LLMs for Nurses: Breaking Down Evidence Dissemination Barriers (2025)

Are nurses being left behind in the digital age? Evidence-based practice is crucial for improving patient care, but a significant barrier is the difficulty nurses face in accessing and implementing the latest research. This study dives deep into how Large Language Models (LLMs) – think advanced AI like ChatGPT – could revolutionize evidence dissemination in nursing. But here's the catch: current LLMs aren't quite cutting it.

Original Paper: Nurses’ Perspectives on Evidence Dissemination Barriers and Large Language Model–Based Support: Qualitative Study Using Focus Groups and Nominal Group Technique

Authors:

Junyi Ruan, MSN
Yimin Tang, MSN
Zhongyu Wei, PhD
Weijie Xing, PhD
Yan Hu, PhD

Affiliations:

School of Nursing, Fudan University, Shanghai, China
School of Data Science, Fudan University, Shanghai, China
Shanghai Innovation Institute, Shanghai, China
JBI Fudan University Centre for Evidence-Based Nursing, Shanghai, China

Corresponding Author: Weijie Xing, PhD, School of Nursing, Fudan University, Shanghai, China. Contact information provided in the original paper.

Background:

The study highlights a critical issue: despite the well-documented benefits of evidence-based practice (things like better patient outcomes and more efficient healthcare systems), its adoption in clinical nursing is far from ideal. A major contributing factor is the inadequacy of current methods for disseminating research findings to nurses. Many nurses, particularly in China, lack access to advanced education and specialized training in evidence-based knowledge. Time constraints, heavy workloads, and insufficient expert guidance further impede their ability to access and apply relevant research. While academic institutions are working to promote evidence dissemination, existing pathways are often ineffective, leading to delays in the uptake of new evidence and limiting nurses' awareness of the latest advancements. Therefore, strengthening evidence dissemination and supporting nurses in using the best available evidence are essential to advance evidence-based nursing practice.

Evidence dissemination is defined as the active and targeted distribution of research findings to promote their application in practice. It acts as a bridge between research and real-world clinical settings. Nurses are key players, acting as both recipients and transmitters of evidence. Effective dissemination not only facilitates nurses' access to, understanding of, and application of the best available evidence but also enables them to share this knowledge with their colleagues. Effective dissemination strategies accelerate the adoption of research findings, enhancing both accessibility and usability of evidence and promoting equity in evidence-based care delivery.

Existing dissemination strategies, ranging from printed materials to social media, often fall short of meeting the specific needs of healthcare professionals. Frontline nurses struggle to access up-to-date, high-quality evidence, especially when facing language barriers and limited access to database resources. Current dissemination pathways tend to be static and text-heavy, lacking the structured formats and efficient retrieval mechanisms necessary for supporting timely clinical decision-making. Furthermore, the lack of interactivity and intelligent recommendation systems hinders personalized evidence delivery. Delays in updating digital platforms compound the problem. Therefore, developing a more scientific and efficient dissemination method is crucial for shifting from static dissemination to dynamic interaction.

Enter Large Language Models (LLMs). These AI systems, with their advanced natural language processing capabilities, hold significant promise for revolutionizing evidence dissemination in healthcare. LLMs can assist healthcare professionals in moving from static internet searches to dynamic, AI-driven knowledge acquisition. In clinical nursing, LLMs are increasingly being used to support and optimize clinical decision-making, generate nursing care plans, and provide medical inquiries and intelligent question-answering. They offer a new pathway to evidence dissemination by enhancing human-computer interaction and improving efficiency, thereby facilitating the integration of evidence into routine clinical workflows. Their scalability and adaptability make them valuable for expanding access to high-quality evidence, particularly for nurses with limited support in evidence retrieval.

However, current general-purpose LLMs, such as ChatGPT and Gemini, have significant limitations. They lack domain-specific knowledge, exhibit poor alignment with users' actual needs, and often provide responses with relatively low accuracy. Given the low fault tolerance in healthcare, developing specialized LLMs tailored to the medical domain to enhance accuracy is essential. While research is focusing on developing medical LLMs by integrating medical knowledge into model training, most of these models are designed primarily for physicians and may not adequately address the unique needs of clinical nurses. Therefore, developing LLMs specifically tailored to nursing practice is crucial for improving patient care outcomes.

Objective:

This study aimed to understand the challenges and barriers clinical nurses encounter in disseminating evidence, their perspectives on using existing LLMs to support this process, and their needs and preferences for an LLM-based nursing evidence question-answering system.

Methods:

Researchers used a qualitative approach, combining focus group discussions and the Nominal Group Technique (NGT). They recruited 22 clinical nurses with diverse specialties, professional titles, and experience via purposive sampling. Two online focus groups were conducted via Tencent Meeting (a Chinese video conferencing platform) in late 2024 to explore the barriers to evidence dissemination and nurses' views on existing LLMs. The data were analyzed using qualitative content analysis. Then, the NGT was used in early 2025 to identify nurses' needs and preferences for a future LLM system. To accommodate geographical constraints and busy schedules, the NGT was conducted entirely online, using questionnaires and WeChat groups. Two rounds of voting were used to prioritize desired functionalities.