Prediction models for early-onset atopic dermatitis in infancy: a prospective cohort study

Abstract Background Atopic dermatitis (AD) is the most common chronic inflammatory skin disease in childhood. Early identification of infants at high risk for AD may support preventive interventions and reduce the subsequent burden of allergic diseases. This study aimed to develop and evaluate prediction models for early-onset AD during the first two years of life. Methods Prediction models were developed using data from a prospective cohort study of healthy infants aged 42–90 days enrolled at Songklanagarind Hospital between August 2020 and October 2021. Demographic and clinical characteristics were collected at baseline. AD was assessed at 12 and 24 months of age using the Hanifin and Rajka diagnostic criteria. Multivariable logistic regression models were constructed, with variable selection performed using stepwise and least absolute shrinkage and selection operator (LASSO) methods. Model performance was evaluated using sensitivity, specificity, area under the receiver operating characteristic curve (AUC), F1 score, and calibration plots. Results Among 671 infants included in the analysis, 55 (8.2%) developed AD within the first two years of life. Higher household income and a family history of allergy were significantly associated with increased AD risk, whereas regular moisturizer use was significantly associated with reduced AD risk. The stepwise-selected model demonstrated lower sensitivity than the LASSO-selected model (68% vs. 74%) but higher specificity (75% vs. 66%) and AUC (74% vs. 67%). Both models were similar in F1 score (26%) and calibration. Conclusions Prediction models based on early-life clinical factors demonstrated promising performance in identifying infants at risk of atopic dermatitis. These models may support early risk stratification and prevention-oriented counseling in infancy. Trial registration The present study is a secondary analysis of data derived from participants enrolled in the V114-032 (PNEU-ERA) clinical trial, which was prospectively registered at ClinicalTrials.gov (NCT04193215). The current cohort analysis evaluating prediction models for atopic dermatitis did not involve a separate interventional protocol.