Elamur Research⌕
☰
home / papers / 2026 / DeID-Clinic: A Risk-Aware Pseudonymization Framework for Clinical Text De-identification and Re-identification Risk Assessment

DeID-Clinic: A Risk-Aware Pseudonymization Framework for Clinical Text De-identification and Re-identification Risk Assessment

Paper · 2026 · Angel Paul, Dhivin Shaji, Lifeng Han et al.

Summary

Masking protected health information (PHI) in clinical notes does not by itself guarantee privacy. Unusual combinations of remaining details or distinctive context can still let someone re-identify a patient, and most de-identification tools never measure this leftover risk. DeID-Clinic tackles the problem by extending the open-source MASK platform in two ways. First, it adds fine-tuned BioBERT and ClinicalBERT taggers, which work alongside MASK's dictionary lookup and regular-expression rules. Second, it adds a document-level module that scores residual re-identification risk. Entities found by all detectors are merged, then either redacted with placeholders or replaced with synthetic substitutes. The risk score combines k-anonymity, l-diversity and t-closeness with embedding-based contextual similarity and analysis of how entities co-occur. On the i2b2 2014 de-identification data, several entity categories reach macro F1 above 0.96. The risk scores also make it possible to rank documents so the riskiest ones get manual review first. The authors argue that pairing neural PHI detection with explicit risk modelling makes data sharing more accountable, and that the approach could transfer to legal or administrative texts.

Key points

  • The authors integrate BioBERT and ClinicalBERT into the MASK framework (Northern Care Alliance NHS Foundation Trust) and combine them with dictionary and rule-based detection through a union of entities.
  • Users can choose between two masking options: redaction (placeholders such as "XXX-Name") or replacement with synthetic values. Replacement mappings are kept for use in the risk assessment step.
  • The risk module merges k-anonymity, l-diversity, t-closeness, contextual similarity and entity co-occurrence into a single document-level score, and the system returns a risk report with each output.
  • Evaluation on i2b2 2014 shows macro F1 above 0.96 for several PHI categories.
  • The system includes a human-in-the-loop interface where users can edit the detected entities. Fine-tuning ran on Google Colab T4 GPUs. The paper extends a 2-page poster presented at ICHI 2025.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of ManchesterManchester · United Kingdom87 works · 44,938 citations🎓Leiden UniversityLeiden · Netherlands32 works · 1,074 citations🤝Leiden University Medical CenterLeiden · Netherlands4 works · 2 citations

 Fields, topics and concepts

📚 Shelves 3The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksInformation Systems · 1,437 works

Cite

Cite as Angel Paul, Dhivin Shaji, Lifeng Han, Warren Del-Pinto, Goran Nenadic and Suzan Verberne. 2026. DeID-Clinic: A Risk-Aware Pseudonymization Framework for Clinical Text De-identification and Re-identification Risk Assessment. Proceedings of the Joint Workshop on Legal and Ethical Issues in Human Language Technologies and Computational Approaches to Language Data Pseudonymization, Anonymization, De-identification, and Data Privacy (LEGAL2026 and CALD-pseudo 2026) @ LREC 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/5muz36muepa2

@inproceedings{paul2026deid,
  title = {{DeID-Clinic: A Risk-Aware Pseudonymization Framework for Clinical Text De-identification and Re-identification Risk Assessment}},
  author = {Angel Paul, Dhivin Shaji, Lifeng Han, Warren Del-Pinto, Goran Nenadic and Suzan Verberne},
  booktitle = {Proceedings of the Joint Workshop on Legal and Ethical Issues in Human Language Technologies and Computational Approaches to Language Data Pseudonymization, Anonymization, De-identification, and Data Privacy (LEGAL2026 and CALD-pseudo 2026) @ LREC 2026},
  year = {2026},
  doi = {10.63317/5muz36muepa2},
  url = {https://doi.org/10.63317/5muz36muepa2}
}

