Learn the difference between intensional and extensional value sets, the challenges of managing them, and strategies to maintain accurate, standardized healthcare data.
In today’s data-driven healthcare landscape, value sets play a critical role in ensuring accurate insights from patient data. Whether used in clinical decision support, quality reporting, prior authorization, utilization management, or any healthcare analytics, well-maintained value sets are foundational to the accuracy, efficacy, and success of data-driven initiatives.
Value sets are a simple concept, but challenging to curate and maintain for three main reasons:
- Difficulty authoring with required precision for specific use cases
- Maintainability when underlying codes are ever-changing
- Governance to ensure accuracy, applicability, auditability, and re-use
I’ll dig into each of those, discussing the critical role of intensional value sets, and highlight the essential features of a value set management platform to manage use case-specific needs.
Before we jump in, what are value sets?
At Wolters Kluwer, we always say, “healthcare runs on codes.” And it’s 100% true, every system throughout the healthcare ecosystem runs and relies on codes, and nearly everything is coded into the Electronic Health Record (EHR). Interoperability regulation, currently defined in the USCDI, ensures that patient data are coded consistently across EHRs to enable interoperability across the ecosystem, and, of course, claims are coded to billing standards so that providers can get paid.
But, there’s an important categorization of codes that’s equally as essential as the codes themselves. The categorizations are called “Value Sets”, also referred to as “Code Groups,” “Code Sets,” “Epic Groupers”, or OMOP’s “Concept Set” (Leave it to the terminology standards to introduce synonyms for basic objects in the models.). These value sets organize codes that have a similar meaning into groups, so each time you need to identify patients with a certain characteristic, you don’t have to search for each individual code.
An example of value set management in clinical decision support
A classic clinical decision support rule is ‘if patient has diabetes, then ensure an A1C test result less than 6 months old’. Variations of this same rule exist for quality reporting, analytics, and population health. Across each, without value sets, this rule would read:
‘if patient has codes in [E11.8, E11.9, E11.40, E11.41, E11.42, E11.43, E11.44, E11.49, E11.51, E11.52, E11.59, E11.65, E11.69, or …], then make sure they have a result for codes in [71875-9, 4548-4, 17855-8, 4549-2, 17856-6, 62388-4, 96595-4, or …] less than 6 months old.
It’s not very readable, but worse, there is a high potential for the code to be incomplete, inaccurate, or out of date.
Value set management tools solve this challenge by centralizing the maintenance of the sets of codes that mean ‘diabetes’ or ‘A1C Test’ and making those definitions reusable and easy to maintain so you can simply have a rule that reads:
‘if patient has ‘diabetes’ then make sure they have a result for ‘A1C’ less than 6 months old.
This enables your informaticists to focus on building the right rules, or your data scientists to create the right prompts to build the AI-agents to deliver:
- Accurate cohort identification for population health or care management
- Trustworthy insights from real-world evidence (RWE) and real-world data (RWD)
- Digitally computable and precise prior authorization rules for utilization management
How intensional value sets help achieve precision
It’s not rocket science to find the ICD-10 diabetes codes. With a little knowledge or an AI-prompt, you can get a list of diabetes codes that will be reasonably complete and accurate.
But, what if you only want Type II Diabetes, or Type II Diabetes codes that are currently valid and billable, or Type II Diabetes codes (SNOMED and ICD-10) that have ever been billable. What if you want to infer the patient is diabetic based on the medications they are taking? Or, differentiate between patients taking oral vs. injectable semaglutide. Or what if you want to consider the problem list, coded in SNOMED rather than ICD-10?
That’s where the precision gets tough and where authoring with intention can help. The opposite of intensional, which takes the intension for the group and converts it to a rule, is extensional, which is just a list of codes. If you define with intention, you can be confident you are getting the desired codes. If you just define the list of codes, you may not have what you want.
