不同数据格式影响大模型药物重整效果，小模型用叙述文本，大模型用原始JSON。
Serialisation Strategy Matters: How FHIR Data Format Affects LLM Medication Reconciliation
- 比较四种FHIR数据格式对大模型药物重整的影响。
- 小模型（≤8B）用临床叙述文本可提升19点F1，大模型（70B）用原始JSON最优。
- 小模型易漏药，且无法处理复杂用药患者，适合医疗安全审计优化。
临床交接中的药物重整是高风险、易出错的任务。大型语言模型正被用于基于结构化FHIR病历辅助该任务，但一个基本且未被充分研究的变量是：在输入模型前，如何序列化FHIR数据。我们首次系统比较了四种序列化策略（原始JSON、Markdown表格、临床叙述、时间线）在五个开源模型（Phi-3.5-mini、Mistral-7B、BioMistral-7B、Llama-3.1-8B、Llama-3.3-70B）上的表现，使用200个合成患者数据，共4000次推理测试。结果表明，序列化策略对参数量≤8B的模型性能有显著影响：临床叙述相比原始JSON，在Mistral-7B上最高提升19点F1（r=0.617，p<10⁻¹⁰）。而在70B模型中，原始JSON达到最佳平均F1 0.9956。所有20种模型与策略组合中，平均精确率均高于召回率：遗漏药物是主要错误模式，模型更倾向于漏掉真实药物而非虚构。小模型在约7-10种并发用药时性能饱和，无法有效处理多药联用患者——这类人群正是最易出错的群体。即使经过领域预训练的BioMistral-7B（无指令微调），在所有条件下均输出无效结果，说明仅靠领域预训练不足以完成结构化提取。本研究为临床大模型部署提供实证格式建议：≤8B模型用临床叙述，≥70B模型用原始JSON。完整流程可在开放源码工具下复现，仅需AWS g6e.xlarge实例（NVIDIA L40S，48GB VRAM）。
原文摘要 · Abstract (English)
Medication reconciliation at clinical handoffs is a high-stakes, error-prone process. Large language models are increasingly proposed to assist with this task using FHIR-structured patient records, but a fundamental and largely unstudied variable is how the FHIR data is serialised before being passed to the model. We present the first systematic comparison of four FHIR serialisation strategies (Raw JSON, Markdown Table, Clinical Narrative, and Chronological Timeline) across five open-weight models (Phi-3.5-mini, Mistral-7B, BioMistral-7B, Llama-3.1-8B, Llama-3.3-70B) on a controlled benchmark of 200 synthetic patients, totalling 4,000 inference runs. We find that serialisation strategy has a large, statistically significant effect on performance for models up to 8B parameters: Clinical Narrative outperforms Raw JSON by up to 19 F1 points for Mistral-7B (r = 0.617, p < 10^{-10}). This advantage reverses at 70B, where Raw JSON achieves the best mean F1 of 0.9956. In all 20 model and strategy combinations, mean precision exceeds mean recall: omission is the dominant failure mode, with models more often missing an active medication than fabricating one, which changes how clinical safety auditing priorities should be set. Smaller models plateau at roughly 7-10 concurrent active medications, leaving polypharmacy patients, the patients most at risk from reconciliation errors, systematically underserved. BioMistral-7B, a domain-pretrained model without instruction tuning, produces zero usable output in all conditions, showing that domain pretraining alone is not sufficient for structured extraction. These results offer practical, evidence-based format recommendations for clinical LLM deployment: Clinical Narrative for models up to 8B, Raw JSON for 70B and above. The complete pipeline is reproducible on open-source tools running on an AWS g6e.xlarge instance (NVIDIA L40S, 48 GB VRAM).
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。