Clinical Phenotypes and Acute Brain Dysfunction Trajectories Among Critically Ill Patients With Delirium: A Real‐World Longitudinal Cohort Study

Background Intensive care unit (ICU) delirium is a heterogeneous and fluctuating form of acute brain dysfunction. Binary classification may obscure underlying mechanisms and temporal patterns. Aim To identify clinical phenotypes and acute brain dysfunction trajectories among critically ill patients with delirium and examine their associations with clinical course indicators. Study Design This real‐world, longitudinal, electronic health record‐based retrospective cohort study included adults admitted to four ICUs in China from 1 January 2023 to 31 December 2025. Delirium was assessed three times daily and classified into four non‐mutually exclusive phenotypes: hypoxic, septic, sedative‐associated and metabolic. Longitudinal latent class analysis modelled normal brain function, delirium and coma across ICU Days 1–7. Multivariable models examined phenotype burden and trajectory class. Results Among 4284 eligible ICU patients, delirium was identified in 767 patients during the ICU stay. The most common phenotypes were sedative‐associated delirium (51.8%) and septic delirium (50.7%), with substantial overlap. Four trajectories were identified: early improvement (22.8%), persistent delirium‐fluctuating (68.4%), coma‐delirium transition (5.6%) and progressive coma‐dominant acute brain dysfunction (3.1%). Each additional phenotype was associated with longer ICU stay (adjusted geometric mean ratio [aGMR] 1.16, 95% confidence interval [CI] 1.09–1.23), hospital stay (aGMR 1.11, 95% CI 1.05–1.18) and mechanical ventilation duration (aGMR 1.57, 95% CI 1.37–1.79), but not with in‐hospital mortality (adjusted odds ratio 1.40, 95% CI 0.92–2.13). Coma‐related trajectories were associated with longer ICU stay and mechanical ventilation. Conclusion Phenotype‐trajectory profiling captured heterogeneity in ICU delirium and provides a descriptive framework for characterising its mechanisms and longitudinal course. Relevance to Clinical Practice Integrating overlapping phenotypes with longitudinal brain dysfunction patterns may help ICU nurses structure the assessment of reversible contributors and identify patients requiring closer monitoring.