Everything on FHIR: Toward a Frictionless Knowledge & Data Ecosystem in Healthcare

Matthew Burton, MD, Vice President Clinical Informatics, Apervita
Matthew Burton, MD
Vice President Clinical Informatics & Analytics, Apervita
Blackford Middleton, MD, MPH, MSc, Apervita
Blackford Middleton, MD, MPH, MSc
Chief Informatics & Innovation Officer, Apervita

In the 2009 ARRA (American Recovery and Reinvestment Act)/HITECH (Health Information and Technology for Economic and Clinical Health) Act, the policy intent was to implement EHRs broadly, ensure interoperability between them for data- sharing, and to use clinical decision support to transform care. In the nearly 12 years since that law was signed, the U.S. has fallen short of these goals. We have not achieved the value proposition that was expected (and demonstrated in select sites) to return on this investment with EHR adoption, interoperability, and informed and transformed care. The reasons are multiple, but it’s clear EHRs suffer usability challenges for some, do not seamlessly exchange data, and it’s exceedingly difficult to get cognitive support to work right.

FHIR (Fast Healthcare Interoperability Resources) to the rescue! Tactfully avoiding a play on words, we have never been more excited than now about the potential for health IT to transform care for providers, healthcare delivery systems, payers, and patients alike. Healthcare certainly benefits from strong federal mandates and regulations from the ONC and CMS in the form of the Information Blocking and the Interoperability and Patient Access final rules. Specifically, CMS’ regulations regarding patient access will require creation of EHR APIs usable ‘without special effort’ to allow patients to receive their healthcare data, for payers and providers to exchange data with other payers and providers, and for patients to use their data as they see fit in applications of their choosing to manage their health and wellness. The ONC and CMS regulations also serve as an essential foundation to decrease the friction in data and knowledge exchange in healthcare opening up vast opportunities that will positively impact many lives.

HL7’s Key Role

These policies and the associated technologies are possible in large part due to a long-standing, broad, deep, and significant effort led by the HL7 International standards development organization (SDO), and the countless professionals from academe and the healthcare industry. These professionals have largely participated in voluntary capacities because of the dire need to establish the requisite standards through deliberate consensus-driven processes.

“To achieve truly ‘frictionless’ data and knowledge exchange across healthcare, we need not only think about the progressive layers of knowledge abstraction and implementation from data to wisdom; we need to also think about the tiers of implementation in our health IT systems: from data layer, to logic layer, to presentation layer.”

As we look back on our combined 55 years of applied clinical informatics work, we recall the Data, Information, Knowledge, Wisdom Pyramid (DIKW) learned in early Fellowship training that describes the desired transformation of data into knowledge. In this pyramid, data is compiled, information is derived, knowledge is inferred, and wisdom is internalized and acted upon. In a way, we can see the decades of standards evolution and expansion crawling up this DIKW mountain to get us to true interoperability, where data is fully and seamlessly interoperable, information and knowledge is sharable and computable and wisdom is internalized and acted upon by all decision-makers, with modern compute capabilities being leveraged to optimize value throughout the healthcare ecosystem.

To achieve truly ‘frictionless’ data and knowledge exchange across healthcare, we need not only think about the progressive layers of knowledge abstraction and implementation from data to wisdom; we need to also think about the tiers of implementation in our health IT systems: from data layer, to logic layer, to presentation layer (creating a grid of “L by T”; see FIG. 58 here.[i] Levels of Representation by Tiers of Functionality with Examples (Rhodes,B, Database Consulting Group, 2019).). The implementation challenge is non-trivial in a heterogeneous IT landscape with multiple EHRs, multiple unique implementation instances, diverse local practice patterns, and an exploding array of patient-facing mobile health (mHealth) apps. Figure 1 depicts the traditional separation of concerns in the implementation tiers (vertical axis), and suggests a progressive abstraction of knowledge artifacts in the levels of knowledge abstraction (horizontal axis). This structure supports the componentization of parts for knowledge artifacts, as well as re-usability. Standards do not yet exist for each cell in the L by T grid, but they do exist for many of these cells or at least touch on portions of them. For example, at the presentation layer, the same logic may be expressed in a SMART-on-FHIR container, a suggestion “Card” with CDSHooks, or as an iFrame delivered via web-service. Similarly, along the horizontal axis an expression written to define a numerator and denominator expression for a measure may be re-used in a CDS ECA rule.

