Current applications of large language models in clinical practice and needs assessment for cardiovascular physicians
10.3760/cma.j.cn112148-20250220-00132
- VernacularTitle:大模型在临床医师中的应用现状及心血管医师的需求分析
- Author:
Wenyu WANG
1
;
Zhixian WANG
;
Yize ZHAO
;
Lixin TIAN
;
Liu HE
;
Changsheng MA
Author Information
1. 首都医科大学附属北京安贞医院心脏内科中心，北京 100029
- Publication Type:Journal Article
- Keywords:
Artificial intelligence;
Large language models;
Awareness;
Application requirements;
Clinical decision-making
- From:
Chinese Journal of Cardiology
2025;53(6):644-652
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the current awareness of large language models (LLM) among Chinese clinical physicians and analyze the application needs of cardiovascular specialists.Methods:This is a cross-sectional study utilized convenience sampling. In December 2023, a self-designed questionnaire was distributed to 7 980 clinical physicians, including 930 cardiologists. The survey collected demographic information, including work city (categorized as first-tier, new first-tier, second-tier, third-tier, and fourth-tier and below), hospital level, professional title, and department. And the awareness of LLM, and their application demands in clinical decision-making support, information filtering, and scientific research work were also collected. Differences in awareness and application requirements across geographic regions, hospital tiers, professional ranks, and medical departments were analyzed. Besides, specific demands of cardiovascular specialists were further examined.Results:Among the 7 980 clinical physicians, the awareness rate of LLM was 76.3% (6 088/7 980), and the utilization rate was 11.8% (942/7 980). For the 930 cardiologists, the awareness rate was 78.5% (730/930) and the utilization rate was 11.4% (106/930). Significant differences in awareness and utilization rates were observed across city tiers, hospital grades, and departments (all P<0.05). No significant difference was found among professional titles ( P=0.053). Among the 6 088 physicians aware of LLM, demand rates for clinical information filtering, clinical decision support, and research assistance were 87.3% (5 312/6 088), 78.4% (4 774/6 088), and 75.8% (4 616/6 088), respectively. For the 730 cardiologists aware of LLM, these rates were 91.0% (664/730), 79.2% (578/730), and 75.9% (554/730), respectively. Significant differences in demands for clinical information filtering and research assistance were observed across city tiers, hospital grades, professional titles, and departments (all P<0.05), while no significant difference was noted for decision support demands across hospital grades ( P=0.085). In clinical information screening and acquisition, cardiologists from different city tiers exhibited statistically significant differences in the demand for literature interpretation. Similarly, variations in the demand for conference summaries, expert biographies, healthcare policies, and social news were noted among cardiologists with different professional titles, while disparities in patient education and science popularization needs were identified across city tiers and hospital grades (all P<0.05). In clinical decision-making support, cardiologists from diverse city tiers and professional titles demonstrated distinct differences in guideline and consensus inquiries, and those from various city tiers showed varied demands for pharmaceutical and medical device-related content (all P<0.05). For research support, cardiologists across city tiers and professional titles exhibited statistically significant differences in trial protocol design requirements, while those from varying city tiers differed in literature search/analysis and research application procedures. Additionally, physicians from different hospital grades displayed divergent needs for data collection (all P<0.05). Conclusions:The adoption of LLM is significantly influenced by regional disparities, institutional resources, and professional backgrounds. Implementing targeted interventions, such as enhancing technical training, optimizing LLM functionalities, and improving accessibility across diverse healthcare settings, could encourage widespread integration of LLM into clinical practice. Such measures could ultimately enhance the quality and efficiency of medical services in China and foster innovations in healthcare delivery.
