Structuring Legacy Pathology Reports by openEHR Archetypes to Enable Semantic Querying.

Area of Science:

• Medical Informatics
• Health Informatics
• Clinical Data Management

Background:

• Clinical information, such as pathology reports, is often unstructured free text.
• Keyword searches on full documents yield many false positives, hindering efficient data retrieval.
• Semi-structured elements like headers exist but are insufficient for precise data access.

Purpose of the Study:

• To semi-automatically transform unstructured pathology reports into a structured information model.
• To enable improved access and retrieval of relevant clinical data through section-sensitive queries.
• To facilitate information reuse by storing data in a standardized, structured format.

Main Methods:

• Developed a processing pipeline including information modeling, section boundary detection, and section-sensitive querying.
• Utilized openEHR archetypes for creating a standardized patient information model.
• Transformed reports into XML-based pathology electronic health records (PEHRs) for querying with XQuery and visualization with XSLT.

Main Results:

• A keyword-based approach reliably structured pathology reports into distinct sections.
• Reusability of openEHR archetypes expedited the information modeling process.
• Standardized PEHRs enabled accurate data retrieval based on user queries.

Conclusions:

• Mapping unstructured reports to a standardized information model offers practical benefits for data access.
• Archetype-based XML facilitates section-sensitive retrieval and visualization using established XML techniques.
• Targeting specific sections for retrieval significantly improves accuracy and reduces search time.