Objectives: To assess the outcome and to identify early prognostic indicators in a global population of patients with hematologic malignancy admitted to the intensive care unit for a life-threatening complication. Design: Retrospective observational study. Setting: Medical intensive care unit at a tertiary university hospital. Patients: A total of 124 consecutive critically ill patients with a hematologic malignancy admitted to the intensive care unit during a 3.5-yr period. Measurements: We collected variables at admission and during admission and identified predictors of in-hospital mortality by stepwise logistic regression analysis. Main Results: Mean Acute Physiology and Chronic Health Evaluation II score was 26 +/- 7.7. Sixty-one percent had a high-grade malignancy, and 27% had active disease. Thirty-five percent were leukopenic (leukocyte count, <1.0 x 10(9)/L) at admission. Respiratory failure (48%), sepsis (18.5%), and neurologic impairment (17%) were the major reasons for admission at the intensive care unit. Seventy-one percent of the patients required ventilatory support for a median duration of 6 (3-17) days, 46% received vasopressors at admission, and 26.6% needed renal replacement therapy during their intensive care unit stay. A recent bacteremia precipitating intensive care unit admission was found in 21.8% of the patients. Crude intensive care unit, in-hospital, and 6-month mortality rates were 42%, 54%, and 66%, respectively. Four variables were independently associated with outcome in a multivariate logistic regression analysis: leukopenia (odds ratio, 2.9; 95% confidence interval, 1.1-7.7), vasopressors (odds ratio, 3.74; 95% confidence interval, 1.4-9.8), and urea of >0.75 g/L (>12 mmol/L) (odds ratio, 9.4; 95% confidence interval, 4.2-26) at admission were associated with poor outcome, whereas recent bacteremia (odds ratio, 0.17; 95% confidence interval, 0.05-0.58) was associated with better prognosis. Using these variables, we arbitrarily categorized our population into three groups for survival analysis: a low-risk group (low urea with or without either leukopenia or vasopressors, n = 60), an intermediate-risk group (high urea or a combination of leukopenia and vasopressors, n = 34), and a high-risk group (high urea in combination with leukopenia or vasopressors, n = 27). Patients with a bacteremia prompting intensive care unit admission were allocated to a one-step-lower risk group. Survival probabilities at 30 days and 6 months were 75% and 55% in the first group, 35% and 21% in the second group, and 4% and 0%, respectively, in the third group (p < .001). Conclusion: The general reluctance to admit patients with a hematologic malignancy to the intensive care unit, even with severe critical illness, is unjustified. However, we identified four early predictors of outcome that may be of value in deciding in which patients advanced or prolonged support should not be continued.
