High computation costs and latency of large language models such as GPT-4 have limited their deployment in clinical settings. Small language models (SLMs) offer a cost-effective alternative, but their limited capacity requires biomedical domain adaptation, which remains challenging. An additional bottleneck is the unavailability and high sensitivity of clinical data. To address these challenges, we propose a novel framework for adapting SLMs into high-performing clinical models. We introduce the MediPhi collection of 3.8B-parameter SLMs developed with our novel framework: pre-instruction tuning of experts on relevant medical and clinical corpora (PMC, Medical Guideline, MedWiki, etc.), model merging, and clinical-tasks alignment. To cover most clinical tasks, we extended the CLUE benchmark to CLUE+, doubling its size. Our expert models deliver relative improvements on this benchmark over the base model without any task-specific fine-tuning: 64.3% on medical entities, 49.5% on radiology reports, and 44% on ICD-10 coding (outperforming GPT-4-0125 by 14%). We unify the expert models into MediPhi via model merging, preserving gains across benchmarks. Furthermore, we built the MediFlow collection, a synthetic dataset of 2.5 million high-quality instructions on 14 medical NLP tasks, 98 fine-grained document types, and JSON format support. Alignment of MediPhi using supervised fine-tuning and direct preference optimization achieves further gains of 18.9% on average.

Large language models (LLMs) such as GPT-4o and o1 have demonstrated strong performance on clinical natural language processing (NLP) tasks across multiple medical benchmarks. Nonetheless, two high-impact NLP tasks — structured tabular reporting from nurse dictations and medical order extraction from doctor-patient consultations — remain underexplored due to data scarcity and sensitivity, despite active industry efforts. Practical solutions to these real-world clinical tasks can significantly reduce the documentation burden on healthcare providers, allowing greater focus on patient care. In this paper, we investigate these two challenging tasks using private and open-source clinical datasets, evaluating the performance of both open- and closed-weight LLMs, and analyzing their respective strengths and limitations. Furthermore, we propose an agentic pipeline for generating realistic, non-sensitive nurse dictations, enabling structured extraction of clinical observations. To support further research in both areas, we release SYNUR and SIMORD, the first open-source datasets for nurse observation extraction and medical order extraction.

Several studies have shown that Large Language Models (LLMs) can answer medical questions correctly, even outperforming the average human score in some medical exams. However, to our knowledge, no study has been conducted to assess the ability of language models to validate existing or generated medical text for correctness and consistency. In this paper, we introduce MEDEC (https://github.com/abachaa/MEDEC), the first publicly available benchmark for medical error detection and correction in clinical notes, covering five types of errors (Diagnosis, Management, Treatment, Pharmacotherapy, and Causal Organism). MEDEC consists of 3,848 clinical texts, including 488 clinical notes from three US hospital systems that were not previously seen by any LLM. The dataset has been used in the MEDIQA-CORR 2024 shared task to evaluate seventeen participating systems. In this paper, we describe the data creation methods and we evaluate recent LLMs (e.g., o1-preview, GPT-4, Claude 3.5 Sonnet, Gemini 2.0 Flash, and DeepSeek-R1) for the tasks of detecting and correcting medical errors requiring both medical knowledge and reasoning capabilities. We also conducted a comparative study where two medical doctors performed the same task on the MEDEC test set. The results showed that MEDEC is a sufficiently challenging benchmark to assess the ability of models to validate existing or generated notes and to correct medical errors. We also found that although recent LLMs have a good performance in error detection and correction, they are still outperformed by medical doctors in these tasks. We discuss the potential factors behind this gap, the insights from our experiments, the limitations of current evaluation metrics, and share potential pointers for future research.

Medical doctors spend on average 52 to 102 minutes per day writing clinical notes from their patient encounters (Hripcsak et al., 2011). Reducing this workload calls for relevant and efficient summarization methods. In this paper, we introduce new resources and empirical investigations for the automatic summarization of doctor-patient conversations in a clinical setting. In particular, we introduce the MTS-Dialog dataset; a new collection of 1,700 doctor-patient dialogues and corresponding clinical notes. We use this new dataset to investigate the feasibility of this task and the relevance of existing language models, data augmentation, and guided summarization techniques. We compare standard evaluation metrics based on n-gram matching, contextual embeddings, and Fact Extraction to assess the accuracy and the factual consistency of the generated summaries. To ground these results, we perform an expert-based evaluation using relevant natural language generation criteria and task-specific criteria such as critical omissions, and study the correlation between the automatic metrics and expert judgments. To the best of our knowledge, this study is the first attempt to introduce an open dataset of doctor-patient conversations and clinical notes, with detailed automated and manual evaluations of clinical note generation.

Recent studies on automatic note generation have shown that doctors can save significant amounts of time when using automatic clinical note generation (Knoll et al., 2022). Summarization models have been used for this task to generate clinical notes as summaries of doctor-patient conversations (Krishna et al., 2021; Cai et al., 2022). However, assessing which model would best serve clinicians in their daily practice is still a challenging task due to the large set of possible correct summaries, and the potential limitations of automatic evaluation metrics. In this paper we study evaluation methods and metrics for the automatic generation of clinical notes from medical conversation. In particular, we propose new task-specific metrics and we compare them to SOTA evaluation metrics in text summarization and generation, including: (i) knowledge-graph embedding-based metrics, (ii) customized model-based metrics with domain-specific weights, (iii) domain-adapted/fine-tuned metrics, and (iv) ensemble metrics. To study the correlation between the automatic metrics and manual judgments, we evaluate automatic notes/summaries by comparing the system and reference facts and computing the factual correctness, and the hallucination and omission rates for critical medical facts. This study relied on seven datasets manually annotated by domain experts. Our experiments show that automatic evaluation metrics can have substantially different behaviors on different types of clinical notes datasets. However, the results highlight one stable subset of metrics as the most correlated with human judgments with a relevant aggregation of different evaluation criteria.

We introduce MedicalSum, a transformer-based sequence-to-sequence architecture for summarizing medical conversations by integrating medical domain knowledge from the Unified Medical Language System (UMLS). The novel knowledge augmentation is performed in three ways: (i) introducing a guidance signal that consists of the medical words in the input sequence, (ii) leveraging semantic type knowledge in UMLS to create clinically meaningful input embeddings, and (iii) making use of a novel weighted loss function that provides a stronger incentive for the model to correctly predict words with a medical meaning. By applying these three strategies, MedicalSum takes clinical knowledge into consideration during the summarization process and achieves state-of-the-art ROUGE score improvements of 0.8-2.1 points (including 6.2% ROUGE-1 error reduction in the PE section) when producing medical summaries of patient-doctor conversations.

We discuss automatic creation of medical reports from ASR-generated patient-doctor conversational transcripts using an end-to-end neural summarization approach. We explore both recurrent neural network (RNN) and Transformer-based sequence-to-sequence architectures for summarizing medical conversations. We have incorporated enhancements to these architectures, such as the pointer-generator network that facilitates copying parts of the conversations to the reports, and a hierarchical RNN encoder that makes RNN training three times faster with long inputs. A comparison of the relative improvements from the different model architectures over an oracle extractive baseline is provided on a dataset of 800k orthopedic encounters. Consistent with observations in literature for machine translation and related tasks, we find the Transformer models outperform RNN in accuracy, while taking less than half the time to train. Significantly large wins over a strong oracle baseline indicate that sequence-to-sequence modeling is a promising approach for automatic generation of medical reports, in the presence of data at scale.