As the healthcare industry moves towards increased sharing of patient data, the challenge has become how to share and store that data between systems without losing the true clinical meaning of the diagnosis. In a conversational way, a paragraph of words by the clinician can most effectively present the full meaning of the diagnosis. But the problem is paragraphs of words are not easily understood by machines. Applications exist today for Clinical Decision Support and Population Health Analysis, but these applications rely on coded data, not sentences.
I previously wrote a blog about the three CDA levels of interoperability, where I briefly discussed the topic of semantic interoperability. Semantic interoperability plays a key role in keeping the intent of the clinician clear while transposing the content to coded data.
So what is semantic interoperability? Wikipedia defines semantic interoperability as “the ability of computer systems to transmit data with unambiguous, shared meaning.”
To better understand semantic interoperability, we need to first compare the differences between syntax and semantic interoperability.
- Data flows from one application to another in the defined structured format.
- The pipes and hats of HL7 v2 are a perfect example of syntax.
- Data is communicated in a way that is understood exactly the same by the sender and receiver.
- Data models are used to establish the meaning, such as the RIM model in HL7 v3.
In the United States healthcare IT industry, HL7 v2 messages have been the staple of exchanging patient data among applications for many years. However, HL7 v3 has started to emerge with the use of CDA documents, which are part of the HL7 v3 standard. The use of CDA documents is mandated by Meaningful Use and is the primary reason for the standard’s emergence in popularity.
Why is this a good thing? Because HL7 v3 is based on a model, the RIM, and has the ability to better achieve semantic interoperability. Semantics are not explicit in HL7 v2.
Let’s take a simple example: A patient develops hives after taking penicillin. There are codes for hives and there are codes for taking penicillin, but how is a link drawn between the two acts using coded data? With HL7 v3, the RIM is in place and can model the relationship between two separation actions. Thus, there is a way to code the entire semantic meaning into data. But without a model, it would be much more difficult to represent the two actions and their link into the coded data. That is why having a model to represent the structure of the coded data is so important to maintaining the meaning of data.
As the healthcare industry progresses through the three stages of Meaningful Use, from having the EHR infrastructure in place to having the data in place to really impact the quality of care, the semantic interoperability of the data will be increasingly important. With more data shared in a way in which the true meaning is not lost, better care will ultimately result. And with CDA and HL7 v3, the architecture is in place so that the data does not lose its true meaning. ♦