Describir: GPT-4 versus human authors in clinically complex MCQ creation: A blinded analysis of item quality.