Predicting 28‑day mortality in critically ill patients: A combined model using APACHE II, lactate clearance, and Pcv-aCO₂

Early identification of prognostic factors in critically ill patients is essential for optimizing ICU management. Although the APACHE II score, lactate clearance rate (LCR), and central venous-to-arterial carbon dioxide partial pressure difference (Pcv-aCO₂) are each associated with outcomes, their combined predictive value remains unclear. This study evaluated their individual and combined performance in predicting 28-day mortality.

We retrospectively analyzed ICU patients admitted to Shaoxing Second Hospital from January 2021 to December 2023. Among 1188 screened patients, 634 were included. Clinical data, APACHE II score, initial Pcv-aCO₂, and 24-hour LCR were collected. Patients were classified as survivors or non-survivors at 28 days. Logistic regression identified independent predictors, and receiver operating characteristic curves assessed predictive performance. The predictive performance of individual indicators and their combined models was evaluated using the area under the curve (AUC), sensitivity, and specificity.

The 28-day mortality rate was 26.8%. Independent predictors included APACHE II score (OR=1.069), 24 h LCR (OR=0.306), Pcv-aCO₂ (OR=1.096), vasoactive drug use (OR=2.201), coagulation dysfunction (OR=1.781), and multiple organ dysfunction (OR=1.906) (all p< 0.05). Among single indicators, 24 h LCR showed the highest accuracy (AUC=0.909), followed by APACHE II score (AUC= 0.753) and Pcv-aCO₂ (AUC=0.739). The combination of 24 h LCR and APACHE II score achieved an AUC of 0.963, while adding Pcv-aCO₂ did not further increase the AUC but improved sensitivity and specificity.

APACHE II score, 24 h LCR, and Pcv-aCO₂ are independent predictors of short-term mortality. Combined assessment improves prognostic accuracy, supporting integrated evaluation for risk stratification in critically ill patients.