#clinithink

CDI Technology-Enabled Services and Solutions

Supercharging clinical trial patient recruitment with Clinithink's AI solution

Get the Best NLP Api - Clinithink

Using Clinithink for Clinical Decision Support

Written by Dr. Peter Johnson, Co-Founder/Chief Terminology Director from Clinithink

Statisticians have been collecting and classifying causes of disease for hundreds of years. In fact, the well-known ICD system started as a classification of causes of death in 1855. Since then, the focus has shifted to include classifying all aspects of care for many different purposes – from billing to early disease warnings, epidemiology, health care planning and point-of-care decision support.

The history and prevalence of classifications in healthcare gives rise to the importance of understanding the fundamental difference between a classification and a terminology:

A classification scheme could be thought of as a collection of buckets into which a care provider throws a particular concept or record. And since there can only be one bucket into which a concept fits, the process of labeling the buckets often leads to catch-all terms like: ‘Disease X, unspecified’ or ‘Y, not elsewhere classified’. As a result, accurately classifying records is rightly seen by most care providers as a separate process from record creation and is typically carried out by specially trained coders who know how to apply the process.

Unlike a classification system, a terminology allows the user to specify precisely what they want to record. Specifically, a terminology doesn’t have any ‘not elsewhere classified’ bucket terms, but is designed to have the terms that a user needs to record what actually happened.

The role of a terminology is thus different to that of a classification: the terminology  allows one to accurately capture what happened, while a classification uses a separate process of abstraction (or choosing the right bucket) to categorize a particular health care event.

While early terminologies aimed to provide a term for everything that a care provider would want to record, that idea quickly became untenable. Even though care providers may only use a vocabulary of approximately 30 000 words, they can combine those words in millions of ways, making it impossible to enumerate all the possible combinations and meanings.

SNOMED Clinical Terminology (SNOMED CT) on the other hand, is a terminology  that allows multiple terms to be used together within a set of rules which a computer can process. In this way it becomes possible to create an infinite number of coded terms, also called ‘post-coordinated terms’.

The result is that SNOMED CT is an extremely expressive terminology that is far more capable of recording precisely what a physician wants to say, rather than asking them to potentially distort the meaning of their clinical notes by fitting them into one of the ‘bucket terms’ of ICD. Even ICD-10CM limits the options to only 70,000 diagnoses and a similar amount of choices for procedures, whereas SNOMED CT provides an almost infinite amount of choices by use of post-coordination.

Although capturing clinical narrative using a suitably advanced terminology allows for an accurate representation of a care episode, that episode still needs to be classified for various purposes. Many would logically assume that this would then raise the same limitations with classification discussed earlier but what makes SNOMED CT different to other terminologies is the classification process can be automated by intelligent IT systems. SNOMED CT is internally structured in such a way that the computer can do the generalizations required to categorize a care episode into whatever classification scheme is required for output.

The beauty of using post-coordinated SNOMED CT as the primary encoding is that everything else can be derived from that – most classification, including ICD-9&10 can then be done in a semi-automated fashion. However, that is only the start – once one has richly encoded records in a machine understandable form, many other analytics become possible, decision support can be added, and communications become less  ambiguous and safer. Unfortunately if one uses a classification as the primary encoding, the result is a more generalized categorization and tends to be only useful for the task it was created for, such as payment. One can’t then derive a rich, terminological encoding for other purposes.

The post-coordinated approach is one of the bases on which Clinithink’s technology  is built. The CLiX engine is not restricted to the limitations of classification systems or terminologies and was designed to maximize the power of computer assisted coding solutions and increase coding productivity. CLiX has the flexibility and cross mapping capabilities to accurately index and structure free-text clinical narrative (without losing the detail that defines a care encounter), into structured medical data that can be used to conduct analyses for many different purposes. By harnessing the power of SNOMED CT as part of the CLiX solution, organizations gain the insight necessary to drive positive change in the healthcare delivery process and improve revenue and outcomes.

About Dr. Peter Johnson: