Imagining A World On FHIR
By Deborah Borfitz
May 2, 2019 | Achieving the aims of value-based healthcare presumes the free flow of all kinds of data—including physician free text notes, radiology images and medical journal articles—whenever and wherever they’re needed so providers can deliver personalized, evidence-backed care and patients know their options and can make informed choices. In this healthcare utopia, patients might directly tap the power of artificial intelligence (AI) for many of their routine care needs to avoid trips to the doctor’s office when their physical presence is unnecessary.
It is a world being built atop FHIR (pronounced “fire”), the Fast Healthcare Interoperability Resources draft standard describing data formats and elements and an application programming interface (API) for exchanging electronic health records (EHRs). FHIR was developed by the nonprofit standards development organization Health Level Seven International (HL7), which has been championing interoperability for more than three decades.
FHIR founder and product director Grahame Grieve says support for FHIR has been mushrooming for several years but got a huge boost in February when the Centers for Medicare & Medicaid Services and the Office of the National Coordinator for Health Information Technology (ONC) released proposed rules that include new certification criteria for health IT tools, including the use of APIs and FHIR. No one knows how many FHIR evangelists are out there because interactions happen on social media as well as face-to-face meetings, but estimates range from 500 to 1,000 people worldwide, he says.
One of its earliest fans was Joshua Mandel, M.D., chief architect at Microsoft Healthcare, who was researching health data APIs back in 2010 as a member of the SMART (Substitutable Medical Applications, Reusable Technologies) team at Boston Children’s Hospital—the primary pediatric program of Harvard Medical School. As a medical student, he developed an automation tool for compiling data on what happened to his patients overnight that shifted his career path to medical informatics, he says.
SMART was one of a handful of ONC-sponsored research projects that since 2010 has been developing tools and platforms that would talk to an EHR. But without standards, Mandel says, the API couldn’t be deployed at scale. SMART ported its app platform to FHIR, helping shape the draft standard, and contributed the SMART app framework to the standards process. HL7’s Argonaut Project brought together commercial EHR vendors and the provider community to help make that a reality.
The work of getting FHIR to full production mode is happening on multiple fronts, says Grieve, and the chief accelerators are the Argonaut Project, Integrating the Healthcare Enterprise (IHE) and two other initiatives—the CARIN Alliance and the Da Vinci project. Participants in the HL7’s FHIR Accelerator Program are also in the trenches trying to learn what reasonably can be standardized, including EHR vendors that have a “strong vested interest in consistent practices between institutions because it reduces the surface space they need to manage in their products.”
The Argonaut Project
In the landmark JASON report, a government advisory panel of scientists heavily criticized the current state of healthcare interoperability and recommended making better use of public APIs. That effectively gave birth in 2014 to the Argonaut Project under the HL7 umbrella, says Grieve, bringing together a half dozen top health systems and a similar number of major EHR vendors to develop a first-generation FHIR-based API built on modern internet conventions widely used in other industries.
“FHIR is really healthcare of the web and one of the key adoption milestones was when Argonaut decided to adopt the standard for development of use cases,” says Grieve. An important first step was building “profiles” (aka implementation guides) to accommodate various data exchange scenarios.
Among the use cases being explored by the Argonaut Project are FHIR-based scheduling APIs, including one to enable better appointment functionality across EHRs, says Grieve. “Here in Australia, if I see my primary care practitioner and he recommends I see a specialist, it’s my problem to figure out when I’m going to see him.” Another, more advanced scheduling API under consideration would help specialists orchestrate multiple treatments, procedures, and tests for individuals with complex conditions, such as breast cancer, as part of their EHR workflow.
The Argonaut Project has also been collaborating with the SMART Health IT team at Boston Children’s Hospital to produce specifications for use cases whereby clinicians and patients use web-based and mobile apps to access FHIR resources. Another focus last year was looking at how to facilitate bulk data export to a data warehouse for analysis—"a gap in functionality that the ONC is particularly interested in closing,” Mandel says.
