Background and objective: To examine the calibration of the prognostic system Acute Physiology and Chronic Health Evaluation Score (APACHE II) regarding hospital mortality and predicting weaning outcome after long-term mechanical ventilation of the lungs.
Methods: Prospective observational cohort study performed in a respiratory intensive care unit including 246 patients whose lungs were ventilated for 42.1+/-37.8 (median 30) days in the referring hospital. APACHE II (24 h after admission to our respiratory intensive care unit) and the cause of respiratory failure, underlying disease, prior duration of mechanical ventilation and gender were recorded. The predictive power was evaluated with sensitivity and specificity for different cut-off points and summarized in a receiver operating characteristic curve.
Results: No difference was found between survivors (APACHE II 16.0+/-4.3) and non-survivors (APACHE II 16.9+/-5.1). In a mean time of 8.0+/-10.3 days, 146 patients (59.3%) were successfully weaned (APACHE II 15.2+/-3.5). One-hundred patients (40.7%) were considered unweanable (APACHE II 17.7+/-5.3). Recalibration of APACHE II to predict weaning failure was possible, resulting in an area under the receiver operating characteristic curve (AUC) of 0.638. Furthermore the AUC improved to 0.723 by changing the weights of selected APACHE items and introducing external factors. Diagnostic accuracy fell from group with mechanical ventilation < or =25 days (AUC 0.770) to group with mechanical ventilation >50 days (AUC 0.517).
Conclusions: APACHE II did not predict hospital mortality after long-term mechanical ventilation of the lungs. Not the original APACHE II but a recalibrated and adapted APACHE II can be useful to predict weaning outcome in patients with less than 25 days of prior lung ventilation.