Figure 1
From: Predicting the complexity and mortality of polytrauma patients with machine learning models
Workflow of data collection and processing. First, a total of 996 patients diagnosed with polytrauma before or upon admission were initially enrolled. Next, patients with ICU stays of more than 24 h and those with injuries involving two or more anatomical sites were included. Then, patients with severe missing data and obvious data errors were excluded, leaving 756 patients (674 survivors and 82 deceased) for model construction. All 756 patients were used to develop and validate the mortality risk model of polytrauma patients. In order to classify the complexity of polytrauma patients, we classified all 674 survived patients into mild, moderate and severe groups. Specifically, 375 polytrauma patients with ICU stays of 3 days or less were grouped into the mild group, while 129 patients with ICU stays of 8 days or more were grouped into the severe group. This resulted in a final dataset of 504 patients for the construction of polytrauma complexity model.
