Uncertainty Estimation of Large Language Models in Medical Question Answering