Topic Modeling of Nursing Records for Analyzing Patient Dissatisfaction in the Emergency Department
10.22650/JKCNR.2025.31.3.292
- Author:
Hyung Bok LEE
1
;
Min Jin CHOI
;
Ji Won KIM
Author Information
1. Department of Nursing, Seoul National University Hospital
- From:
Journal of Korean Clinical Nursing Research
2025;31(3):292-304
- CountryRepublic of Korea
- Language:English
-
Abstract:
Purpose:This study aimed to identify and prevent emergency department patient dissatisfaction. To achieve this purpose, we extracted unstructured nursing records related to patient dissatisfaction and applied Natural Language Processing (NLP) and Latent Dirichlet Allocation (LDA) topic modeling to identify key topics and associated keywords.
Methods:Unstructured nursing records from dissatisfied patients who visited a tertiary hospital ED between 2016 and 2022 were analyzed. NLP techniques were used to extract the top 30 keywords based on term frequency (TF) and term frequency-inverse document frequency (TF-IDF). LDA was applied to uncover latent themes, and an intertopic distance map was generated to visualize the relationships among the topics.
Results:The most frequent keywords were 'explanation' (1,083), followed by 'patient' (1,078), 'examination' (837), 'waiting' (752), and 'caregiver' (731). The study also found significant verbal assault words, including 'yelling' (496), 'verbal abuse' (135), and 'anger' (114). The LDA approach identified four topics:dissatisfaction with ED examinations, procedures, results, and physicians' care by topic modeling.
Conclusion:The findings indicate that ED patient dissatisfaction commonly begins at arrival, largely due to insufficient explanations. These situations are often accompanied by verbal aggression.
