Objective: To identify the risk for prolonged mechanical ventilation in cardiac surgical patients.
Design: Prospective study with retrospective combination of a second database.
Patients: Six hundred and eighty-seven patients after cardiac surgery over a period of 12 months.
Measurements: Demographic data were recorded preoperatively, and surgical procedures intraoperatively using a surgical database designed for quality control. Length of ICU and hospital stay, and hospital outcome were recorded. Severity of illness was assessed daily using APACHE II, SAPS II, and Organ Failure Score. Intensity of treatment and nursing care was monitored by means of the Therapeutic Intervention Scoring System (TISS). Univariate and multivariate analyses were performed using logistic regression. The predictive value of the identified variables was tested by the Wilcoxon test using the receiver operating characteristic curve.
Main results: Sixty-two patients (9.0%) were ventilated for > 48 h and accounted for 42.8% of the total costs in the ICU. The pre- and intraoperatively collected data produced a model with weak predictive capacity for prolonged ventilation [area under curve (AUC) 73.22 and 71.08, respectively]. The use of TISS and SAPS postoperatively resulted in an effective model of prediction (AUC 93.76). Adding the occurrence of reoperation, reintubation, emergency transfusion, intraaortic balloon pumping, and need for total parenteral nutrition to the model further improved its predictive capacity (AUC 94.74).
Conclusions: The present results strongly suggest that data collected postoperatively using established scoring systems as well as documented events of high clinical impact for risk assessment and quality control are reliable predictors of prolonged ventilation.