Exploration of Early Warning Models for Critical Risk in Emergency Patients

Objective To establish an early warning model for assessing the critical risk of emergency patients and evaluate its clinical benefits, providing a reference for the early identification of critically ill patients in emergency departments. Method The 3859 enrolled patients were randomly divided into a training set and a validation set in a 7:3 ratio. In the training set, a predictive model was established on the basis of the results of multivariate logistic stepwise regression analysis. Moreover, risk levels were divided, and the predictive efficacy and clinical benefits of the predictive model were verified. Results Multivariate logistic stepwise regression analysis revealed that sex, age, heart rate (HR), respiratory rate (R), systolic blood pressure (SBP), pulse oxygen (SPO2), consciousness, pupils, mental state, and pain score were independent risk factors for early assessment of critical risk (P < 0.05), and a predictive model was established on this basis. Using a conditional inference tree, critical risks are classified into low risk, medium risk, and high risk. Furthermore, the prediction model was internally validated in both the training and validation sets, with a training set area under the subject working characteristic curve (AUC) of 0.926 (95% confidence interval [CI] 0.912–0.939, P < 0.001) and a validation set AUC of 0.911 (95% CI 0.886–0.936, P < 0.001), indicating good discrimination. The calibration curve of the training set fits the standard curve, whereas the calibration curve of the validation set model slightly deviates from the standard curve, indicating good calibration of the predicted model. The decision curve analysis (DCA) and the clinical impact curve (CIC) suggest that both groups of patients can achieve good clinical effectiveness. Conclusion Establishing a predictive model for the early assessment of emergency critical risk is helpful for the early identification and intervention of emergency critical patients.