United StatesHealthcare and life sciences2024 – ongoingSCAN Health Plan
Machine learning and language processing to reduce member disenrollment
SCAN Health Plan cut voluntary disenrollment by 4.59% during annual enrollment using predictive models
Problem
SCAN is a not-for-profit Medicare Advantage health plan serving high-needs seniors. Much of what it knew about members sat in unstructured data such as call center notes, medical charts and PDFs. It wanted to spot members who were likely to leave or to have a major health event.
Approach
SCAN used natural language processing on its unstructured records to find patterns in member behavior. It built machine learning models on the Databricks platform to predict major health events, voluntary disenrollment and flare-ups of chronic conditions. The model pipeline can trigger case management and home care. For some members it moved up the reassignment of their primary care physician.
Results
Takeaway
Models that flagged at-risk members let the plan act earlier, and the page ties that to lower disenrollment. The page did not give the size of the member groups compared.
Categories
- Industry
- Healthcare and life sciences
- Business function
- Customer service
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Databricks describes how SCAN used language processing and machine learning on the Databricks platform to predict health events and member disenrollment. The page reports a 3.6% drop in voluntary disenrollment for members targeted for earlier physician reassignment and a 4.59% reduction during annual enrollment.
Full details are in the original sources.