Objective: This study aimed to develop a decision tree-based model to identify prognostic patient profiles associated with discharge destination during intensive care unit admission.
Design: Historical cohort study using decision tree analysis.
Subjects/patients: 300 critically ill patients selected from 857 intensive care unit admissions after applying exclusion criteria.
Methods: The primary outcome was discharge destination (home vs non-home). Candidate variables included demographics, frailty, illness severity scores, delirium, ventilator duration, intensive care unit length of stay, and physical function on intensive care unit discharge. Terminal nodes were interpreted as prognostic profiles.
Results: Decision tree analysis identified profiles based on Acute Physiology and Chronic Health Evaluation II score, intensive care unit length of stay, Functional Status Score for the Intensive Care Unit, and age. A profile with lower initial severity and prolonged stay was associated with non-home discharge (area under the curve 0.76). Conversely, a profile with higher initial severity, preserved functional mobility (Functional Status Score for the Intensive Care Unit ≥ 5), and age < 80 years was associated with home discharge (area under the curve 0.77).
Conclusion: Combinations of illness severity, length of stay, functional status, and age-defined prognostic profiles were associated with discharge destination during intensive care unit admission.