Summary
Sonata delivered a Gen AI-driven patient summary solution for a large healthcare data platform used extensively by providers. By leveraging Phi-3 small language models and Sonata's Harmoni.AI framework, the solution generates concise patient summaries from extensive medical records, enabling healthcare providers to quickly access critical patient information. The implementation resulted in a 70% reduction in manual patient data processing time and a 50% lower administrative workload.
Client Overview
A leading healthcare technology firm delivers advanced population health management solutions, integrating clinical and administrative data to improve care coordination, quality outcomes and analytics for underserved communities.
Pressure Points
Difficulty navigating patient data spread across multiple platform pages
Time-intensive manual summarization processes consuming valuable resources
Delays in diagnosis and treatment due to inefficient data access
Challenges in quickly accessing critical patient information during care decisions
Solutions
Sonata implemented a Gen AI-driven patient summary and insights solution powered by small language models to address the challenges of fragmented and time-consuming patient data processing. The solution automatically generates concise and contextually relevant summaries by intelligently collating information from across medical records. Additionally, it delivers actionable insights derived from the summarized data, empowering healthcare providers to make faster and more informed clinical decisions.
Leverage Phi-3 small language models (SLM) to reign-in Gen AI models costs
Implemented Sonata's Harmoni.AI framework for responsible Gen AI consumption
Deployed serverless API for small language model in Azure AI Studio
Created a simple user interface displaying patient summary information in pop-up windows
Integrated data store with summarization application for seamless prompt building
Results that Speak Volumes
Up to 70% reduction in time taken for manual patient data processing
~50% lower administrative workload related to patient data management
Reduced chances of human error in patient data interpretation
