We retrospectively reviewed the hospital records of 53 patients admitted for 73 episodes of myasthenic crisis at Columbia-Presbyterian Medical Center over a period of 12 years, from 1983 to 1994. Median age at the onset of first crisis was 55 (range, 20 to 82), the ratio of women to men was 2:1, and the median interval from onset of symptoms to first crisis was 8 months. Infection (usually pneumonia or upper respiratory infection) was the most common precipitating factor (38%), followed by no obvious cause (30%) and aspiration (10%). Twenty-five percent of patients were extubated at 7 days, 50% at 13 days, and 75% at 31 days; the longest crisis exceeded 5 months. Using survival analysis and backward stepwise Cox regression, we identified three independent predictors of prolonged intubation: (1) pre-intubation serum bicarbonate > or = 30 mg/dl (p = 0.0004, relative hazard 4.5), (2) peak vital capacity day 1 to 6 post-intubation < 25 ml/kg (p = 0.001, relative hazard 2.9), and (3) age > 50 (p = 0.01, relative hazard 2.4). The proportion of patients intubated longer than 2 weeks was 0% among those with no risk factors, 21% with one risk factor, 46% with two risk factors, and 88% with three risk factors (p = 0.0004). Complications independently associated with prolonged intubation included atelectasis (p = 0.002), anemia treated with transfusion (p = 0.03), Clostridium difficile infection (p = 0.01), and congestive heart failure (p = 0.03). Three episodes of crisis were fatal, for a mortality rate of 4% (3/73); four additional patients died after extubation. All seven deaths were due to overwhelming medical comorbidity. Over half of those who survived were functionally dependent (home or institutionalized) at discharge. In addition to prospective controlled studies of immunotherapies, the prevention and treatment of medical complications offers the best opportunity for further improving the outcome of myasthenic crisis.