Abstract 126: Derivation and Validation of a Proposed Long Length of Stay (≥ 7 days) Score in Patients Hospitalized for Acute Ischemic Stroke
Introduction: Long length of stay (LLOS) is one of the main factors in the determination of high cost in hospitalized stroke patients. Our aim was to prospectively predict patients more likely to have a long length of stay using our hospitals Get with the guidelines (GWTG) ischemic stroke registry. Methods: We selected 5,400 patients from our database of which 3,400 (~70%) were used for the derivation cohort and 2,000 (~30%) were used for internal validation. For external validation, 730 patients were included from the University of Arkansas. Long length of stay was defined ≥ 7 days. A predictive score was developed using stepwise logistic regression, and its performance assessed using ROC curve analysis. Result: Patients with LLOS in the derivation cohort were more likely to female, self-pay, more often have diabetes mellitus, atrial fibrillation, heart failure, previous stroke and carotid stenosis, and more often presented with weakness. They were more likely to have received IV or IA thrombolysis and early antithrombotics, and had higher rates of pneumonia (18.8% vs. 2.6%) and UTI (16.7% vs. 5.3%). Independent predictors of LLOS were Medicare/Medicaid insurance, self pay, history of atrial fibrillation, CAD, previous stroke, carotid stenosis, higher NIHSS and altered level of consciousness at presentation. The LLOS score (Table 2) performed well on ROC analysis (Derivation cohort AUC=0.72, Internal validation AUC=0.73 and External validation AUC=0.77). Conclusion: Many factors play a role in determining the length of stay for AIS patients. Our study provides a scoring system that may help physicians predict which patients are more likely to have a prolonged hospital stay.