Unwarranted long delays in treating asthmatic patients requiring assisted mechanical ventilation (AMV) could be associated with higher death rates. The objective of the study was to identify which were the best predictive characteristics of patients with acute severe asthma (ASA) who required AMV. For this purpose, we reviewed retrospectively the hospital record of every patient with ASA admitted to the intensive care unit (ICU) from 1981 to 1991 (38 with AMV and 38 without AMV). Twenty-seven variables were obtained from the history, physical examination, laboratory tests, blood gases and treatment. Using a multivariate discriminant analysis, the most powerful predictor of patients needing AMV was a function that consisted of eight variables: arterial pH, number of previous admissions to ICU, asthma severity, time elapsed since last visiting a physician, respiratory rate, age, systolic pressure and heart rate. With this function, 33 of 38 mechanically ventilated patients were well classified (sensitivity: 0.89) and the overall accuracy of the test was 92% (70 out of 76 cases). The positive and negative predictive values of the function for mechanical ventilation were 0.96 and 0.90, respectively. We conclude that the application of the calculated final discriminant function could be appropriate to decide which patients with ASA require AMV.