Detecting effective categories of medical incident reports for patient safety management

The analysis of medical incident reports is indispensable for patient safety management. The cycles between analysis of incident reports and proposals to medical staffs are a key point for improving the patient safety management in the hospital. Most of the incident reports include free descriptions, however, the analysis of free descriptions aren't enough in the medical area. We aimed to accumulate, to interpret information again by structured incident information, and to clarify the point that should be improved for the cause of the accident and safe medical treatment improvements in the present study. We employ the natural language processing and the network analysis for detecting effective categories of Medical Incident Report. The network analysis can find various relationships that are not only direct relationships but also indirect relationships. First, some important characteristic words were extracted in three categories of the accident's background, details, and solutions using TF-IDF measure. Next, we show the co-occurrence networks using the extracted words. Then, we detect the new categories based on the network analysis and compare between existing categories based on experts' decisions and bottom-up ones. By the network analysis, some of new perceptions for improving the patient safety management are appeared.