Value sets and the problem of redundancy in value set repositories.

OBJECTIVE

Crafting high-quality value sets is time-consuming and requires a range of clinical, terminological, and informatics expertise. Despite widespread agreement on the importance of reusing value sets, value set repositories suffer from clutter and redundancy, greatly complicating efforts at reuse. When users encounter multiple value sets with the same name or ostensibly representing the same clinical condition, it can be difficult to choose amongst them or determine if any differences among them are due to error or intentional decision.

METHODS

This paper offers a view of value set development and reuse based on a field study of researchers and informaticists. The results emerge from an analysis of relevant literature, reflective practice, and the field research data.

RESULTS

Qualitative analysis of our study data, the relevant literature, and our own professional experience led us to three dichotomous concepts that frame an understanding of diverse practices and perspectives surrounding value set development: Permissible values versus analytic value sets;Prescriptive versus descriptive approaches to controlled medical vocabulary use; andSemantic and empirical types of value set development and evaluation practices and the data they rely on.This three-fold framework opens up the redundancy problem, explaining why multiple value sets may or may not be needed and advancing academic understanding of value set development.

CONCLUSION

In order for value set repositories to become more rather than less useful over time, software must channel user efforts into either improving existing value sets or making new ones only when absolutely necessary. This would require major, innovative changes to value set repository platforms. We believe the most direct path to giving value set developers the ability to leverage prior work is by encouraging them to compare existing value sets using advanced interfaces like VS-Hub, and by collecting and using metadata about code inclusion and exclusion decisions during the authoring process.