Interpreting free-text cardiac catheterisation reports: A machine learning approach informed by focused ethnography

Aim: To examine how focused ethnographic insights can inform the development of a machine learning pipeline to improve the extraction of clinically relevant information from percutaneous coronary intervention (PCI) documentation and support nursing education and practice. Background: Cardiac catheterisation procedures produce detailed documentation, often embedded in free-text fields in electronic health records. For nurses delivering post-procedural care, extracting this information is time-consuming and prone to error. While machine learning (ML) offers automation potential, many models struggle to handle contextual and structural inconsistencies in real-world documentation. Design: A qualitative-informed machine learning study using focused ethnography and rule-based model development. Methods: The study was conducted at a tertiary medical centre in Taiwan and included 200 h of non-participant ethnographic observation to explore documentation practices in PCI reporting. Ethnographic data were thematically analysed to identify structural patterns, linguistic variability and workflow behaviours. These insights informed the iterative development of a rule-based ML pipeline, which was tested on 4128 de-identified PCI reports to evaluate extraction accuracy. Results: Three key patterns were identified: structured use of templates, formatting inconsistencies and free-form narrative variability. These informed the application of four extraction strategies: (1) rule-based and ontology-driven methods; (2) statistical topic modelling; (3) deep learning models and (4) large language models. A rule-based approach was selected for its adaptability and interpretability. Extraction accuracy exceeded 99 % in structured fields and approximately 50 % in narrative-rich sections. Conclusion: Combining ethnography with machine learning enhances automated clinical documentation interpretation and supports AI-informed nursing education through improved digital literacy and contextual awareness.