[1H4-J-13] AI application: medicine

[1H4-J-13-05] Cluster Features and Transition Patterns in the Analysis Results of Sepsis-related DPC Data with Weighted PLSA Methods

〇Chisae Murata¹, Kazuya Yamashita¹, Yuichiro Sakamoto², Eiichi Sakurai^1,2, Yoichi Motomura^1,2 (1. National Institute of Advanced Industrial Science and Technology, 2. Saga University)

Keywords:AI, sepsis, diagnosis procedure combination

This study aimed to investigate factors related to mortality in patients with sepsis by analyzing all items of their diagnosis procedure combination (DPC) data with weighted probabilistic latent semantic analysis (PLSA) methods and to clarify features and transition patterns of the clusters extracted from the analysis. The results showed that 12 clusters were optimal and among those the one with high severity and the one receiving palliative care tended to have a high mortality rate. In addition, focusing on the main cluster for each patient, five cluster transition patterns were revealed.