Is the duration of mechanical ventilation predictable?

BACKGROUND Prolonged mechanical ventilation (MV) is associated with high morbidity, mortality, and cost. However, few and limited data are available on the prediction of duration of MV. We conducted an observational cohort study to seek predictive criteria. METHODS The study was performed in a surgical ICU (SICU) in a university hospital. One hundred ninety-five consecutive unselected patients and 203 episodes of MV were prospectively analyzed to determine if clinical features, physiologic parameters, or multifactor scoring systems, at the time of admission or intubation, could be used as predictors of MV > or = 15 days. A univariate statistical analysis and a multiple logistic regression were used. A prospective validation study was then conducted to determine the accuracy of the results. RESULTS (1) Univariate statistical analysis indicated that SICU length of stay, emergent endotracheal intubation as opposed to elective intubation, indication for MV, sepsis score at the time of admission and intubation, lung injury score (LIS) at the time of admission and intubation, number of organ system failures at the time of admission and intubation, and serum albumin concentration were significantly different between the two groups. (2) Only the circumstances (emergency) of endotracheal intubation (odds ratio [OR]=3.5, p=0.02) and the LIS (OR=3.7, p=0.004) independently predicted a duration of endotracheal intubation > or = 15 days. One hundred twenty-eight consecutive patients requiring emergent intubation and MV were included in the prospective validation. The accuracy of the LIS > or = 1 used to predict MV > or = 15 days was as follows: sensitivity=0.88; specificity=0.28; positive predictive value=0.24; negative predictive value=0.91. CONCLUSION Low incidence of MV > or = 15 days was observed (13% and 20%, respectively, in observational cohort study and validation study) in unselected SICU patients. LIS > or = 1 at the time of intubation provides excellent negative predictive value (0.93 and 0.91) of duration of MV > or = 15 days. These data suggest that tracheotomy should not be considered for patients with LIS < 1.