Do I need a problem or diagnosis 'priority' subset for clinical data entry?
Collection: Implementation dilemmas
Summary: No, the best practice recommendation is to avoid depending on priority subsets for user data entry. Concept selection should ideally be restricted only by domain, and all the SNOMED concepts within the acceptable range for the field should be available to the clinical users. This approach provides greater detail and benefits from frequent updates in SNOMED content.
Implementation dilemma
One of the first use cases in clinical implementations of SNOMED CT is to record clinical problems, diagnoses, or reasons for encounters. When defining the rules for which SNOMED concepts can be used in these use cases, some implementers face the question of whether it is better to define wide terminology bindings, including all SNOMED options for these domains, or to create a "priority subset", a smaller selection of relevant concepts according to some criteria or previous experience. The benefits pursued by creating a subset may include a simpler integration of smaller lists in legacy systems or facilitating user entry by offering a smaller set of terms.
Advice
As a general approach, it is preferable to avoid depending on manually enumerating a subset for user data entry. Concept selection should ideally be restricted only by domain, and all the SNOMED concepts within the acceptable range for the field should be available to the clinical users. The Expressions Constraint Language (ECL)1 provides functions for implementers to restrict data entry domains to relevant content.
Restricting the list of relevant concepts by frequency of use or priority is not recommended, as it limits clinical users' expressivity and prevents them from taking advantage of the constant content updates in the terminology. Also, current trends like precision medicine and genomics require a high level of detail in data capture, recording very specific clinical assertions and rare diseases.
Instead, it is recommended that best practice search features applied by the systems support rapid searching across all available concepts within a predefined domain. Terminology servers and search engines can use frequency of use or priority in their sorting algorithms to boost the visibility of certain concepts while maintaining full access to the complete set of concepts relevant to the field.
There are some situations where SNOMED needs to be incorporated into an application that is not compatible with the complete SNOMED CT Edition content. The application may only support importing code systems as plain lists of concepts with a code and a term and uses limited search capabilities. In these situations, extracting a subset of SNOMED CT as a plain list is acceptable. However, using a subset restricts the expressivity of clinical users when selecting SNOMED concepts; therefore, in the long term, it is advisable to upgrade the application to support FHIR terminology services or any other technologies to provide access to the complete SNOMED CT content. Some commonly used sources for frequency-based concept selections are usually derived from the original work in the NLM CORE Subset in the US, resulting from the contribution of frequency of use in seven big healthcare institutions2. This resource was used for the initial selection of concepts for the SNOMED IPS Refset and the IPS Terminology3; these are also relevant sources of subset selections.
Tips for best results implementing data entry with SNOMED CT without arbitrary subset restrictions
Focus on creating the best possible terminology bindings: use value sets or ECL (the Expressions Constraint Language)1 to ensure that only acceptable concepts are available for data entry.
Use a terminology server for out-of-the-box best-practice text-matching algorithms or implement these algorithms locally4
Train clinical users on how the multiple prefixes, no order, string matching technique works.
If priority or frequency of use information is available, consider implementing boosting techniques to raise the visibility of the result.
Only use extracted lists of concepts as a temporary solution in legacy systems.
Restrictions for mapping purposes are acceptable; consider including all descendants of mapped concepts to enhance expressivity.
References
Expression Constraints Language Guide: http://snomed.org/ecl
NLM Core Subset: https://www.nlm.nih.gov/research/umls/Snomed/core_subset.html
International Patient Summary Terminology User Guide: http://snomed.org/ipstug
SNOMED Terminology Services: https://www.implementation.snomed.org/terminology-services
Learn More
To learn more, please reach out to [email protected].
Last updated