Expanding the Conversation on Interoperability & the Products that Power It

BD: I am formally trained in medical informatics with a Master’s degree from the University of Utah Biomedical Informatics department. As a grad student, I was introduced to many of the challenges we still face today: interoperability, data modeling, ontologies, natural language processing, clinical decision support, analytics, automation, etc. My career has never strayed from health IT and medical informatics. I have worked with large healthcare organizations, such as the VA (Veterans Affairs) in the US and Queensland Health in Australia, to solve these kinds of problems with innovative solutions anchored in medical informatics. Prior to joining Smile, I was also a Smile customer. I previously led a team building a healthcare interoperability platform called HealthConcourse: an end-to-end interop solution that ingested and standardized all customer data to FHIR (using Smile’s Clinical Data Repository) and analyzed that data to provide actionable insights (using Smile’s Clinical Reasoning tools). 

So, in many ways, my prior training and career path have been preparing me for exactly this role. Through these experiences, I developed my skills in architecture and product development — skills I am excited to bring to my new role at Smile. Smile presents an opportunity for me to apply what I have learned and make a big difference in healthcare delivery. 

BD: When Smile started six short years ago, we were well positioned to ride the wave of excitement around the emerging FHIR standard. We had the right people working on the right products at the right time. But now, FHIR is table stakes, in the US at least. It is commonplace and accepted, and as such, it is no longer good enough to have a product strategy based solely on a FHIR management platform. 

I am excited to join Smile as a CPO to deliver an expanded product portfolio that fills out other aspects of our envisioned health data fabric (HDF). The way I see it, the opportunity in front of us is to provide solutions that address more of the interoperability landscape (we dive more into this in the next question). This includes products that: augment our ability to ingest customer data (data that is rarely in FHIR to begin with), optimize data for use by downstream applications, help compute and reason upon our customer’s data (including AI/ML), package and deliver our data in novel ways, and use the data to automate clinical workflows. 

We are beginning to realize the goal of using data to enable a learning health system with deep insights and automation. This is essential now because research shows that it takes 17 years to get clinical best practices from research into implementation. Now that we have robust solutions to get data to a computable state, we can greatly shorten this timeline by building the evidence-based clinical reasoning tools directly into our product suite. In other words, we have the opportunity to expand from “getting data to a computable state” to being able to learn from the data. I am excited to build a product strategy that positions us to deliver greater value across more of the interoperability landscape, and enable intelligent use of our customer’s data.

BD: Interoperability, and the solutions that enable it, are often viewed too narrowly, so let’s broaden the definition. It is simple to think that you either have interoperability or you don’t. In reality, interoperability exists along a continuum. Some use cases are satisfied simply by aggregating data into one source or through one data access mechanism. Other use cases require data to be standardized to some canonical schema. If you want truly computable data—data that can be used for decision support, advanced analytics and automation—then you need semantic interoperability. 

By definition, semantic interoperability requires that data satisfy two principal requirements. First, it must conform to a rich and expressive data model to ensure a common record syntax and structure. Second, the contents of the record must also link to industry standard terminologies so that the embedded clinical concepts have an unambiguous meaning and are understood with the relationships defined by a rich medical ontology. By providing solutions that achieve semantic interoperability, we deliver the data quality our customers expect.

Taking this one step further, data interoperability needs to be considered within its usage context to ensure it is optimized for downstream use. By usage context, I am referring to the business logic that executes rules against the data (e.g., clinical decision support) and the workflow that describes the activities and processes where the data is used. The artifacts that define the computation logic or knowledge, as well as the artifacts that define the workflow, all need to be well defined, computable, and ideally, interoperable. Thus, “Knowledge Interoperability” and “Process Interoperability” are necessary overlays on top of semantic data interoperability to realize the complete vision for interoperability. 

As we continue to grow, our products need to consider this broader perspective on what interoperability is and offer solutions that span the continuum.

BD: Data fragmentation resulting from siloed healthcare indeed contributes to a lack of interoperability. However, siloed and fragmented data are not synonymous concepts with interoperability. You can have siloed data and still have interoperability if the data silos conform to a common data model and terminology at either the database schema or API level. Conversely, you can aggregate data to eliminate silos but still not have interoperability. This is commonly seen with data lakes, whereby data is ingested and co-located without standardizing the data to a canonical state. To address this, we need to advocate the use of rigorous, computable standards and provide products that help our customers standardize their data, regardless of how the data is physically stored.

BD: For too long, electronic medical record (EMR) companies have held their customers hostage to predatory business practices enabled through proprietary data and knowledge models. By building their products on non-standard data models, EMR companies position themselves as the only ones that can make changes to their products to address an ever-changing set of data requirements. The same is true for the decision support rules and logic that are too often hard-coded into proprietary systems. This is the vendor lock-in paradigm.

Data and knowledge are our customers’ true enterprise assets and as such, should not be restricted by how they are implemented within health applications. It is the applications that are the commodities and should be able to be replaced without throwing out the data and knowledge. To achieve this, we need to flip the paradigm and democratize the data and knowledge assets.

We need to move the industry to recognize and enforce a healthy “separation of concerns” whereby data and knowledge assets are modeled and authored using industry standards, and EMRs and similar applications provide their business functions by using these standards-based assets without forcing them into proprietary systems.

BD: Anyone who has attended the HL7^® working group meetings over the years knows that much of the value comes from face-to-face interactions and the cookies that are served between Q3 and Q4 every day. During the peak of COVID, the working group meetings were conducted virtually. To engage participants virtually, the creative people at HL7 introduced a few fun and interactive ideas. One of these was “HL7’s Got Talent”. They invited all willing participants to submit a one-minute video of themselves doing something related to FHIR. 

One of my interests and hobbies is slam poetry. For those who don’t know, slam poetry is a style of performance poetry where the poets recite a three minute poem and are judged by a panel of five judges. This allows for scoring and competition. Given this background, I was practiced at producing and reciting poems that were fast, engaging (hopefully), and somewhat edgy. 

I concluded my one-minute, FHIR-based slam poem with the following proclamation: “I know your system wasn’t designed to cross the line from your organization to mine, but these boundaries are killing me. I need you to acquire solutions that can span the wire. I know what you require. You need FHIR.” 

I won second place and was awarded with free entry to the next HL7 WGM!

Follow Bo on LinkedIn

Join the movement to #BetterGlobalHealth