Pii: S1010-7940(02)00303-2

Objective: To identify predictors of requirement for readmission to the intensive care unit (ICU) for patients undergoing cardiac surgery. Methods: The setting was a 17-bedded ICU in a tertiary level institute for specialist adult cardiorespiratory disease. The case notes and ICU charts of 65 ICU readmissions and 65 controls, matched for day of initial ICU discharge, were analysed. Patient variables assessed included preoperative risk stratification, ICU admission APACHE III score and intensive therapy interventions, complications and indication for readmission if readmitted. Results: Twenty of 65 patients (31%) readmitted to the cardiac ICU died, compared with no mortality among the control group. Significant univariate determinants of ICU readmission (odds ratio, 95% confidence interval) included worse angina (1.38, 0.99–1.91) and dyspnoea (1.70, 1.10–2.61) classes and corresponding non-elective surgery (2.04, 1.31–3.19), higher Parsonnet score (1.06, 1.01–1.11) or EuroSCORE (1.14, 1.01–1.28), APACHE III score (1.03, 1.00–1.05), body mass index . 27 (4.25, 1.43–12.63), non-usage of beta-blockers (1.53, 1.03–2.26), emergency resternotomy (5.00, 1.10–22.79), and lower haemoglobin (0.75, 0.58–0.96), higher required inspiratory oxygen (1.05, 1.02–1.08), and higher respiratory rate upon ICU discharge (1.09, 1.01–1.18). Renal failure, respiratory failure and cardiac arrest were the most common indications for ICU readmission. Thirty-five of 65 patients readmitted to the ICU required ventilation for a mean of 7.1 days. The mean ICU readmission duration for all 65 cases was 5.7 days. Conclusions: Readmission of cardiac surgical patients to the ICU is associated with high morbidity and mortality, and substantial resource consumption. Parsonnet or EuroSCORE risk stratification models in combination with obesity, operative urgency, resternotomy and respiratory indices at time of intended ICU discharge are strongly associated with readmission to ICU. q 2002 Elsevier Science B.V. All rights reserved.