We prospectively followed 314 children discharged from a children's hospital emergency department (ED) following an asthma attack, to identify risk, factors for relapse, i.e. a second ED visit for asthma within the next 10 days. Parents were surveyed concerning their child's past medical history, drugs received prior to the index visit, triggering factors, physician availability, parental anxiety, and sociodemographic variables. Data on severity of the attack, emergency treatment, response to treatment and drugs prescribed on discharge were extracted from the medical record. Ninety-six of the 314 children (31%) relapsed, most (68%) within 24 hours. Using multiple logistic regression, a predictive model was developed on 211 patients ("test sample"). The best model contained two variables: (1) the number of ED visits for acute asthma in the previous year (odds ratio [OR] = 2.4 for 4 or more vs fewer visits, 95% CI = 1.3-4.4) and (2) the intake of an oral short-acting theophylline preparation during the course of the acute treatment (OR = 0.4, 95% CI = 0.2-0.7). The sensitivity, specificity and positive predictive values of this model for predicting relapse were 73, 53, and 40%, respectively. When applied to a second randomly selected "validation sample" of 103 children, sensitivity was 73%, specificity 50%, and PPV 41%, thus indicating the stability of the model. The model identifies the number of ED visits in the previous year as an important risk factor for relapse. It also suggests that oral short-acting theophylline may still have a role in the treatment of patients in whom the contribution of inflammation to airway obstruction is minimal.