Improving Clinical Information Extraction from Electronic Health Records: Leveraging Large Language Models and Evaluating Their Outputs