Sometimes a document seems so impenetrable, so dense and so dry that it doesn’t seem worth the effort to read it. But strangely enough, sometimes plowing ahead is worth the trouble, and this is one of those times. If you dig through the layers of a recent letter AMIA wrote to the National Library of Medicine, you’ll pick up some interesting ideas about the nature of data science and its role in managing healthcare data.
AMIA wrote the letter in response to an NLM Request for Information on next-generation data science challenges in health and biomedicine. What’s interesting about this request is the reality that data science overall is still evolving, which means that defining its applications in health and biomedicine is a particularly challenging exercise. Still, why wait ’till the answers are obvious. Where’s the fun in that?
In its RFI, the National Library asked for input on new data science research initiatives that could address key problems in health and biomedicine, along with suggestions on which data science research directions in health and biomedicine it should pursue.
The health IT trade group argued that the best thing it could do was focus on the basic science of data standards, including “development of granular data specifications to enable a ‘periodic table of elements’ approach to biomedical data standards.”
The group rightly notes that at present, data elements may be grouped together in different ways depending on the purpose for their use, which can rob the data of much-needed context. “Such an approach would enable the combination and substitution of discrete data elements for specific use cases, such as quality measures, and facilitate data re-use more readily than is the case today,” AMIA notes.
The trade group also called for the development of metadata in health and biomedicine which, as it notes, could make data more traceable and simplify the process of tracking its provenance and accuracy. In the world where big data is king, anything that makes data elements knowable is a good thing. Again, context counts.
Over time, AMIA suggests, the industry should look for ways to categorize and manage data in finer-grained ways, including:
- Further development of disease-specific data elements (probably, extensions for existing standards) along with standards for nursing and other clinical areas
- Figuring out how to tell when differences in coding are appropriate and when they are inappropriate, as well as establishing baseline standards harmonizing coding variations
- Developing standard imaging recognition algorithms for key image-based health disciplines like radiology and pathology
The letter also suggests some interesting ideas for the future of health and biomedical data science, notably finding ways to use data from consumer sources such as wearable device, mHealth applications and social media tools, along with industrial-level technologies such as genomics, Internet of Things and geospatial sensors.
All told, to drag out an old cliché, it’s certainly an interesting time to be in healthcare data science. Those who master its arcane arts will shape the future of medicine along with helping others decipher medical data. Hey, in any business it never hurts to be the idea-driver and the oracle.