References

33 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Kazuma Hara, Takuya Matsuzaki, and Yusuke Miyao. 2018. Risk analysis of de-identification methods for anonymizing biomedical text data. Journal of Biomedical Informatics, 84:136–146. find
  2. Emily Alsentzer, John R Murphy, William Boag, Wei-Hung Weng, Di Jin, Tristan Naumann, and Matthew McDermott. 2019. Publicly available clinical bert embeddings. arXiv preprint arXiv:1904.03323.
  3. Tal Hartman, Michael D Howell, Jeffrey Dean, Shahar Hoory, Ronit Slyper, Irit Laish, Oren Gilon, and Yossi Matias. 2020. Customization scenarios for de-identification of clinical notes. BMC Medical Informatics and Decision Making, 20(1):1–9. find
  4. Fida K Dankar, Khaled El Emam, Angelica Neisa, and Tyson Roffey. 2012. Estimating the re-identification risk of clinical data sets. BMC Medical Informatics and Decision Making, 12(1):1–13. find
  5. Youngjun Kim, Patricia Heider, and Stéphane Meystre. 2018. Ensemble-based methods to improve de-identification of electronic health record narratives. In AMIA Annual Symposium Proceedings, pages 663–672. find
  6. Yves-Alexandre De Montjoye, César A Hidalgo, Michel Verleysen, and Vincent D Blondel. 2013. find
  7. Aleksandar Kovačević, Bojana Bašaragin, Nikola Milošević, and Goran Nenadić. 2024. De-identification of clinical free text using natural language processing: A systematic review of current approaches. Artificial Intelligence in Medicine, 151:102845. find
  8. Stcaféphane M Meystre, F Jeffrey Friedlin, Brett R South, Shuying Shen, and Matthew H Samore. 2010. Automatic de-identification of textual documents in the electronic health record: a review of recent research. BMC Medical Research Methodology, 10(1):1–16. find
  9. Zeljko Kraljevic, Anthony Shek, Joshua Au-Yeung, Ewart Sheldon, Mohammad Al-Agil, Haris Shuaib, Bai Xi, Kawsar Noor, Anoop Shah, Richard Dobson, and James Teo. 2023. Deploying transformers for redaction of text from electronic health records in real world healthcare. find
  10. Nikola Milosevic, Gangamma Kalappa, Hesam Dadafarin, Mahmoud Azimaee, and Goran Nenadic. 2020. Mask: A flexible framework to facilitate de-identification of clinical texts. arXiv preprint arXiv:2005.11687. — [arXiv:2005.11687](https://arxiv.org/abs/2005.11687) find
  11. NCA NHS Foundation Trust. 2021. Mask api. https://github.com/ NCA-NHS-Foundation-Trust/MASK_API_ Copy/tree/master. Accessed: 2024-10-15. ↗
  12. Jinhyuk Lee, Wonjin Yoon, Sungdong Kim, Donghyeon Kim, Sunkyu Kim, Chan Ho So, and Jaewoo Kang. 2020. Biobert: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics, 36(4):1234–1240.
  13. Ishna Neamatullah, Margaret M Douglass, Li-wei H Lehman, Andrew T Reisner, Mauricio Villarroel, William J Long, Peter Szolovits, George B Moody, Roger G Mark, and Gari D Clifford. 2008. Automated de-identification of free-text medical records. BMC Medical Informatics and Decision Making, 8(1):32. find
  14. Ninghui Li, Tiancheng Li, and Suresh Venkatasubramanian. 2007. T-closeness: Privacy beyond k-anonymity and l-diversity. In 2007 IEEE 23rd International Conference on Data Engineering, pages 106–115. IEEE. find
  15. Atiquer Rahman Sarkar, Yao-Shun Chuang, Noman Mohammed, and Xiaoqian Jiang. 2024. De-identification is not enough: a comparison between de-identified and synthetic clinical notes. Scientific reports, 14(1):29669. find
  16. Jennifer Lin. 2020. A comparison of recurrent neural networks and conditional random fields for named entity recognition. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 2324– 2335. find
  17. Martin Scaiano, Grant Middleton, Luk Arbuckle, Varada Kolhatkar, Liam Peyton, Moira Dowling, Debbie S Gipson, and Khaled El Emam. 2016. A unified framework for evaluating the risk of re-identification of text de-identification tools. Journal of biomedical informatics, 63:174–183. find
