Describir: A framework for human evaluation of large language models in healthcare derived from literature review.