Abstract 3247: Predicting Prolonged Intensive Care Unit Stay In Patients Undergoing Coronary Artery Bypass Grafting
Background: Prolonged length of stay (LOS) in the intensive care unit (ICU) following coronary artery bypass grafting (CABG) places a considerable burden on the health care system. Few tools allow one to predict patients at risk for prolonged ICU LOS. We sought to develop a predictive model based on pre-operative clinical and non-clinical factors to identify patients at risk for prolonged ICU LOS following CABG.
Methods: Retrospective analysis was performed on all patients undergoing isolated CABG at a single center between June 1998 and December 2002. Prolonged ICU LOS was defined as initial admission to ICU exceeding 72 hours. Patients with prolonged ICU LOS were compared to those within standard. An internally validated parsimonious risk-predictive model was constructed.
Results: Of 3476 patients undergoing isolated CABG between June 1998 and December 2002, 411 (11.8%) experienced prolonged ICU LOS. These patients were older (68.2 ±10.1 vs 63.7 ± 10.4; p= < 0.0001) and more likely to have diabetes (45.0% vs 35.4%; p=0.0001), pre-op renal failure (13.9% vs 3.6%; p= < 0.0001), triple vessel and/or left main disease (90.0% vs 78.8%; p= < 0.0001) and NYHA IV symptoms (69.1% vs 42.4%; p= < 0.0001). Income or education did not differ between the two groups. Overall in-hospital mortality was higher among patients requiring prolonged ICU LOS (14.4% vs. 0.0%, p= < 0.0001). The following variables were found to be independent predictors of prolonged ICU LOS: Subsequent validation of this model on 2000 consecutive patients undergoing isolated CABG between January 2003 and September 2005 patients demonstrated a ROC of 78%.
Conclusion: An internally-validated, risk predictive model of prolonged ICU LOS in patients undergoing CABG was constructed on the basis of pre-operative clinical factors. Identification of these patients allows for strategies aimed at optimizing hospital resources.