A Deep Learning-Based Text Classification of Adverse Nursing Events

Abstract

Adverse nursing events occur suddenly, unpredictably, or unexpectedly during course of clinical diagnosis and treatment processes in the hospitals. These events adversely affect the patient's diagnosis and treatment results and even increase the patient's pain and burden. Additionally, It is high likely to cause accidents and disputes and affect normal medical work and personnel safety and is not conducive to the development of the health system. Due to the rapid development of modern medicine, health and safety of patients have become the most concerned issue in society and patient safety is an important part of medical care management. Research and events have shown that classified management of adverse nursing events, event analysis, and improvement measures are beneficial, specifically to the health system, to continuously improve the quality of medical care and reduce the occurrence of adverse nursing events. In the management of adverse nursing events, it is very important to categorize the text reports of adverse nursing events and divide these into different categories and levels. Traditional reports of adverse nursing events are mostly unstructured and simple data, often relying on manual classification, which is difficult to analyze. Furthermore, data is relatively inaccurate and practical reference significance is not obvious. In this paper, we have extensively evaluated various deep learning-based classification methods which are specifically designed for the healthcare systems. It becomes possible with the development of science and technology; text classification methods based on deep learning are gradually entering people's field of vision. Additionally, we have proposed a text classification model for adverse nursing events in the health system. Experiments and data comparison test of both the proposed deep learning-based method and existing methods in the text classification of nursing adverse events effect are performed. These results show the exceptional performance of the proposed mechanism in terms of various evaluation metrics.