Case Study: Sharp HealthCare achieves 80% accurate prediction of patient decline for proactive RRT interventions with Cloudera

Using Machine Learning and EMR Data to Predict Patient Decline

Sharp HealthCare, a large not-for-profit system in San Diego, explored whether predictive analytics could help its rapid-response teams (RRTs) detect patient deterioration sooner. Facing the challenge of manually reviewing charts to spot crises, Sharp ran an eight-week proof-of-concept to see if machine learning on Cerner EMR data could automatically identify patients likely to need an RRT intervention within the next hour.

Working with Intel and ProKarma, Sharp built a scalable ML model on a Cloudera cluster using vital signs and other EMR features; the PoC achieved about 80% accuracy (≈82% precision) when tested against historical data. The result demonstrated the potential to trigger earlier, data-driven interventions, improve outcomes and resource utilization, and extract more value from existing EMR investments while guiding further model refinements.

Sharp HealthCare

Brett MacLaren

Vice President of Enterprise Analytics