Development and Validation of an Automated Cancer Stage Classifier for Real-World Oncology Data Mapped to OMOP CDM
At ISPOR Europe 2025, we presented our latest work on developing an automated cancer stage classifier for real-world oncology data mapped to OMOP-CDM. A major step toward enabling scalable oncology research — with thanks to the physicians taking part in the FAIR-ICI project (*).
Challenges:
• Cancer stage is critical for oncology RWE but is often missing/inconsistent in EHRs.
• Heterogeneous recording across sites/tumor types limits comparability and reproducibility.
• Mapping to OMOP CDM enables harmonized inputs from structured fields and NLP-extracted text.
Study aims:
• Develop and validate an automated, rule-based cancer stage classifier covering >20 cancer types.
• Integrate structured TNM and NLP-derived evidence to maximize completeness.
• Encode UICC 8th edition staging per cancer type with pathological > clinical precedence and inference from partial TNM data sources.
(*) Prof. Christof Vulsteke & team at AZ Maria Middelares, Prof. Philip Debruyne & team at az groeninge, Dr. Annelies Verbiest & team at UZA, Prof. Sandrine Aspeslagh & team at UZ Brussel
Study aims:
• Develop and validate an automated, rule-based cancer stage classifier covering >20 cancer types.
• Integrate structured TNM and NLP-derived evidence to maximize completeness.
• Encode UICC 8th edition staging per cancer type with pathological > clinical precedence and inference from partial TNM data sources.
(*) Prof. Christof Vulsteke & team at AZ Maria Middelares, Prof. Philip Debruyne & team at az groeninge, Dr. Annelies Verbiest & team at UZA, Prof. Sandrine Aspeslagh & team at UZ Brussel
