USE OF POSTOPERATIVE INFORMATION TO PREDICT MORTALITY-RATES FOR PATIENTS WHO HAVE LONG STAYS IN THE INTENSIVE-CARE UNIT AFTER CORONARY-ARTERY BYPASS-GRAFTING

OBJECTIVE: to derive and evaluate prediction models for patients who h ad undergone coronary artery bypass grafting (CABG) and had a difficul t postoperative course. DESIGN: Observational. SETTING: Midwestern hos pital specializing in high-risk CABG procedures. PATIENTS: One hundred eight-three patients who stayed at least 10 consecutive days in the i ntensive care unit after a CABG procedure. OUTCOME MEASURE: Death with in 60 days of surgery. INTERVENTION: None. RESULTS: The final logistic regression prediction models included the following findings: pulmona ry capillary wedge pressure, cardiac index, heart rate, urine output, positive end-expiratory pressure, blood urea nitrogen levels, and the arterial pressure of carbon dioxide. The model was able to stratify pa tients into four risk groups with observed 60-day mortality rates of 0 .0% (n = 107), 21% (n = 39), 55% (n = 20), and 88% (n = 17). Preoperat ive patient information was not associated with prognosis for these pa tients. CONCLUSIONS: The findings suggest that a risk model that is sp ecific for patients who have undergone CABG and is based on postoperat ive findings may provide useful prognostic information for patients wh o are having a difficult postoperative course.