In my healthcare IT career of more than 30 years, I have worked diligently on the interoperability of content, with a natural focus on precisely defined and coded structures for the exchange of information.
However, the computer scientist in me has always pondered: when will we be able to move to systems of cognitive understanding and the exchange of full narrative and so-called unstructured data? I believe that we are on the cusp of that time, and that there is good reason to place more and more emphasis on the narrative aspects of health care.
Whither the Code? A brief history of health IT coding
We have been at this coding business in health care for a very long time. The use of a coding system for healthcare IT dates to 1348, in Italy. The current ICD coding system can be traced back to the classification of mortality in the 1600s by John Gaunt. Did you know that in 1655 there were 8,297 deaths in London, of which bubonic plague alone claimed 7,165? That same year, 3 deaths were coded as frightened to death (yes, fright got its own code that year)?
In 1948 there was a revolutionary paper in the Bell Systems Journal by Claude Shannon: “A Mathematical Theory of Communication”. The original discourse was on communications as applied to signal processing, but the establishment of Information Theory, deriving from Shannon’s definition of information entropy, has become the foundation of established works in many fields.
Challenge: Preserving the content’s meaning with context
How does this relate to our consideration of healthcare IT and data interoperability? Quite simply, Shannon’s theory tells us that every time we encode content we lose information (there is a loss of entropy). Not only is this an inherent issue with encoding, but we know that in health care the meaning we seek to preserve and share with others involves the context of the data. It is also very likely that there is the use of professional slang, sometimes terms of negation, co-references to other data and other semantic structures not captured when data are encoded.
Consider the content “allergic to peanuts since childhood.” In today’s world we can encode this using a code system to convey this condition as relates to peanuts, but we are significantly challenged to encode the contextual nature of the condition - “since childhood.” But let’s place the full content within a larger narrative, such as a reason for visit or a summarization of events leading up to some episodic care. We clearly see where the use of coding, although interoperable at best (not all systems today can or do process inbound discrete data), omits some of the most useful information.
Are we ready, after almost 800 years of coding healthcare content, for systems that handle, understand and use the full power of narrative, graphical and other so-called unstructured data? There were several presentations at IHE World in 2016 on this topic, and at one of the keynote addresses at HIMSS 17, we heard from Ginni Rometty that we are arriving at the age of cognition.
Perhaps it really is time to refocus our efforts and bring the narrative of our healthcare data back to the forefront.