Uncertainty Quantification for Clinical Outcome Predictions with (Large) Language Models