Background: Both delayed and premature extubation increase complication rate, the need for tracheostomy (TT), the duration of intensive care unit stay, and mortality. In this study, we therefore investigated factors associated with primary TT and predictors for extubation failure (EF) in a sample of severely affected ventilated stroke patients. Methods: One hundred eighty five intubated stroke patients were prospectively analyzed in this observational study. Patients not meeting predefined clinical and respiratory extubation criteria received a TT. All other patients were extubated and followed up for the need of reintubation. Characteristics of patients with and without extubation attempt were examined. Additionally, within the group of extubated patients, subgroups of successfully vs. unsuccessfully extubated patients were compared. Clinical factors associated with reintubation, including a previously established semi-quantitative airway score, were determined and predictors of EF were assessed. Results: Ninety-eight of 185 patients (53%) were primarily extubated; EF rate was 37% (36 patients). Eighty-seven (47%) were tracheostomized without a prior extubation attempt. Primarily tracheostomized patients had more severe strokes, which were more often hemorrhagic, presented with a lower level of consciousness, needed neurosurgical intervention more often, had a higher rate of obesity, and were more frequently intubated because of suspicion of compromised protective reflexes. EF was independently predicted by prior neurosurgical treatment and low airway management scores. No differences were found for the ability to follow simple commands and classical weaning criteria. Conclusion: Airway management decisions in intubated stroke patients represent a clinical challenge. Classical weaning criteria and parameters reflecting the patient's state of consciousness are not reliably predictive of extubation success. Criteria more closely related to airway safety and secretion handling may provide the most relevant information and should therefore be assessed by specific clinical scoring systems.