Background: Early and accessible screening of patients with polytrauma at a high risk of hospital death is essential. The purpose of this research was to seek an accurate and convenient solution to predict deaths occurring within 72 h after admission of these patients.
Methods: A secondary analysis was conducted on 3,075 patients with polytrauma from the Dryad database. We imputed missing values in eligible individuals with the k-nearest neighbor algorithm and then randomly stratified them into the training group (n = 2,461) and the validation group (n = 614) based on a proportion of 8:2. The restricted cubic spline, univariate, backward stepwise, and multivariate logistic regression methods were employed to determine the suitable predictors. Calibration and receiver operating characteristic (ROC) curves were applied to assess the calibration and discrimination of the obtained model. The decision curve analysis was then chosen as the measure to examine the clinical usage.
Results: Age, the Glasgow Coma Scale score, the Injury Severity Score, base excess, and the initial lactate level were inferred as independent prognostic factors related to mortality. These factors were then integrated and applied to construct a model. The performance of calibration plots, ROC curves, and decision curve analysis indicated that the model had satisfactory predictive power for 72-h mortality after admission of patients with polytrauma. Moreover, we developed a nomogram for visualization and a web-based calculator for convenient application (https://songandwen.shinyapps.io/DynNomapp/).
Conclusions: A convenient web-based calculator was constructed to robustly estimate the risk of death in patients with polytrauma within 72 h after admission, which may aid in further rationalization of clinical decision-making and accurate individual treatment.
Keywords: Glasgow Coma Scale; base excess; injury severity score; lactate; mortality; nomogram; polytrauma.
Copyright © 2022 Chen, Liu, Feng, Lv and Wei.