Latent class growth analysis of dynamic PaCo

To analyze dynamic patterns of arterial carbon dioxide partial pressure (PaCO₂) using latent class growth analysis in acute brain injury patients and investigate their associations with 28-day ICU mortality and 60-day in-hospital mortality. This retrospective study utilized the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. We applied latent class growth analysis to analyze PaCO₂ dynamic patterns during the first 72 h after ICU admission in adult patients with acute brain injury. Associations between trajectories and outcomes were evaluated using Cox proportional hazards models. Among the included acute brain injury patients (n = 1,145), three distinct PaCO₂ trajectories were identified: persistent hypocapnia pattern (23.7%), normal-mild regulation pattern (65.9%), and hypercapnia improvement pattern (10.5%). Cox proportional hazards regression analysis revealed that, compared with the normal-mild regulation pattern, the persistent hypocapnia pattern was significantly associated with higher risk of 28-day ICU mortality (HR = 1.28, 95% CI: 1.02-1.60) and 60-day in-hospital mortality (HR = 1.28, 95% CI: 1.03-1.59) after adjusting for confounding factors. The hypercapnia improvement pattern demonstrated a potentially protective association. This study identified distinct patterns of PaCO₂ dynamic changes in acute brain injury patients. Our analysis demonstrated that persistent hypocapnia was associated with higher risk of 28-day ICU and 60-day in-hospital mortality, while the hypercapnia improvement pattern showed potentially protective association. These results suggest the complexity of PaCO₂ management and the need for more individualized respiratory management strategies.