Transforming unstructured breast cancer pathology reports into the Observational Medical Outcomes Partnership Common...

Area of Science:

• Biomedical Informatics
• Oncology Research
• Data Standardization

Background:

• Breast cancer pathology reports contain vital clinical data but are unstructured free text.
• Previous efforts utilized natural language processing (NLP) for data extraction from these reports.
• This study focuses on standardizing extracted pathology data into the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM).

Purpose of the Study:

• To present a process for standardizing breast cancer pathology data into the OMOP CDM.
• To demonstrate the utility of this standardized data for retrospective oncology research.
• To provide a framework for other institutions to adopt.

Main Methods:

• Extracted data from 13,857 surgical pathology, 11,374 immunohistochemistry, and 5,904 molecular pathology reports.
• Employed rule-based NLP, including text pre-processing, segmentation, and pattern-based extraction.
• Mapped extracted data to OMOP standard vocabularies and integrated into an existing OMOP CDM.

Main Results:

• NLP-derived pathology data populated OMOP tables like NOTE_NLP, CONDITION_OCCURRENCE, and MEASUREMENT.
• A feasibility study demonstrated the utility of the integrated data.
• Patient-level prediction analysis yielded an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.840 for 5-year all-cause mortality prediction.

Conclusions:

• Integrating breast cancer pathology data into OMOP CDM facilitates unified analysis of clinical and pathological data.
• The NLP-based standardization framework is adaptable for other institutions.
• This approach supports retrospective oncology research and the expanding use of OMOP CDM in cancer studies.