Large Language Models as Medical Codes Selectors: a benchmark using the International Classification of Primary Care