Why is unstructured data a challenge in healthcare?

An estimated 60–80% of clinical information exists only in unstructured form (clinical letters, notes, radiology reports, pathology reports). This data is human-readable but difficult for computers to process without natural language processing (NLP). The richest clinical detail is often ‘trapped’ in free text, and unlocking it safely and accurately using NLP and large language models remains an active area of research.