To determine the incidence of, and the factors contributing to spontaneous extubation (SE), we followed prospectively all intubated children admitted to a pediatric ICU. Eleven potential risk factors were monitored and scored twice daily for 8 consecutive months. Using data from the first 204 patient admissions, we evaluated the risk factors by orthogonal discriminant analysis and found that four factors (patient age, amount of secretions, endotracheal tube slippage, and state of consciousness), when considered together, had good discriminating power for SE vs. intentional extubation. We tested this method on the next 45 patient admissions and identified all seven spontaneously extubated patients as high risk. Analysis of covariance revealed successful discrimination between low-risk and high-risk patients for up to three days after patient admission. Overall, the incidence of SE was 13% (33 of 249). The effect of extubation on gas exchange was the same for spontaneously and intentionally extubated patients. No morbidity or deaths were attributed to SE. Standard ventilator low-pressure alarms did not reliably signal the presence of SE, nor did upper extremity restraints keep patients from extubating themselves.