve itself while rapidly integrating this knowledge into best practice models, moving research out of journals and into the clinical environment.
So says Charles Friedman, chair of the University of Michigan Medical School’s Department of Learning Health Sciences and the Josiah Macy Jr. Professor of Medical Education.
“We need to begin to think about what evidence means in an age where academic and large health centers, in particular, are routinely collecting data in electronic form as a byproduct of their patient care,” contends Friedman. “The idea that any practitioner might be able to keep up with the latest knowledge in his or her field by reading journals is simply not realistic. Too much of it is changing too quickly.”
“The value of big data is to generate big knowledge,” he observes. “The power of big data is to provide better models. If all those models do is sit in journal articles, no one’s going to be any healthier.”
Writing in an article in the December issue of the Journal of General Internal Medicine, Friedman and his co-authors note that while evidence has traditionally been represented in human readable form—in text, tables, and figures—and published in books and journals, knowledge in this format cannot support its rapid and routine translation into practice. They contend that evidence of all types must be available in standardized, computable formats to be leveraged for the curation, representation, dissemination and application of knowledge.
Towards that end, the University of Michigan Medical School’s Department of Learning Health Sciences has created an open computer platform called the Knowledge Grid, which is designed to make medical best practice knowledge computable and readily available for widespread use in digital libraries that generate patient-specific advice.
“A lot of scientific studies result in some kind of model—an equation, a guideline, a statistical relationship, or an algorithm—all of these kinds of models can be expressed as computer code that can automatically generate advice about a specific patient,” according to Friedman, who notes that when both “local” models and published models are available in computable forms, it is suddenly possible to generate advice that reflects both kinds of sources.
“To achieve optimal use of knowledge in support of care decisions and choices, the main theme of the paper is how to combine the two kinds of knowledge,” says Friedman. “We need both but we need a way to put them together.”
As part of a preliminary initiative, Friedman reveals that a Knowledge Grid activator—a system for deploying Knowledge Objects in a computable form to process health data—has been installed in the Michigan Health Information Network, which facilitates information sharing among the state’s HIEs and providers, so that as clinical data moves through MiHIN the network has the capability to apply Knowledge Objects to the data and draw conclusions about individual patients.
“With all this information and data available to us, the field of biomedicine and particularly healthcare is in a new era,” adds Friedman. “There are knowns that we don’t know because we haven’t done the analyses yet.”
At the same time, he emphasizes that the privacy of patients is paramount and that “there are a whole set of ethical, legal, and social issues that bear on these processes that must be taken into account—this is not just a technical exercise.”