Toward a Learning Health System at Scale

To truly enable the Learning Health System, we envision a loosely coupled array of standards within and across these cells that promotes both data access and reuse, as well as implementation of computable knowledge artifacts that can be ‘played’ or rendered in a variety of presentation mechanisms—to both healthcare providers, and to patients. The HL7 community has enabled just such a vision. When most think of HL7 FHIR, they focus on standards for APIs and data resources, which the FHIR community has put much effort into maturing and the broader ecosystem has adopted broadly. However, the HL7 standards stack and FHIR in particular was fundamentally architected and has further evolved to enable much of the DIKW pyramid across the L by T grid. For instance, FHIR Definitional resources (e.g. Structure Definition, Plan Definition, Measure), based on the HL7 Knowledge Artifact Specification, together with the Clinical Quality Language (CQL) are used to express computable, shareable knowledge across the stack from explicit data semantics to some of the highest orders of knowledge in the entire domain entirely through the application of the FHIR foundational architecture (using the HL7 Implementation Guide (IG) mechanism). Examples of this higher order domain knowledge include the Evidence-Based Medicine (EBM-on-FHIR IG) and expert curated, evidence-based Clinical Practice Guidelines (CPG-on-FHIR IG). Many may not be aware of the FHIR architecture and development framework’s ability to compose and then derive numerous computable high-value knowledge artifacts specified to fundamental domain concepts.

The HL7 standards stack, and the development of the Clinical Practice Guideline on FHIR Implementation Guide (CPG-on-FHIR or CPG-IG), give us strong reason to hope for acceleration in the creation and delivery of computable practice guidelines to decision makers at the point of clinical decision making, at the point of opportunity to enable best practice care. Building upon the full array of Foundation Standards, Data Standards, Interoperability, and CQL, the CPG IG describes a standard that can not only describe the essential clinical pathway of a guideline, but many derivative knowledge artifacts, sharing the very same best practice logic as well. FIG. 01. here [ii] shows the interfacing of standards across the clinical quality improvement ecosystem. The full knowledge management lifecycle, along with the elaboration of dependent artifacts, can be considered together to support the entire lifecycle. In this manner, we can dramatically increase the reuse of knowledge artifact component parts such as terminologies, value sets, and expressions (CQL defines). This will help dramatically to reduce the knowledge engineering burden for CPGs, and reduce unwarranted variability in their implementation. It enables a closed-loop, feedback and feedforward characteristics of a truly Learning Health System.

This is a truly exciting time to be in healthcare informatics! We have seen the explosion of new knowledge arising from the care of COVID-19 patients, and ongoing in other areas, which has forced innovation and collaboration in ways unseen until now. Apervita is pleased to be at the forefront of this digital transformation of healthcare bringing the best of clinical informatics innovation and analytics to the practice of medicine.

Have a question or comment on this blog? Or a suggestion for future blog topics? Let us know!



[i] http://build.fhir.org/ig/HL7/cqf-recommendations/documentation-approach-08-levels-of-representation-by-tiers-of-functionality.html
[ii] http://build.fhir.org/ig/HL7/cqf-recommendations/documentation-approach-01-main-page.html


Blackford Middleton, MD, MPH, MSc
Chief Informatics & Innovation Officer, Apervita

Dr. Middleton is the Chief Informatics and Innovation Officer at Apervita. He is also an adjunct faculty member at the Stanford University Medical Center Clinical Excellence Research Center. Previously, he was a professor of Biomedical Informatics, and/or of Medicine, at Stanford, Harvard, and Vanderbilt Universities, and he held executive leadership roles at MedicaLogic/Medscape, Partners Healthcare System, and at Vanderbilt. His work is focused on clinical informatics—the applied science surrounding strategy, design, implementation, and evaluation of clinical information systems in complex environments.
Matthew Burton, MD
VP Clinical Informatics & Analytics, Apervita

Dr. Burton is Apervita’s Vice President of Clinical Informatics, leading innovation efforts and industry knowledge architecture activities with key Apervita partner institutions. He is also an Assistant Professor of Biomedical Informatics at Mayo Clinic and has served as the Lead Clinical Informatician for the Applied Clinical Informatics Program. He has taught graduate level courses, most recently as an Associate Research Professor of Biomedical Informatics at Arizona State University.

Recent Posts