Improving Clinical Validity in Synthetic Electronic Health Record Generation Using Best-of-N Sampling: Comparative Evaluation Study
{{output}}
Background: Synthetic electronic health record generation is limited not only by statistical fidelity but also by clinical validity. Records that appear statistically plausible may still violate hard structural, physiological, or... ...