Comparison of machine learning algorithms to SAPS II in predicting in-hospital mortality of fractures of the pelvis and acetabulum: analyzes based on MIMIC-III database

Fractures of pelvis and acetabulum are at high risk death worldwide. However, the capability mortality prediction by conventional system Simple Acute Physiologic Score (SAPS) II remains limited. Here, we hypothesized that use machine learning (ML) algorithms could provide better performance than SAPS for traumatic patients in intensive care unit (ICU). Customized ML models were built including support vector machine, decision tree, logistic regression random forest based on MIMIC-III clinical database. 307 enrolled with an ICD-9 diagnosis pelvic, acetabular or combined pelvic fractures. Feature expansion method was used experimental extending features to first 72 hours after ICU admission, compared traditional first-24-hours ones build respective control models. A comparison each model’s made through area under receiver-operating characteristic curve (AUROC). All outperformed (AUROC = 0.73), among which model had supreme 0.90). The results suggested aid injuries potentially decision-making orthopedics practitioners.