A number of electronic health record vendors now grant access to EHR data to programmers at user sites through an application programming interface (API). Hospitals and ambulatory practices are making great use of APIs to extract data on patients and analyze it to improve care (or at least their bottom lines–it’s good for marketing to patients as well as protecting them from adverse medical events).
Far fewer EHRs allow totally new applications to be built by outsiders, suitable for sale or free distribution to health care providers. To find out more about how this can be jump-started, I talked recently to Chip Ach at athenahealth, whose outspoken CEO Jonathan Bush has committed to dealing with health care in new ways. (I reviewed Bush’s book, Where Does It Hurt?,in another article.) Chip Ach is senior architect for athenahealth’s More Disruption Please (MDP) program.
MDP is building an open, API-based ecosystem on athenahealth’s cloud-based platform for entrepreneurs to connect, collaborate, and scale more efficiently in health care. athenahealth allows both established and start-up health IT companies to sign up as partners, creating a large marketplace of innovative, third-party health care solutions. Applications solve such problems as scheduling, speeding up payments, and increasing engagement with patients.
The value of a third-party market
Let’s step back a minute to look at why both vendors and health care providers would want a marketplace for companies to write apps based on their EHR data. Start with the obscurity of the data itself: very few health care providers can find out from their EHRs which patients are frequent ER visitors or which surgeons have consistently better outcomes–data that could trigger critical interventions. Even fewer EHRs provide information to patients, and the patient portals that some providers have made available usually offer extremely limited views.
Second, most EHRs have poor interfaces, a gawkiness that becomes especially evident on the tablets and cell phones that are just as popular among clinicians as the general population. A healthy competition and atmosphere of experimentation among interface developers will greatly improve EHR usability.
Clinicians are just too diverse for one vendor to take into account all their needs. More often than now, clinicians report a drop in productivity after the introduction of EHRs, even after the users get acclimated. If an agile development process could create unique and customizable interfaces, doctors and nurses could go back to focusing on patients.
Finally, unanticipated uses for patient data can emerge from an open marketplace. The use of analytics and clinical decision support could skyrocket.
athenahealth management decided several years ago that the health care system needs a radical change toward generating and accepting new solutions to its pain points. Calling this a “Health Care Internet,” they see it as a place for health care providers to compare and “shop” for nearly any product or service they might want or need, whether it be data management, digital check-in, self-pay, mobile charge capture, or transcription. MDP and the athenahealth Marketplace are their contributions to change.
Steps toward athenahealth’s Marketplace
It takes a special program like MDP to draw developers to a platform, particularly in health care where so many developers are deterred by the field’s complexity and issues of liability. In keeping with the company vision to straighten out the operation of the health care system, athenahealth embraced an open platform, adhering to industry standards such as HL7 and investigating the publication of open APIs.
At the same time, their programmers were modularizing a rather monolithic code to reap the productivity benefits that this would offer. Modularization facilitated the design of APIs to connect one module with another. After announcing a few open APIs and seeing the program catch on, the MDP team asked itself: why not make all their modules available to outside programmers, including those that were previously considered internal?
Early in the program, they found a company eager to develop an appointment scheduling application, and worked closely with them to make it possible. Ach’s team promised to get the API working in three months, which was quite a tight deadline. None of the team had prior experience doing an open API for a market, and they needed to face a lot of new requirements that openness created–for instance, adding an authentication layer, deciding how to deliver documentation on the APIs, and offering the opportunity for developers to get a feel for using the APIs. They now have a developer’s portal open to those who sign their modest terms and conditions. There is no fee or vetting process.
To speed up delivery of an API, athenahealth developers sought out a management service and chose Mashery to do the job. In the end, after a lot of long workdays, they met their three-month deadline. The partner was so sure they’d miss the deadline that it wasn’t ready to use the API–but it caught up and successfully released the app. Along the way, the MDP team learned some of the range of data that needed to be exposed by their APIs.
Eventually, athenahealth dedicated itself to rigorously modularizing their code with well-defined interfaces between all components–like Amazon.com–and to exposing as many interfaces as they could to partners. Naturally, this was potentially frightening; both managers and programmers have to take a leap of faith. What would design decisions made years ago look like when put out in the light of day? How many changes would they have to make?
According to Ach, developers were not resistant for ego reasons (they didn’t mind showing the architecture to outsiders), but worried about their ability to seamlessly upgrade the API in the future. A lot of decisions had to be made quickly before opening the APIs to third-party developers, who would come to depend on them and assume they would remain unchanged. Ach says the teams have been happy with their choices, although early on, in the effort to keep the process moving forward, they didn’t create all their naming standards and left that to be done later.
So the process went smoothly on the inside–but how did it fare externally? That’s the subject of part 2 of this article.