Objective: To develop a model able to determine the right time for extubation and to validate its performance. paragraph sign
Design: A prospective clinical study.
Setting: 14-bed medical intensive care unit in a university hospital.
Patients: 101 patients (37 women/64 men) ventilated over more than 48 h (mean 10.4 +/- 10.3 days) and considered ready to be weaned by the medical team (February 1996-February 1998).
Methods: This study included two series: a development series with 53 patients and a validation series with 48 patients. Before extubation, a weaning test was performed measuring tidal volume (V(T)), respiratory rate (f), f/V(T) ratio, minute ventilation, vital capacity (VC) and maximum inspiratory and expiratory pressures (MIP and MEP). The success of extubation was assessed after 48 h. Receiver operating characteristic (ROC) curves allowed the analysis of the discriminating power of each parameter. Threshold values were determined using the Youden's index. To create the best predictive model, we performed a multiple logistic regression analysis. To assess the calibration and the discrimination of the model, the Hosmer- Lemeshow goodness-of-fit test and area under ROC curves (AUC) were adopted.
Measurements and results: In a development series, 60 tests were carried out with 38 successful extubations and 22 extubation failures. The multivariate analysis found three significant variables: VC (threshold value = 635 ml), f/V(T) ratio (threshold value = 88 breaths/min.l) and MEP (threshold value = 28 cmH(2)O). The validation cohort included 59 tests (38 successes and 21 failures). The validation series shows a good discrimination (AUC = 0.855 +/- 0.059) and calibration (goodness-of-fit test C: p = 0.224) of the model.
Conclusion: VC together with the f/V(T) ratio and MEP offer accurate prediction of early extubation.