In January 2018, after Apple announced it was adding the Argonaut implementation guide to the operating system of iPhones, dozens of institutions quickly began supporting the FHIR-based API allowing patients to securely and directly download their health records using a smartphone app. Apple has also announced that it is collaborating with the U.S. Department of Veterans Affairs on a Lighthouse Gateway project to develop API tools for veterans. The VA Health API is designed to help veterans self-manage their health and share their data and uses the Argonaut API and SMART on FHIR to launch and authorize apps.
Numerous provider-facing apps also employ SMART on FHIR, says Mandel. One of the earliest EHR-based apps was Meducation, which creates personalized medication instructions for patients who have low health literacy, impaired vision, and language barriers. The SMART app gallery today includes 70 apps in categories ranging from care coordination and clinical research to genomics and risk calculation.
Movement toward a standardized EHR has been understandably slow going. It’s an ongoing process that HL7 can only facilitate via efforts to reach consensus on mandatory data fields. “Every institution is maybe $100 million worth of different,” says Grieve, and not so much because EHR vendors like Cerner and Epic are blocking progress but because decision-making at provider organizations is spread across individual department heads with different needs and priorities. EHRs are also notoriously unpopular with physicians who have little interest in being the primary data collectors.
Opening the Spigots
Digitizing manual, error-prone processes that waste people’s time is the goal of the Da Vinci project, making it a particularly active group with 14 use cases underway that Grieve expects will “deliver real change to the industry” this year. The use case for streamlining prior authorization has received a lot of attention because it would enable providers to request and receive immediate authorization at the point of service, he notes. “A lot of work is being done with passion and focus because CMS is expecting to reference it as a requirement in forthcoming [HIPAA] rules.”
The Da Vinci group is developing guidance to ensure that information flows between payers and providers, and payers and patients, in ways that support value-based care, Mandel says. The CARIN Alliance is doing the same for patient app development, including best practices for privacy agreements and terms of service.
The larger objective of the CARIN Alliance is to bring about consumer-directed health information exchange but remains “a pipe dream at present” due both to a lack of specific regulations and enabling technologies that are still under development, says Grieve.
Other FHIR Hot Spots
One group of FHIR enthusiasts is actively working on a set of agreements about how genetic test results get reported to enable development of decision support algorithms for genomic testing laboratories around the world, says Grieve. More generally in research, the focus is on automating the consent management process around bio-samples.
The nation’s largest health information exchange, The Sequoia Project, has separately launched its own FHIR-based provider directory. Meanwhile, IHE International is creating profiles that leverage the new FHIR specification to standardize information sharing between computer systems. The FHIR standard is being adopted in some countries for diagnostic ordering and reporting but in the absence of regulation in this arena mandating standards be used it has been slow going, Grieve says.
The health data interoperability movement has been a boon for companies providing cloud storage, computing power and machine learning tools, says Mandel, including Microsoft. Last fall, the company released its open source FHIR Server for the Azure cloud—the first of a growing set of FHIR services that now also includes Azure API for FHIR.
The day will come when providers will routinely be making clinical treatment decisions using all available evidence and clinical guidelines rather than just the portion they can humanly recall, Mandel says, and the information will flow seamlessly into the clinician experience. Research results will be easier to query across a population of patients, as will the distribution of outcomes for patients who look a lot like the one sitting in front of them. “FHIR is one of the critical foundational components.”
While hard to fully imagine, healthcare in another three to five years will truly be a world on FHIR, says Grieve. Getting there will require “difficult conversations” with provider organizations and solution vendors that fear participating in a truly consumer-mediated exchange might cost them patient market share to competitors or distract them from providing real healthcare, says Grieve.
Still to be solved is the question of how to deal with less empowered patients who may not want to take ownership of their data or consult digitally with AI service providers, says Grieve. “It’ll be interesting to see how this plays out, and if this is even a good idea.”