Role of Large Language Models in Patient Safety in an Expert Evaluation Study
DOI:
https://doi.org/10.59681/2175-4411.v18.2026.1570Keywords:
Clinical Decision Support Systems, Artificial Intelligence, Large Language ModelsAbstract
Objective: This study comparatively evaluates the performance of different Large Language Models (LLMs) as tools to support medical prescription, considering criteria associated with patient safety and clinical applicability, based on a blinded expert evaluation. Methods: Six LLMs answered four questions about contraindications, drug interactions, and dosages. A panel of 34 physicians blindly evaluated 24 responses based on five criteria: consistency, focus, coherence, completeness, and detail. Results: Performance varied according to the criteria and question types; LLM6 showed better completeness and detail, especially in complex cases. Simple questions, such as contraindications, received higher scores, while complex questions showed greater variation. Conclusion: The findings indicate that the use of LLMs to support medical prescription requires careful model selection and consideration of the level of clinical complexity, reinforcing the need for contextual validation before their adoption in healthcare settings.
Downloads
References
Tariq RA, Vashisht R, Sinha A. Medication dispensing errors and prevention. In: StatPearls. Treasure Island (FL): StatPearls Publishing; [cited 2025 May 14]. Available from: https://www.ncbi.nlm.nih.gov/books/NBK519065
National Coordinating Council for Medication Error Reporting and Prevention. About medication errors: what is a medication error? [cited 2025 Feb 2]. Available from: https://www.nccmerp.org/about-medication-errors
Cruzeta APS, Dourado ACL, Monteiro MTM, Martins RO, Calegario TA, Galato D. Fatores associados à compreensão da prescrição médica no Sistema Único de Saúde de um município do Sul do Brasil. Cien Saude Colet. 2013;18(12):3731-7. doi: https://doi.org/10.1590/S1413-81232013001200029
World Health Organization. Medication without harm. 2025 [cited 2025 Feb 3]. Available from: https://www.who.int/initiatives/medication-without-harm
Academy of Managed Care Pharmacy. Medication errors. 2025 [cited 2025 Feb 3]. Available from: https://www.amcp.org/concepts-managed-care-pharmacy/medication-errors
Cohen MR, Smetzer JL. ISMP medication error report analysis. Hosp Pharm. 2017;52(6):390-3. doi: https://doi.org/10.1177/0018578717715346
Shah K, Xu AY, Sharma Y, Daher M, McDonald C, Diebo BG, et al. Large language model prompting techniques for advancement in clinical medicine. J Clin Med. 2024;13(17):5101. doi: https://doi.org/10.3390/jcm13175101
Netto AV, Berton L, Takahata AK. Ciência de dados e a inteligência artificial na área da saúde. São Paulo: Editora dos Editores; 2021.
Netto AV. Ciência de dados em saúde: contribuições e tendências para aplicações. Rev Saúde.Com. 2021;17(3):1-5. doi: https://doi.org/10.22481/rsc.v17i3.6290
The Healthcare Technology Report. Software & Data: Oxford’s DrugGPT AI tool enhances medication prescriptions. 2024 [cited 2025 Feb 4]. Available from: https://thehealthcaretechnologyreport.com/oxfords-druggpt-ai-tool-enhances-medication-prescriptions
Medeiros CH, Kauark FD, Manhães FC. Metodologia da pesquisa: guia prático. Itabuna: Via Litterarum; 2010.
OpenAI. Hello GPT-4o. 2024 [cited 2025 Feb 8]. Available from: https://openai.com/index/hello-gpt-4o
Cortex. Your exclusive corporate AI. 2025 [cited 2025 Feb 8]. Available from: https://sinapse.tech/cortex
Anthropic. Introducing Claude 3.5 Sonnet. 2024 [cited 2025 Feb 8]. Available from: https://www.anthropic.com/news/claude-3-5-sonnet
Meta. Introducing Llama 3.1: our most capable models to date. 2024 [cited 2025 Feb 8]. Available from: https://ai.meta.com/blog/meta-llama-3-1
Cohere. The all-in-one platform for private and secure AI. 2024 [cited 2025 Feb 8]. Available from: https://cohere.com/
Google. Gemini models. 2025 [cited 2025 Feb 8]. Available from: https://ai.google.dev/gemini-api/docs/models/gemini
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Antonio Valerio Netto, Camila de Brito Pontes

This work is licensed under a Creative Commons Attribution 4.0 International License.