4 years ago when I took the Healthcare IT Portfolio of UMSC, I learned the horror of integration in healthcare, to get various bits of clinical information about a patient from its different Silos to the consolidated Medical Record. When most vendors spoke of integration, they were purely speaking on the level of syntactic interoperability, where protocols and middleware allowed for the flow of badly coded HL7 messages from one system to another. This process was prone to problems of version control, the lost of dimensions and context to the data and is an expensive and complex process. The biggest proof that this is a broken practice is that Clinical analytics was often not possible from the aggregated data. Most importantly, the machines pushing the data around were agnostic to its semantic meaning or context and therefore this limited ability to unlock automation and analytics that normally follows other industries that digitize their data.
So when I explored the possibility of semantic interoperability on top of syntactic interoperability, I was pointed to data dictionaries and codified medical knowledge and told that the only solution was getting Clinicians and practitioners to abstract and code data manually so that it was machine understandable. So I went on a quest, to look at the use of AI and Natural Language Processing to solve this problem, and was pleasantly surprised that IBM has developed some solutions on these principles and capabilities. Finally we can leverage IT for Semantic Interoperability and unlock the value of Clinical Analytics.