Background: Plug-and-play data analytics or finding ways to more efficiently export algorithmic models to use with large data-sets has been steadily entering fields outside healthcare for the past two years.

Data markets have been around for several years with offerings such as data.com (Salesforce), the Azure Marketplace, Factual, Socrata’s OpenData, Infochimps.com and DataMarket.com to name a few. These markets exist so that companies have the option of outsourcing parts of the data value chain to third parties who can extract insights.

Algorithmia.com is one example of a general marketplace for building and sharing algorithms and for making them available as a web service. We’ve seen over the years a number of data markets emerge and the next step in their evolution could also be in the direction of these new algorithmic markets. LexisNexis Health Market Science has another type of health data market with claims data in a national level warehouse containing data from all of the major payers totaling nearly 1.2 billion claims that have been scrubbed and anonymized. Clients have the ability to obtain market insights on referrals – market dynamics for facilities.

Orion Health: Enabling Analytics on an Open Platform

Over the past year and a half or so, clinical network management (CNM) vendor Orion Health has been building a platform to support plug-and-play analytics across the provider, payer, and government spectrum of clients. The difference here is that the platform is not only an analytics platform, but rather a comprehensive platform that offers tools for entire pathways and workflows. They built the platform to plug the gap that exists between HIEs and analytics systems.

The Orion Health platform is built on open source tools including Apache’s Cassandra, Spark and Elastic Search. Cassandra will serve as the data lake for the Orion Health platform within their Amadeus Data Engine. With Elastic Search, Orion Health and its customers are able to build patient registries with all of the longitudinal data around a patient including clinical, claims, and patient relationship data and have ad hoc querying capabilities through the interface.

On the algorithmic side, Apache Spark – which is Hadoop compatible but some believe is the next evolutionary step beyond Hadoop – offers strong machine learning (ML) integration. Apache Spark has most of the relevant ML libraries for logistic regression, linear models, naïve Bayes, and clustering, making it easier to run other algorithms. The platform employs a rich set of APIs that can support algorithms in any base language. Many of their clients with more advanced analytics capabilities, such as the large payers, are including algorithms that they have internally developed and will prefer running their algorithms on the Orion Health system.

Orion Health’s approach to an algorithm marketplace is somewhat different from the Apervita example. Their approach is to use an API layer that serves as an app store that offers APIs and Analytics APIs and is not solely concerned with algorithms. In the future the plan is to build a clinical pathways forum. As a CNM vendor first and foremost, their view is that the platform must be able to support the entire care pathway and not just analytics.

Summary

While these two examples for different models of algorithm markets are in the early days, they do offer important new options for providers, startups or smaller players who may not have the data science chops for developing robust algorithms in-house. One additional model that may enter this market from the margins is the IBM Watson approach to building an eco-system of collaborators who can tap into the analytics engines of Watson via APIs.

While the business models are very different, this is another way of extending the platform potential of Watson while gaining access to ever widening datasets that can further the development of the Watson analytics capabilities.