The use of machine learning for investigating the role of plastic surgeons in anatomical injuries: A retrospective observational study

Abstract

While plastic surgeons have been historically indispensable in the reconstruction of posttraumatic defects, their role in trauma centers worldwide has not been clearly defined. Therefore, we aimed to investigate the contribution of plastic surgeons in trauma care using machine learning from an anatomic injury viewpoint. We conducted a retrospective study reviewing the data for all trauma patients of our hospital from March 2019 to February 2021. In total, 4809 patients were classified in duplicate according to the 17 trauma-related departments while conducting the initial treatment. We evaluated several covariates, including age, sex, cause of trauma, treatment outcomes, surgical data, and severity indices, such as the Injury Severity Score and Abbreviated Injury Scale (AIS). A random forest algorithm was used to rank the relevance of 17 trauma-related departments in each category for the AIS and outcomes. Additionally, t test and chi-square test were performed to compare two groups, which were based on whether the patients had received initial treatment in the trauma bay from the plastic surgery department (PS group) or not (non-PS group), in each AIS category. The department of PS was ranked first in the face and external categories after analyzing the relevance of the 17 trauma-related departments in six categories of AIS, through the random forest algorithm. Of the 1108 patients in the face category of AIS, the PS group was not correlated with all outcomes, except for the rate of discharge to home (P