Last spring, Partners HealthCare, founded by Brigham and Women’s Hospital (BWH) and Massachusetts General Hospital, and California-based software company Persistent Systems, announced a strategic collaboration to develop a new industry-wide open-source platform with the goal of bringing digital transformation to clinical care.
Indeed, with the digital platform, Partners’ leaders hope to enable greater exchange of information across healthcare providers everywhere, and make available open source applications to any health system. At the time of the 2017 announcement, officials said that the co-developed digital platform will be based on Substitutable Medical Applications & Reusable Technologies (SMART), an open, standards-based technology platform along with Fast Healthcare Interoperability Resources (FHIR). “The platform will enable provider systems across the country to rapidly and cost effectively deploy industry-leading best practices in clinical care across their ecosystems,” according to the announcement.
Healthcare Informatics Managing Editor Rajiv Leventhal recently spoke with Sandy Aronson, executive director of information technology at Partners Healthcare, about this collaboration, its specific goals and outlook, and how things have come along so far. Below are excerpts of that discussion.
What would you say is the greatest significance behind this collaboration?
I have been at Partners for about 15 years, and the first 13 of those years were primarily focused on the clinical use of genetics and genomics. In that space, we created a suite of applications that was architected differently than health IT applications are typically architected. These were applications that helped with the generation of interpreted reports for genetics and genomics sequencing test results. So, where normally in health IT applications you create a transaction system and then try to bolt a knowledge base on top of it to the extent you can, we decided to architect this in the opposite way.
We built a knowledge base that deeply modeled the tests that a laboratory offers, the genes that are covered by that test, variants known to exist in these genes, variants that are learned over time, and the state of knowledge linking those variances to clinically relevant facts—so disease states, drug response, drug efficacy, etc. So we built this deep knowledge base and built a transaction system on top of it, and made a rule that you can’t report out test results unless you keep the knowledge base up-to-date and consistent with your test results. And that enables you to automate the generation of reports.
But as a result, we wound up with this continually-updated knowledge base, so based on that we created what would now be a SMART on FHIR app that plugs into the EHR [electronic health record] and provides clinicians with alerts if something new and potentially clinically relevant is learned about a variant previously identified in one of their patients. So it created this notion of a knowledge base alert being interjected into clinical care.
We studied this and found that clinicians liked it, but the rate at which this learned was dependent on the number of transactions that flow through the system, because that’s how geneticists would gather the data that would enable them to improve their assessment of variants. So we registered this as a medical device, distributed it outside of Partners, and networked the different instances together, so it could learn not just based on our volume, but other folks’ volume as well. Ultimately, we sold that to Sunquest [Information Systems]. The thing we feel was most important was creating this infrastructure that facilitated new clinical processes and captured, shared, and federated data in a way that enabled learning to care.
After having done that, we took a step back and said OK, what should we do next? The infrastructure we built was very specific to issues where genetics and genomics are the major components to deciding what to do for a patient. So we wanted to look at all of the things that made that infrastructure hard to do, and build a platform to make it easier to build things like GeneInsight [an IT platform company owned and developed by Partners], and then distribute that platform, so that in addition to building examples of a similar infrastructure, others can build those examples, too. We wanted that platform to make it easier to distribute apps that are created by different folks in different organizations, ultimately with the goal of networking those apps together.
We are at a unique point in time where you have these new data types coming online that can be helpful to the care delivery process, you have algorithmic-based medicine starting to come into use, both machine learning-based and not, and you have people looking at transformative ideas on how to alter clinical processes where in order to incorporate these new data types and incorporate algorithmic-based approaches to care, you need new kinds of IT support in order to enable these transitions to occur. And that creates an opportunity, not only related to the specific transitions, but also to start collecting data for specific clinical problems in a much finer-grained way that lays the groundwork for these networks that can build the data that’s required to underlie continuous learning processes.
All of this is happening in a time with incredible cost pressure in healthcare, which does constrain internal investment but also makes organizations far less resistant to change. The goal here is to fundamentally enable clinicals to evolve their practices, their care, new data, ideas, and techniques in ways they haven’t done in the past.
Sandy Aronson
And how are you working with Persistent Systems on this, specifically?
We are building this platform together. The platform is called HIP, or health innovation platform, and the platform itself will be open-source, and it sits on top of the current clinical IT ecosystem. You interface it to underlying systems, and then it handles things like some aspects of security, authentication, and HIPAA, but also access to data as well as incorporating shared algorithms.
The goal is having different places hook up the platform, and once it is hooked up, it should create a uniform surface on top of the platform so that apps built on top of the platform become more shareable and distributable. We are now focused on both building the platform and building certain apps. And the apps get interjected to the EHR as SMART on FHIR apps.
Can you give some examples and details of the apps that are being built?
One example is that we have been working with BWH’s cardiology [department] on this program that they have, where if you look at heart failure, which affects about 2 percent of the population and has a very high mortality rate with a great deal of costs associated with it, there are guidelines that have been shown to really be helpful, yet very few people are treated in a way that actually adheres to guidelines. And that’s because the process of getting them to guideline-based care involves this drug selection and titration process that requires a lot of interaction, some of which can make patients unconformable.
But as it turns out, you can instantiate a process where you use patient navigators to take patients through this drug selection and titration process, interacting with them far more frequently than a cardiologist would ever be able to, to get them to guidelines. It’s a data-intensive process. So we are providing support for that program through the HIP platform today and we are really focused on deepening that support.
What are your goals in the next 12 to 24 months regarding this partnership? What would you like to see happen?
The ideal world is that our group and Persistent Systems will continue to add more capabilities to the platform, and that the platform is reducing costs. So many clinicians have ideas on how to fundamentally improve care but they can’t put those ideas into use without these kinds of IT interventions.
One thing I hope is that this will continuously reduce the cost of building those interventions and as a result, our team, and others, too, will develop more of these apps. We hope to see some cross-institutional adoption of apps built here and elsewhere, that the sharing will begin at the app level and ideally, in two years or so, we will be having real conversations about how we can get the networking between apps really going.
