Combined lactate, base excess, and MEWS score as predictors of ICU transfer from the emergency department: a retrospective cohort study

Objective Accurate risk stratification of emergency department (ED) patients transferred to the intensive care unit (ICU) is essential for improving outcomes. The traditional Modified Early Warning Score (MEWS) alone has limited utility in predicting mortality risk. Combining MEWS with arterial blood gas parameters—lactate (Lac) and base excess (BE)—may enhance predictive performance. Methods A retrospective cohort study was conducted involving 262 ED patients admitted to the ICU, who were divided into a non-survivor group ( n  = 83, 28-day mortality rate: 31.68%) and a survivor group ( n  = 179). Multivariable Cox regression analysis was used to identify independent predictors of 28-day mortality. Model performance was evaluated using receiver operating characteristic (ROC) curve analysis, DeLong test, net reclassification improvement (NRI)/integrated discrimination improvement (IDI) metrics, and stratified validation. Results Multivariate Cox regression analysis showed that both Lac were independent risk factors for 28-day mortality (both p  < 0.05). ROC curve analysis revealed that the Lac + BE + MEWS combined model achieved the highest area under the curve (AUC) of 0.819 (95% CI: 0.760–0.870), which was significantly superior to other models. The BE + MEWS model yielded an AUC of 0.805 (95% CI: 0.749–0.864), with no statistically significant difference from the Lac + BE + MEWS model ( p  = 0.098). NRI and IDI analyses indicated that both the BE + MEWS and Lac + BE + MEWS models markedly improved predictive performance (NRI: 85.9 and 83.1%, respectively; IDI: 0.249 and 0.255, respectively; both p  < 0.001). Notably, BE conferred greater incremental value to MEWS than Lac. Stratified validation confirmed that the Lac + BE + MEWS model exhibited the best stability and risk-stratification capacity, especially in the intermediate-risk stratum (5 ≤ MEWS ≤ 8, AUC = 0.812). Conclusion Lac and BE are independent predictors of 28-day mortality in ED patients admitted to the ICU. BE adds significantly more predictive value to MEWS than Lac. The Lac + BE + MEWS combined model demonstrates the strongest stability and optimal risk-stratification performance, particularly for patients with MEWS scores of 5–8.