# Synthetic Data for Veterinary EHR De-identification: Benefits, Limits, and Safety Trade-offs Under Fixed Compute

arXiv: https://arxiv.org/abs/2601.09756 · HTML: https://mlsanity.com/paper/2601.09756/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Medical synthetic data and agents, one of 13 directions in Synthetic data and self-improvement for LLMs.
- Builds on 17 of its 33 references that are in the map.
- First posted on arXiv 13 Jan 2026.

## Placed in

- Synthetic data and self-improvement for LLMs / [Medical synthetic data and agents](https://mlsanity.com/synth/medical-synthetic-data-and-agents/)

## Builds on (most cited first, 17 of 17)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [BioBERT: a pre-trained biomedical language representation model for biomedical text mining](https://mlsanity.com/paper/1901.08746/) (2019)
- [Domain-Specific Language Model Pretraining for Biomedical Natural Language Processing](https://mlsanity.com/paper/2007.15779/) (2020)
- [Publicly Available Clinical BERT Embeddings](https://mlsanity.com/paper/1904.03323/) (2019)
- [De-identification of Patient Notes with Recurrent Neural Networks](https://mlsanity.com/paper/1606.03475/) (2016)
- [Two Directions for Clinical Data Generation with Large Language Models: Data-to-Label and Label-to-Data](https://mlsanity.com/paper/2401.06774/) (2023)
- [Training and Evaluating Language Models with Template-based Data Generation](https://mlsanity.com/paper/2411.18104/) (2024)
- [De-identification of medical records using conditional random fields and long short-term memory networks](https://mlsanity.com/paper/1709.06901/) (2017)
- [A Comparative Evaluation Of Transformer Models For De-Identification Of Clinical Text Data](https://mlsanity.com/paper/2204.07056/) (2022)
- [How do we measure privacy in text? A survey of text anonymization metrics](https://mlsanity.com/paper/2512.01109/) (2025)
- [Validating transformers for redaction of text from electronic health records in real-world healthcare](https://mlsanity.com/paper/2310.04468/) (2023)
- [Generation of Synthetic Clinical Text: A Systematic Review](https://mlsanity.com/paper/2507.18451/) (2025)
- [Disease Entity Recognition and Normalization is Improved with Large Language Model Derived Synthetic Normalized Mentions](https://mlsanity.com/paper/2410.07951/) (2024)
- [Fine-tuning foundational models to code diagnoses from veterinary health records](https://mlsanity.com/paper/2410.15186/) (2024)
- [Evaluation of Language Models in the Medical Context Under Resource-Constrained Settings](https://mlsanity.com/paper/2406.16611/) (2024)
- [Generating Synthetic Free-text Medical Records with Low Re-identification Risk using Masked Language Modeling](https://mlsanity.com/paper/2409.09831/) (2024)
- [Data Augmentation Method Utilizing Template Sentences for Variable Definition Extraction](https://mlsanity.com/paper/2405.14962/) (2024)

## Abstract (arXiv)

Veterinary electronic health records (vEHRs) contain privacy-sensitive identifiers that limit secondary use. While PetEVAL provides a benchmark for veterinary de-identification, the domain remains low-resource. This study evaluates whether large language model (LLM)-generated synthetic narratives improve de-identification safety under distinct training regimes, emphasizing (i) synthetic augmentation and (ii) fixed-budget substitution. We conducted a controlled simulation using a PetEVAL-derived corpus (3,750 holdout/1,249 train). We generated 10,382 synthetic notes using a privacy-preserving "template-only" regime where identifiers were removed prior to LLM prompting. Three transformer backbones (PetBERT, VetBERT, Bio_ClinicalBERT) were trained under varying mixtures. Evaluation prioritized document-level leakage rate (the fraction of documents with at least one missed identifier) as the primary safety outcome. Results show that under fixed-sample substitution, replacing real notes with synthetic ones monotonically increased leakage, indicating synthetic data cannot safely replace real supervision. Under compute-matched training, moderate synthetic mixing matched real-only performance, but high synthetic dominance degraded utility. Conversely, epoch-scaled augmentation improved performance: PetBERT span-overlap F1 increased from 0.831 to 0.850 +/- 0.014, and leakage decreased from 6.32% to 4.02% +/- 0.19%. However, these gains largely reflect increased training exposure rather than intrinsic synthetic data quality. Corpus diagnostics revealed systematic synthetic-real mismatches in note length and label distribution that align with persistent leakage. We conclude that synthetic augmentation is effective for expanding exposure but is complementary, not substitutive, for safety-critical veterinary de-identification.