The aim of the present study was to predict which patients with severe or difficult-to-treat asthma are at highest risk for healthcare utilisation can be predicted so as to optimise clinical management. Data were derived from 2,821 adults with asthma enrolled in The Epidemiology and Natural History of Asthma: Outcomes and Treatment Regimens (TENOR) study. Multiple potential predictors were assessed at baseline using a systematic algorithm employing stepwise logistic regression. Outcomes were asthma-related hospitalisations or emergency department (ED) visits within 6 months following baseline. Overall, 239 subjects (8.5%) reported hospitalisation or ED visits at follow-up. Predictors retained after multivariate analysis were as follows: younger age; female sex; non-white race; body mass index > or =35 kg x m(-2); post-bronchodilator per cent predicted forced vital capacity <70%; history of pneumonia; diabetes; cataracts; intubation for asthma; and three or more steroid bursts in the prior 3 months. A final risk score derived from the logistic regression model ranged from 0-18 and was highly predictive (c-index: 0.78) of hospitalisation or ED visits. This tool was re-tested in a prospective validation using outcomes at 12- to 18-months follow-up among the same cohort (c-index: 0.77). The risk score derived is a clinically useful tool for assessing the likelihood of asthma-related hospitalisation or emergency department visits in adults with severe and difficult-to-treat asthma.