Abstract
INTRODUCTION: Atopic dermatitis (AD) is the non-fatal skin disease with the highest global disease burden. AIM: This study investigates the factors influencing AD incidence and aims to construct a predictive risk nomogram model. MATERIAL AND METHODS: A total of 469 patients with dermatitis and eczema who were treated at the Dermatology Department of the Fourth Affiliated Hospital of Soochow University from January 2021 to May 2025 were included, and categorized into an AD group (n = 216) and a non-AD group (n = 253). Differences between the two groups were analysed through univariate analyses, followed by multivariate logistic regression. A predictive model was then constructed. The diagnostic performance of the model was evaluated using calibration curve, Hosmer-Lemeshow (H-L) test, receiver operating characteristic (ROC) curve, and clinical decision curve analysis (DCA) approaches. RESULTS: Significant differences were observed between the AD and non-AD groups in terms of age, history of allergic diseases, family history of allergy, infantile onset history, number of medical visits, blood eosinophil count, and serum total immunoglobulin E levels (p < 0.001). Logistic regression analysis showed that the number of medical visits (OR = 1.16, 95% CI: 1.08-1.24), history of allergic diseases (OR = 2.68, 95% CI: 1.44-5.00), family history of allergy (OR = 2.30, 95% CI: 1.27-4.17), and infantile onset history (OR = 23.80, 95% CI: 2.92-193.62) were significant factors influencing AD incidence. Calibration curve, H-L test (χ(2) = 6.68, p = 0.57 in the training set and χ(2) = 7.96, p = 0.44 in the validation set) and area under the ROC curve tests (0.808 in the training set and 0.812 in the validation set) demonstrated that the model exhibited good fit and predictive accuracy. The DCA results indicated that the model provided certain net clinical benefits in predicting AD. CONCLUSIONS: Factors including frequent hospital visits, history of allergic diseases, family history of allergy, and infantile onset are closely associated with the occurrence of AD. The model based on these risk factors developed herein offers significant value as a potential tool for predicting the occurrence of AD.