The purpose of this study is to visualize possible factors in healthcare incidents by analyzing the text in incident reports employing the Self-Organizing Map method.
An ordinal method of analyzing incident reports in a hospital is to count the number of factors that are predefined, or to investigate possible factors by focusing on individual cases. Incident reports include text data, which presented the situation under which the incident occurred, although we have yet to utilize these data efficiently. Specifically, the method to analyze large volumes of text documents in medical incident reports has yet to be established.
This study conducted an analysis with 4,505 incident reports obtained from April 1, 2020, to March 31, 2021,in Hospital A. We applied the Self-Organizing Map method to the text documents in the reports.
The results showed that the following seven groups were distinguished: medicine (Prescription), medicine (Internal [oral] medicine, Intravenous drip), Self-removal of the stomach tube, Self-removal of the tube for intravenous drip, Blood test, Incomplete consent form, and Medicines remaining in pockets of staff uniform.
Meanwhile, patient falls have not appeared on the Self-Organizing Map despite an increasing number of such incidents.
In conclusion, the Self-Organizing Map method would be effective to determine a new tendency, which may not be extracted with other ordinal analyses. This is considering that the Self-Organizing Map method can categorize similar or dissimilar patterns.
The results of the study suggest that the Self-Organizing Map method is capable of demonstrating objective factors from a significantly large number of incidents.
View full abstract