Prognostic accuracy of CURB-65, serum lactate, and a combined model for predicting intensive care unit admission and in-hospital mortality in emergency department patients with pneumonia: a retrospective cohort study

Background: Community-acquired pneumonia is a leading cause of emergency department (ED) presentation, hospitalization, and in-hospital mortality. CURB-65 is the most widely used bedside risk stratification tool; however, it does not capture tissue hypoperfusion. We evaluated the combined discriminative performance of CURB-65 and admission serum lactate for predicting intensive care unit (ICU) admission and in-hospital mortality. Methods: This was a retrospective, single-center cohort study conducted at a university hospital ED (January 2023–July 2025). After excluding patients with missing lactate (n = 204), incomplete CURB-65 components (n = 307), or other reasons (n = 36), a total of 317 out of 864 consecutive patients with pneumonia were included. Predictors were assessed using binary logistic regression; the area under the curve (AUC) were compared using the DeLong method. Calibration was assessed with the Hosmer–Lemeshow test; internal validation used bootstrap resampling (1000 iterations). Results: Median age was 70 years (interquartile range: 57–80); 47.9% were female. ICU admission occurred in 60 patients (18.9%), and in-hospital mortality occurred in 25 (7.9%). Both CURB-65 (odds ratio: 2.48; 95% confidence interval: 1.75–3.50; P < 0.001) and lactate (odds ratio: 1.40; 95% confidence interval: 1.23–1.59; P < 0.001) were independently associated with ICU admission. For ICU admission, AUCs were 0.778 (CURB-65), 0.796 (lactate), and 0.850 (combined; ΔAUC: 0.071, P < 0.001). For in-hospital mortality, AUCs were 0.763, 0.806, and 0.858 (ΔAUC: 0.094, P = 0.003). The combined model showed excellent calibration (Hosmer–Lemeshow P > 0.90) and high negative predictive value (91.1% for ICU; 98.0% for mortality). Conclusion: Adding serum lactate to CURB-65 significantly improved discriminative performance for ICU admission and in-hospital mortality in ED pneumonia patients. The high negative predictive value supports its utility for ruling out adverse outcomes in low-risk patients.