Unstructured clinical data

What does Unstructured clinical data mean in clinical research?

Unstructured data are clinical records that do not fit predefined fields: progress notes, consult letters, pathology and radiology reports, discharge summaries and scanned outside records. A 2023 systematic review in PLOS Digital Health notes that most digital data in healthcare are unstructured and often not readily accessible for research, lacking standardization and needing significant preprocessing.^⁠[1] A widely repeated estimate, drawn from industry sources, puts the unstructured share of medical data at about 80%.^⁠[2]

For trial eligibility, the deciding evidence is often in that text: cancer stage and biomarker results in a pathology report, ejection fraction in an echo report, prior therapies and the reasons they stopped in clinic notes. Screening that reads only diagnosis codes and structured labs will miss eligible patients and pass along ineligible ones.

Practical note: when evaluating a screening tool, ask which document types it reads and whether it shows the passage that supports each criterion decision. Identify reads structured and unstructured records and shows criterion-to-evidence rationale. See also the post on unstructured data as eligibility evidence.