In a previous post, we dove into some of the nitty-gritty on "value sets." Recall that these are lists of standard codes (e.g., ICD-10-CM, CPT) that define a specific clinical concept. In this post, I want to provide a review of the evolution of Azara's value set management process and highlight some of today's issues.
For the first several years at Azara, our value set content integrated with our code; meaning that only engineering could update value set content, it had to be documented through our internal ticketing system, and the changes went out with significant product releases (about every 8-10 weeks). The transparency and consistency were excellent, but this process became a time sink for engineering, and the wait for a product release was often too long to be acceptable to clients.
As Azara grows, our need to formalize value set management has increased. We now get support tickets and change requests regarding value set content on a daily basis. We've hired staff with clinical credentials that are deeply involved in helping define value set content. To remove engineering as a bottleneck in this process, we've developed what we call "terminology services" in our internal tools. Now our clinical data experts can view and edit value set content without any engineering involvement.
In early 2017, we licensed a commercial terminology service to help us manage value sets and related terminology data. By combining this service with Azara’s homegrown tools and intellectual property, we have developed a value set management process that allows us full transparency and versioning of value set content.
Just in time too – as we currently have almost 2,000 value sets that define everything from diabetes diagnoses to statin medications to qualifying encounters for different quality measures. The average value set has hundreds of codes, so that's a lot of content to manage! Moreover, it indeed is active management as the business definition of these value sets is continuously coming under fire. In some cases, we control the content internally; in other cases, we have to interface with the ONC, NCQA or another organization acting as the authority for the value set.
The underlying code systems change too; if a new statin RxNorm code is introduced, we need to make sure it gets in our value set. That presents an interesting problem around keeping value set content accurate. As I see it there are two ways to tackle this problem:
- Constrain the code systems you use to define a certain kind of clinical concept. Just use RxNorm for medications and use ICD-10-CM for diagnoses. That way you only have to work with one code system when defining a value set, so the impact of volatility in the underlying code systems is limited.
- Be pro-active about value set management. As Azara has grown, we have hired more clinical and data experts to help manage this content as opposed to reacting to it through engineering.
While we certainly have been busy improving our software and processes here at Azara, the industry at large has also matured around value set management. Within the last few years, the National Library of Medicine launched the Value Set Authority Center, and CMS has published a blueprint for developing quality measures that provide guidance on value set management. The ONC manages a ticket system where you can submit change requests regarding value set content (Azara often forwards your requests to this system).
In the early days of Azara, value set content was simple and straightforward - diabetes is 250.**, asthma is 493.** (remember ICD-9?). Both Azara and our clients have come a long way since DRVS was first launched in 2008. Today, we have thousands of value sets and hundreds of thousands of codes, all contributing to hundreds of measures. This has pushed us to scale and evolve both our software systems as well as our internal processes and is just another part of the journey as we all try to get better at the complicated process of clinical quality measurement.