  18. Zengjian Liu, Buzhou Tang, Xiaolong Wang, and Qingcai Chen. 2017. De-identification of clinical notes via recurrent neural network and conditional random field. Journal of Biomedical Informatics, 75:S34–S42. find
  19. Dhivin Shaji, Angel Paul, Lifeng Han, Warren Del-Pinto, Goran Nenadic, and Suzan Verberne. 2025. De-identifying clinical texts using biomed-clinical berts and comprehensive risk assessment. In 2025 IEEE 13th International Conference on Healthcare Informatics (ICHI), pages 683–684. IEEE. find
  20. Ashwin Machanavajjhala, Daniel Kifer, Johannes Gehrke, and Muthuramakrishnan Venkitasubramaniam. 2007. l-diversity: Privacy beyond k-anonymity. ACM Transactions on Knowledge Discovery from Data (TKDD), 1(1):3–es. find
  21. Louis Philippe Sondeck and Maryline Laurent. 2025. Practical and ready-to-use methodology to assess the re-identification risk in anonymized datasets. Scientific Reports, 15(1):23223. find
  22. Sara Mazzucato, Tom M Seinen, Sara Moccia, Silvestro Micera, Andrea Bandini, and Erik M van Mulligen. 2026. Advancements in multilingual biomedical natural language processing: exploring large language models for named entity recognition and linking. medRxiv, pages 2026– 01. find
  23. Amber Stubbs, Christopher Kotfila, and Özlem Uzuner. 2015. Automated systems for the de-identification of longitudinal clinical narratives: Overview of 2014 i2b2/uthealth shared task track 1. Journal of Biomedical Informatics, 58:S11– S19. find
  24. Stcaféphane M Meystre, Óscar Ferrcafández, F Jeffrey Friedlin, Brett R South, Shuying Shen, and Matthew H Samore. 2014. Text de-identification for privacy protection: A study of its impact on clinical text information content. Journal of Biomedical Informatics, 50:142–150. find
  25. Amber Stubbs and Özlem Uzuner. 2014. Annotation guidelines for de-identification of medical records. Version 1.0, i2b2/UTHealth. find
  26. Hemang Subramanian, Arijit Sengupta, and Yilin Xu. 2024. Patient health record protection beyond the health insurance portability and accountability act: mixed methods study. Journal of Medical Internet Research, 26:e59674. find
  27. Latanya Sweeney. 2002a. Achieving k-anonymity privacy protection using generalisation and suppression. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 10(05):571–588. find
  28. Latanya Sweeney. 2002b. k-anonymity: A model for protecting privacy. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 10(05):557–570. find
  29. Özlem Uzuner, Yuan Luo, and Peter Szolovits. 2007. Evaluating the state-of-the-art in automatic de-identification. Journal of the American Medical Informatics Association, 14(5):550–563. find
  30. Taher Vakili and Hercules Dalianis. 2022. Utility preservation of clinical text after de-identification. In Proceedings of the 21st Workshop on Biomedical Language Processing, pages 383–388, Dublin, Ireland. find
  31. Paul Voigt and Axel Von dem Bussche. 2017. The EU General Data Protection Regulation (GDPR). A Practical Guide, 1 edition. Springer International Publishing. find
  32. Xiaofeng Yang, Tian Lyu, Qing Li, Chih-Yin Lee, Jiang Bian, William R Hogan, and Yonghui Wu. 2019. A study of deep learning methods for de-identification of clinical notes in cross-institute settings. BMC Medical Informatics and Decision Making, 19(S5). find
  33. Wenguang Yu, Yu Weng, Ronghua Lin, and Yong Tang. 2022. Cosbert: A cosine-based siamese bert-networks using for semantic textual similarity. In CCF Conference on Computer Supported Cooperative Work and Social Computing, pages 376–389. Springer. 9. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/paul-2026-deid-clinic-risk-aware-pseudonymization/