Evaluating Large Language Models in extracting cognitive exam dates and scores