As discussed earlier, our study suggested OI sellekchem is more sensitive than the traditional PaO2/FiO2 ratio in assessing the oxygen exchanging status. For this reason, OI can be a potential predictor for risks of development of ALI and ARDS in patients with acute respiratory failure and thus enables earlier modification of ventilation strategy.Our study showed preexisting CVA was an independent predictor for ventilator weaning failure. Although acute stroke patients who require mechanical ventilation are known to carry poor outcomes [27, 28], preexisting cerebral infarction, and cerebral hemorrhage in patients admitted to ICU for a different disease are not necessarily associated with prolonged mechanical ventilation [29], which instead, may be related to the extent of neurological deficit.

However, preexisting CVA was shown to be a risk factor for weaning failure in our study. This may be explained by the older age (78.3 year old versus 69.4 year old, P = 0.003) and higher number of comorbidities (2.9 versus 2.19, P = 0.01) in patients with preexisting CVA in our study.Change in OI in the 1st 3 days was shown to correlate with weaning outcome in our study, though multivariate analysis failed to establish its role as an independent predictor. The relationship between OI and weaning outcome was discussed in several studies. In the study by Tseng et al., they demonstrated that congestive cardiac failure (P = 0.009), initial high oxygenation index value (P = 0.04), increased SOFA scores (P = 0.01), and increased APACHE II scores are independent predictors of ventilator dependence in patients with ventilator-associated pneumonia [30].

The study by Gajic et al. also suggested that age, OI, and cardiovascular failure three days after intubation are predictors of death or prolonged mechanical ventilation [31]. Comparing with the above mentioned studies, the heterogeneous nature of our ventilated patients may account for the different result in our study.Our study has several limitations. First, a retrospective review of existing data was conducted, inevitably, disadvantage such as missing key data in small amount of patients would occur. This may reduce the representativeness of the sample. Second, the relatively small sample size implies a single data may have a greater influence on final results. Despite this, it did not affect our final conclusion or inference because our main results were highly significant since their P values were less than 0.05 or even less than 0.001. Third, it Brefeldin_A is not known which ventilator strategies (e.g., low tidal volume strategy, lung recruitment, etc.) were used in each patient and for how long, and if any additional therapies were used (e.g.