Development of a prediction model for infants at high risk of food allergy

Background: Identification of risk factors for food allergy (FA) in infants is an active research area. An important reason is to identify optimal target infants for early introduction of specific food antigens. Although eczema has been used for this purpose, multivariable prediction scores have not been reported. Objective: The aim of this research is to develop a multivariable prediction score for infants at high risk of FA. Methods: We performed a cross-sectional analysis of a self-administered questionnaire for the parents of 18-month-old children at well-child visits between April 2016 and March 2017 (development dataset) and between April 2017 and March 2018 (validation dataset). We developed and validated the prediction score. Results: The questionnaire collection rate was 18,549 of 20,198 (92%) in the development dataset and 18,620 of 19,977 (93%) in the validation dataset. Risk factors for FA were being born in August-December, first child, eczema, atopic dermatitis in father and mother, and FA in mother and sibling(s). For identifying infants with FA, the developed multivariable prediction score showed higher discrimination ability (area tinder the curve [AUC] = 0.75) than focusing on eczema (AUC = 0.70) in the validation dataset. The score was also useful for identifying infants with a history of anaphylaxis (AUC = 0.73) than focusing on eczema (AUC = 0.67) in the validation dataset. Conclusion: The new prediction score enables more efficient identification of infants at high risk of FA, who may be the optimal target group for the early introduction of specific antigens.