Predictors of 28-day mortality and development of a clinical risk model for immunoparalyzed patients with severe pneumonia surviving the initial 72-hour stabilization: a real-world ICU retrospective observational study

While ceftazidime-avibactam (CZA) is effective against multidrug-resistant bacteria, clinical survival in immunoparalyzed patients often remains suboptimal despite microbiological clearance. The specific drivers of mortality unique to this immune-exhausted state remain a significant knowledge gap. This study aimed to identify predictors of 28-day mortality and develop a clinical risk model for this high-risk population surviving the initial 72-hour stabilization phase of therapy. We conducted a retrospective analysis of ICU patients with severe pneumonia and immunoparalysis (defined as a peripheral CD4⁺ T-lymphocyte count < 200 /μL or a total lymphocyte count < 1.0×10⁹/L) treated with CZA between 2022 and 2025. To ensure meaningful assessment of therapeutic response, only patients surviving at least 72 hours of CZA therapy were included. Independent risk factors were identified using multivariable logistic regression. Model performance was evaluated via ROC curves and decision curve analysis (DCA). Of the 135 patients analyzed, the 28-day mortality rate was 45.2%. Multivariate analysis confirmed pulmonary aspergillosis co-infection (OR=16.05, P < 0.001) and elevated SOFA scores (OR=1.71, P < 0.001) as independent predictors of death. The predictive model demonstrated robust discrimination (AUC=0.847) and good calibration. In immunoparalyzed patients who survive the initial stabilization phase of therapy, 28-day mortality is independently associated with fungal co-infection and organ failure, rather than bacterial burden alone. Our model provides a tool for early risk stratification, suggesting that survival may require strategies beyond antibacterial therapy alone.