Researchers have developed DiSAH, a novel federated algorithm for survival data analysis that enables multi-site collaborations without sharing patient-level data. This method uses a non-iterative structure to aggregate summary statistics, removing the need for a central server and preserving privacy. DiSAH is the first federated approach capable of estimating hazard differences, which are crucial for understanding the impact of risk factors on event rates. Applied to a dataset of nearly 50,000 emergency department patients from the United States and Singapore, DiSAH achieved accuracy comparable to centralized analysis and identified mortality risk factors that individual sites could not detect. AI
IMPACT This new federated learning approach could enable more robust medical research by allowing institutions to collaborate on sensitive data without compromising privacy.
RANK_REASON The cluster contains an arXiv preprint detailing a new algorithm for statistical analysis. [lever_c_demoted from research: ic=1 ai=0.7]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
