The interoperability problems that plague healthcare are compounded by the paucity of developer options for modern software development. Software as a Service (SaaS) models are increasingly commonplace for EHRs, population health and other finished applications. Platform as a Service (PaaS) models to support application developers are essentially non-existent in the healthcare sector, though we have been arguing for this need since our first HIE report several years ago.
For this reason, we were pleased to talk this week to Orion Health about its “Healthier Populations” roadmap to support population health efforts of providers. While the roadmap has many components, one aspect is especially relevant to developers: Orion Health’s plans to leverage “open” technologies to create a PaaS for Healthier Populations.
Briefly, Healthier Populations consolidates all data sources into a NoSQL database (Cassandra) with an API management layer to support clinical applications – both clinician and consumer facing. The new Cassandra data platform also connects out to an in-memory Map-Reduce engine, Spark, which has some very strong performance metrics when compared to the currently popular Hadoop.
Unlike most of the vendors we routinely talk to, Orion Health’s plan uses technologies and development ideas that leverage modern distributed processing concepts, and web-based client functionality. It plans to assemble a comprehensive clinical and financial data source (EHR, HIEs, payer sources) with an API that will provide roughly the data-sharing network envisioned by the JTF. For now, the API is proprietary but we expect the company to provide the JTF’s recommended FHIR-based open API if and when it becomes a reality.
Orion Health is currently piloting this solution at Cal INDEX, a health information organization (HIO) formed by California’s two largest payers but open to any healthcare data contributor in California. It has information on 9 million patients and aspirations to encompass the health data of most everyone in the state (some 25M). The technology that Orion Health’s new Healthier Populations platform is built upon can certainly handle that scale. What remains to be seen is how quickly Orion Health and Cal INDEX will open up the platform for independent software vendors (ISVs) as the current plan is to do internal development of apps, especially consumer-facing.
This table provides our general impressions of this new offering from Orion Health:
Orion Health Open Platform |
|
Advantages |
Challenges |
|
|
Orion Health’s Healthier Populations platform is their attempt to assemble a general-purpose, open infrastructure to support a healthcare organization’s (HCO) population health management (PHM) strategy. Orion Health has rightly assumed that they alone will not be able to provide all the capabilities required for PHM and have instead developed a truly state-of-the-art PaaS platform, built on open technologies. This provides a certain level of future-proofing of the solution suite and the flexibility to optimize the platform to meet specific community PHM needs. We are hard pressed to name any other HIT vendor that is in a position to offer anything similar although many will say that they do.
Let’s fix the error on comparing Spark to Hadoop. You wrote:
“The new Cassandra data platform also connects out to an in-memory Map-Reduce engine, Spark, which has some very strong performance metrics when compared to the currently popular Hadoop.”
Try this:
Orion Health is leveraging Cassandra’s non-HDFS data model and connecting to an in-memory compute engine called Spark, which has some very strong performance metrics when compared with its Map Reduce predecessor.
Why:
Even though Map Reduce was part of Hadoop it feels wrong to make a comparison between Spark and Hadoop. Cassandra, Hadoop, Map Reduce, Spark, they’re all Apache projects. There’s nothing new being done here by Orion Health at all even in healthcare. It’s only news because Orion is a company that used to apply a big iron mentality like others in the data rationalization or integration space. It’d be super great to see the EHR and HIE vendors alike adopt a more modern data architecture but none of us are holding our breaths while they are making money hand over fist selling 1995 technology still.
Christopher,
Thanks for taking the time to provide feedback on Cassandra, Spark and Hadoop. The secondary research I did for this write-up led me to the conclusion (based on experiences of others) that Spark is different (higher performing) than Hadoop when it comes to Map Reduce functionality. Yes, I do know that all of them are Apache projects but Hadoop is one that has been around awhile whereas Spark was only recently released.
And as to the healthcare vendors themselves, I beg to differ with your opinion. Virtually all vendors, from Orion Health to athenahealth have been building their products on dated tech as that is what their customers, for good or bad, have come to expect, or maybe they just have very low expectations! Fact is, Orion Health is doing something different than here than what we have seen from other major mind share HIT vendors and that needs to be reported. After all, the biggest selling enterprise solution in the market today is not using tech from 1995, but far earlier – MUMPS, which predates UNIX!