Accelerating Evidence Generation and Time to Insight With Clinical AI

There are countless insights waiting to be uncovered in real-world healthcare data, but it takes heroic effort to prepare these datasets for analysis. Manual abstraction is difficult to scale, while defining cohorts, exposures, and endpoints rigorously in RWD is burdensome – requiring both clinical and technical expertise to approximate medical logic with code and business rules. To date, AI has also been unable to decipher clinical data. Current approaches can currently read isolated notes or documents within a medical record, but lack the ability to comprehend full patient journeys. Current approaches can help you write generic code, but not help you to reason through how to define "metastatic colorectal cancer patients treated with FOLFOX" from a list of diagnostic, procedure, and medication codes in a real world dataset. To be useful for evidence generation, clinical AI must be able to reason over real-world data with physician-like intelligence.

This session will:

• Explore how to evaluate clinical-specific reasoning capacity in AI & suitability for clinical applications
• Share data on how clinical AI-enabled analytics compare to unassisted analytics
• Delve into AI explainability and controlling hallucinations