Risk factors for mortality in emergency patients with dysglycemia based on Logistic regression

Zhang Lu; Zheng Lvmei; Xu Fanwen; Huang Shan; Wu Run; Zeng Jingni; Peng Ling; Xiao Shan; Huang Qiulan; Lin Ruixiang

doi:10.3969/j.issn.1002-1949.2026.02.004

Abstract：Objective This study aimed to identify risk factors for mortality in Emergency Department(ED)patients with dysglycemia and develop a predictive model to support clinical decision-making.Methods A single-center retrospective cohort study was conducted,enrolling 5 869 ED patients with dysglycemia from Bao'an District People's Hospital in Shenzhen between 2013 and 2023.Using visit duration(season/time slot),resuscitation room interventions,and clinical characteristics as independent variables(X),and clinical outcomes as dependent variables(Y).Multivariate Logistic regression was used to identify mortality risk factors,and model performance was evaluated using receiver operating characteristic(ROC)curves.Results Among enrolled patients,hyperglycemia accounted for 5 702 cases(97.15％)and hypoglycemia for 167 cases(2.85％),with 76 in-hospital deaths(mortality rate 1.29％).The median resuscitation room length of stay was 5.11 hours(IQR 3.00-8.00),showing significant variation across diseases(Kruskal-Wallis H=56.23,P＜0.001).The prediction model was validated using 10-fold cross-validation,yielding a mean AUC of 0.790[95％confidence interval(CI)0.616-0.958],with a sensitivity of 61.8％and a specificity of 76.0％.Multivariate Logistic regression analysis indicated that age progression(OR=0.97),admission to the emergency resuscitation room(OR=0.14),use of vasoactive drugs(OR=0.11),and sodium bicarbonate correction of acidosis(OR=0.28)were significantly associated with an increased risk of mortality.The predictive weights of variables on clinical outcomes,calculated via the model_10cv_LR(X,Y)model,were as follows:cardiopulmonary resuscitation therapy(-0.724),mechanical ventilation therapy(-0.921),administration of vasoactive drugs(-1.960),correction of acidosis therapy(-1.216),intravenous fluid therapy(-0.103),whether admitted to the resuscitation room(-1.434),time to medical attention within 24 hours(-0.033),age(per one-year increase)(-0.032);insulin therapy(0.443).Conclusions The predictive model integrates patient characteristics,seasonal patterns,temporal factors,treatment measures,and systemic factors,demonstrating strong discriminatory power.It provides a quantitative tool for early identification of high-risk patients and can be applied to construct mortality risk prediction models for precision treatment.