Advancing privacy-preserving causal inference methods for COVID-19 treatment analysis with innovative AI-assisted collaborative modeling
Challenge Overview
Develop Structural Causal Models (SCMs) with LLM assistance to estimate effects of glucocorticoids on COVID-19 patient survival, evaluated through secure computing environments.
Key Dates
- Launch: May 15, 2025
- Webinar: Two weeks after launch
- Submission: July 15, 2025
- Results: Early August 2025
Innovation
Combining Structural Causal Models (SCMs) with real-time collaborative canvas, generative AI, and a secure federated learning environment - maintaining privacy while enabling robust analysis.
How It Works
- Build causal diagrams in cStructure
- Benchmark using de-identified COVID-19 EHR data
- Get weekly performance feedback
- Final validation with COVID-19 cohort study data in EscrowAI
Evaluation Criteria
- Alignment with RCT results
- Proper adjustment for confounding
- Calibration-in-the-large metrics
- Plausibility of causal relationships