Background: Although the predictive criteria for duration of mechanical ventilation may help to evaluate the right time of disconnecting a patient from the ventilator, the efficacy of the APACHE ( Acute physiology, Age and Chronic Health Evaluation) or IPS (Infection Probability Score) systems in the prediction of the need to ventilator and its duration needs further evaluation. The study was performed in Alzahra Medical Center of Isfahan in 2006 – 2007 to evaluate the efficacy of the predicting power needing intubation (NI), mechanical ventilation (NMV), and duration of MV.

Materials and Methods: One hundred eighty critically ill patients were included in this evaluative diagnostic study. On admission day in the ICU, patients' data were collected to compare the APACHE II and III, and IPS scores. The necessity of mechanical ventilation and its duration were recorded for each patient. Sensitivity, specificity, area under the curve and corrected prediction of outcome for each cut-off point were calculated for three scores.

Results: The best cut-off points for prediction of NI were 50 for APACHE III, 12 for APACHE II and 12 for IPS. The Youden index had best cut-off points for APACHE III=0.62, APACHE II=0.36 and IPS=0.4. The area under the Receiver Operating Characteristic (ROC) curve was 0.63, 0.77 and 0.83 for APACHE III, APACHE II and IPS respectively. There were statistical differences between APACHE III , APACHE II and IPS in terms of Youden index and the area under the ROC curve (P<0.05). For the prediction of NMV, APACHE III yield significantly better results in the area under the ROC curve and Youden index than those of APACHE II or IPS (p<0.05). In the prediction of more than 5 days respiratory support under MV none of the three scoring systems could provide a good discrimination.

Conclusion: To predict for NI or NMV, the APACHE III had better accuracy than the APACHE II or IPS. Concerning estimation of the requirement of mechanical ventilation in critically ill patients although the APACHE III showed to be sufficiently accurate the sensitivity and calibration of the scoring systems should be improved.