To assess the potential utility of different screening questions, we calculated the proportions of participants in each survey group who could accurately recall various combinations of the following: diagnosis of atopic dermatitis, diagnosis of eczema, or history of recurrent rash. Generalized additive models (31) and logistic regression were used to determine whether the time since last diagnosis of atopic dermatitis influenced recall. To select the appropriate representation of time since last diagnosis in logistic regression models, we examined graphical plots and accompanying statistical tests derived from generalized additive models, with time since last diagnosis entered as a nonparametric term. For each type of respondent (parent, patient, household contact), we fit a logistic regression model and generated predictions within the observed range of values for time since last diagnosis. Generalized additive models were fit by using S-PLUS software (Insightful Corp., Seattle, Washington), and logistic regression analyses were performed by using SAS software (SAS Institute, Inc., Cary, North Carolina).