The major goals are to create a validated phenotype for clinically relevant bleeding using natural language processing incorporating large language models, and to derive an optimal risk prediction model for bleeding using available clinical and biomarker data.