Advanced Analytic Models Within the Ursa Health Platform

By Colin Beam | 11/26/2019

Since its inception, the Ursa Health Platform has helped its user community accelerate through the many steps involved in transforming raw healthcare data into a unified, reliable, and secure data asset. Flexible tools also allow users to derive value from their data through the construction of custom measures, dashboards, and visualizations.

Once users can create and maintain a set of quality measures, a natural—and increasingly more common—extension is to use these measures as inputs to complex machine learning algorithms and statistical models. To achieve objectives such as risk prediction, health provider rankings, natural language processing, and time-series forecasting, Ursa clients have used their measures within methods such as elastic net regression models, random forests, gradient boosted trees, neural networks, principal components analysis, and hierarchical linear models. However, doing so has required the organizations to download data for use in other computing environments, multiplying the steps in the process and requiring additional security precautions.

With advanced insight discovery methods becoming the standard for innovative healthcare organizations, Ursa Health added a unique capability to the toolset in Data Studio. Our new Advanced Analytic Model feature allows the Ursa Community to implement all of the previously listed methods—as well as many more—within the Ursa Health Platform. This allows for a seamless transition from measures stored in Ursa tables, to complex transformations with machine learning methods, to reports and visualizations that summarize these results. This new capability both speeds up analyses and enhances security because data does not need to be transferred between environments to complete these distinct tasks.

Using the Advanced Analytic Model feature, analysts can write and upload bespoke Python or R code to the Ursa Health Platform, then apply this code to any table on the platform. Python and R are rich, flexible languages that are the top choices of data scientists. These languages both have large, highly collaborative communities that are continuously adding new tools in the form of shared libraries that reflect state-of-the-art methodologies from the fields of statistics and machine learning.

The Advanced Analytic Model feature now makes all of these methods available to use within the Ursa platform, giving healthcare organizations the ability to quickly and securely gain the most information from their data.