Our Recipe for Analytics-Ready Health Data

The following “recipe” summarizes our motivation and approach for transforming health data into concepts that are ready for analysis and model-building. If you’d like to skip our “story” and jump right into the technical details, you can learn more about the tech stack behind our analytics pipeline in this technical overview.

Our thesis: Insightful analysis requires data that has been substantively enriched beyond the “raw ingredients” that can be obtained from most health data sources. Here are the steps we follow to prepare and serve your health data.

Ingredients

Longitudinal health information from patients, providers, and payers
An extensive library of data transformations
One high performance data pipeline
One self-service data platform API

Instructions

Use HIEBus™ to acquire and integrate a broad spectrum of health data from sources including EHRs, claims, and patient-reported data captured via mobile devices. Post-integration, HIEBus will link person records across data sources and rationalize terminology.
Expect that transactional health data will be difficult to use directly in many analytics use cases. Transformation into analytics-ready structures will be required. Use simple transformations to cleanse, standardize, and reorganize data into denormalized structures that are suitable for dashboarding applications and modeling tools. Use complex transformations to generate new structures to represent higher-order concepts such as: up-to-the-day risk scores, disease profiles, medication therapies, and clinical quality measures.
Use a high performance data pipeline to compute data transformations on large populations in near-realtime. Insight and patterns will be more useful and actionable if they can ultimately be expressed as opportunities using up-to-date trends and worklists.
Reveal the analytics-ready data structures using Amazon Redshift. From Redshift, hundreds of analytics applications, data science libraries, and programming frameworks enable interactive dashboards, statistical analysis, and predictive models.

Learn about the technical implementation of the Data Platform API.