What happened?
The MIRROR study compared manual electronic case report form entry to automated extraction from electronic health records using an AI system for 113 breast cancer patients enrolled in 11 trials at Virgen del Rocío University Hospital (Sevilla, Spain). The system returned source-text references and achieved high accuracy, precision, recall and F1-scores for most structured and many unstructured clinical variables.
Why does it matter?
Manual data entry is time-consuming and error-prone. Automating extraction could speed data collection for trials and real-world evidence while reducing transcription mistakes. The study demonstrates feasibility of AI-assisted extraction but highlights remaining challenges, especially for unstructured clinical notes and cross-site standardisation of eCRFs.
How to interpret the evidence
Results are promising but come from a single-centre retrospective sample (113 patients) and specific local EHRs. Performance may vary with different hospital systems, languages, documentation styles and multicentre eCRF formats. European or Maltese sites should regard these findings as preliminary evidence requiring local validation before operational use.
What remains uncertain?
High accuracy here does not imply regulatory approval, universal performance, or immediate interchangeability with manual review. The study does not establish clinical decision-making safety, nor guarantee similar results in other hospitals, larger populations, different languages, or non-trial care records; further validation and standardisation are needed.
Medical News provides general health education. Individual treatment decisions depend on your circumstances and the advice of your clinical team.
This summary was prepared with AI assistance and checked automatically against the linked primary source. It has not been individually reviewed by a clinician.