Prediction Model of Physician-diagnosed Asthma among Children Aged 1–5 Years Presented with Wheezing Episode in Uttaradit Hospital

Abstract

ABSTRACT
Background: Wheezing is a common symptom in young children, particularly those under 5 years of age, and is associated with the subsequent development of asthma - one of the most prevalent chronic respiratory diseases in children worldwide. Diagnosing asthma before the age of 5 remains clinically challenging. Data from Uttaradit Hospital indicate that respiratory diseases represent a major health burden, with asthma ranking third among all respiratory conditions, underscoring the importance of studying the relationship between early wheezing episodes and asthma diagnosis in children.
Objective: To develop a prediction model estimating the probability of receiving a physician diagnosed asthma in children aged 1-5 years who present with wheezing episodes.
Methods: Children aged 1-5 years admitted to the pediatric ward of Uttaradit Hospital with a wheezing episode between January 1, 2020, and July 31, 2025, were included. Those with a prior physician-diagnosed asthma were excluded. Categorical and continuous variables were compared, and associations between potential risk factors and asthma diagnosis were assessed using univariable and multivariable logistic regression with backward elimination, applying a p-value threshold of 0.5. Odds ratios were subsequently converted to log-odds coefficients for model development. Model performance was evaluated across four domains: discrimination, calibration, internal validation, and clinical utility.
Results: The developed prediction model demonstrated moderate discriminative ability (AUROC 0.76; bootstrap-corrected 0.73) with good calibration (O:E ratio ≈ 1; calibration-in-the-large ≈ 0). Internal
validation confirmed model stability with no evidence of overfitting. Decision curve analysis demonstrated clinical utility beginning at a threshold probability of approximately 0.30.
Conclusions: The prediction model can satisfactorily classify and estimate the probability of physician-diagnosed asthma in children, with predictions consistent with observed outcomes. Further studies in larger and more diverse populations are warranted to confirm its accuracy and generalizability.