Prediction Model for Skin Damage Related to Medical Adhesives in Adult Patients Undergoing Degenerative Spine Surgery

ABSTRACT

This study aimed to develop a prediction model for the occurrence of medical adhesive-related skin injuries (MARSIs) based on electronic medical records (EMRs) of adult patients who underwent degenerative spine surgery. This study used the EMR data of adult patients who underwent degenerative spine surgery at a university hospital in Seoul between January 2020 and December 2024. Seven machine learning algorithms and the SuperLearner algorithm were used to evaluate the performance of the SuperLearner model. Performance was focused on the area under the curve (AUC), accuracy, sensitivity, specificity, precision and F1 score. Among the machine learning algorithms, the RuleFit algorithm showed the best performance, with an AUC of 0.723, accuracy of 0.689, sensitivity of 0.959, specificity of 0.276, precision of 0.762 and F1 score of 0.789. In contrast, predicting MARSI using the SuperLearner algorithm had an AUC of 0.951, accuracy of 0.834, sensitivity of 0.635, specificity of 0.964, precision of 0.921 and F1 score of 0.752. This study provides practical evidence for the early identification of high-risk patients and establishment of customized nursing plans by presenting a MARSI prediction model using the SuperLearner ensemble. Future research is recommended to verify the external validity of the model through prospective studies and integration of clinical decision support systems.

Trial Registration: ClinicalTrials.gov Identifier KCT0010601.