Blue Health Intelligence and Decode Health Team Up to Validate the Accuracy of Predictive Models for Chronic Disease
CHICAGO (March 10, 2020) – Blue Health Intelligence® (BHI®) and Decode Health are showing how a machine learning (ML) engine can be applied to large data sets to more accurately forecast which patients will get chronic diseases. Decode’s analysis used a representative subset of BHI’s National Data Repository, the largest and most longitudinally rich data set in the nation, to show how healthcare claims data can be segmented to focus on specific sub-populations.
Decode, a Nashville-based analytics company, developed an ML engine that uncovers specific patterns of chronic disease risk. Decode was initially focused on inflammatory autoimmune diseases, which are difficult to diagnose and cost the healthcare system approximately $90 billion a year through serious adverse events and hospitalization.
The company wanted to test the theory that its ML platform could be used to construct predictive models for multiple chronic diseases. To do so, Decode needed a comprehensive, multi-year, longitudinal data source that captured a wide range of patient populations across different geographies, so it turned to BHI.
BHI worked with Decode to develop a custom claims data set that included three years of continuous information from 2 million people located in a region where autoimmune conditions, such as multiple sclerosis (MS) and Crohn’s disease, are most prevalent. Decode trained and tested a series of disease-prediction models to identify patients who had not received a conventional autoimmune diagnosis, but who were predicted to be diagnosed in the future.
The company further used BHI data to see if clinicians appropriately documented the diagnoses of autoimmune diseases and treatments, and used data from those claims to set up cost profiles based on the actual date of diagnosis versus the predicted date. Using Crohn’s as an example, the cost profile for patients who received a diagnosis two years after Decode’s predictions suggested they were twice as high, on average than patients who were diagnosed within one year of the predicted date.
“By mining BHI’s data, we identified specific patterns for MS and Crohn’s that were indicators of future clinical events and periods of higher or lower future healthcare utilization,” said Chase Spurlock, PhD, CEO of Decode Health. “Being able to track specific patients who exhibit patterns of future disease risk is a significant breakthrough.”
Decode also proved the ability of its ML to identify and monitor patients who had had an undiagnosed disease by using BHI’s data to perform detailed studies of those patients. Decode’s model correctly identified a subset of 200 patients whose treatment costs made up 90% of the total cost profile. This approach resulted in additional cost-savings opportunities.
“Our work with Decode Health is exactly what we want to see when we collaborate on research projects with other healthcare stakeholders,” said Swati Abbott, BHI’s CEO. “With the depth and breadth of our data, which includes enough members to study populations with rare conditions, we can support HIPAA-compliant analyses that ultimately improve healthcare costs and quality for all Americans.”
About Blue Health Intelligence
Leveraging the power of medical and pharmacy claims data from more than 200 million Americans, Blue Health Intelligence® (BHI®) delivers insights that empower healthcare organizations to improve care, reduce costs, and optimize performance. With the largest, most up-to-date, and uniform data set in healthcare, BHI provides an accurate representation of the health profile of commercially insured Americans. Our team of data analysts, clinicians, IT experts, and epidemiologists provide analytics, software-as-a-service, and in-depth consulting to payers, providers, employers, medical device companies, and other healthcare stakeholders. BHI is an independent licensee of the Blue Cross Blue Shield Association and carries the trade name of Health Intelligence Company, LLC. For more, visit bluehealthintelligence.com.
