Abstract
BACKGROUND
The present study considers patients with spontaneous intracerebral hemorrhage (ICH) admitted to the neurocritical care unit (NCCU) through the Emergency Department (ED). It aims to identify patient-specific clinical variables that can be assessed on presentation and that are associated with prolonged NCCU length of stay (LOS).
METHODS
A cross-sectional, single-center, retrospective analysis of ICH patients directly admitted from the ED to the NCCU over an 8-year period was performed. Patients' demographics, clinical exam characteristics, serum laboratory values, intubation status, and neurosurgical procedures at presentation were recorded. Head computed tomography scans obtained on presentation were reviewed. LOS was calculated based on the number of midnights spent in the NCCU. Prolonged LOS was determined using a change point analysis, adopting the method of Taylor which utilizes CUMSUM charts and bootstrap analysis. A decision tree model was trained and validated to identify reliable variables associated with prolonged LOS.
RESULTS
Two hundred and five patients with ICH were analyzed. Prolonged LOS was calculated to be a stay that exceeds 8 days; 68 patients (33%) had a prolonged LOS in NCCU. Median LOS did not differ between survivors and patients who died in hospital. Clinical variables explored through the decision tree model were intubation status, neurosurgical intervention (EVD, decompression or evacuation within 24 h from presentation), and components of the ICH score: age, GCS, hematoma volume, the presence of intraventricular hemorrhage (IVH), and infratentorial location. The model accuracy was 0.8 and AUC was 0.83 (95% CI 0.78-0.89).
CONCLUSION
We propose an ICH-LOS model based on neurosurgical intervention, intubation status and GCS at presentation to predict prolonged LOS in the NCCU in patients with ICH. This simple clinical tool, if prospectively validated, could help with medical planning, contribute to patient care-directed conversations, assist in optimizing hospital resource utilization, and, more importantly, motivating patient-specific interventions aimed at optimizing outcomes and decreasing LOS.
Collapse