Developing a prediction model of children asthma risk using population-based family history health records

Development of a predictive model for pediatric atopic dermatitis: A retrospective cross-sectional nationwide database study

BACKGROUND

The rise in prevalence of atopic dermatitis (AD) has been correlated with numerous elements of the exposome, modern-day lifestyle, and familial history. The combined analysis of familial history and other risk elements may allow us to understand the driving factors behind the development of AD.

OBJECTIVE

To develop prediction models to assess the risk of developing AD using a large and diverse cohort (N = 77,525) and easily assessed risk factors.

METHODS

We analyzed electronic medical record data from Leumit Health System. Documented predictive factors include sex, season of birth, environment (urban/rural), socioeconomic status, household smoking, diagnosed skin conditions, number of siblings, a paternal, maternal, or sibling history of an atopic condition, and antibiotic prescriptions during pregnancy or after birth. Predictive models were trained and validated on the data set.

RESULTS

Medium (odds ratio [OR] 2.04, CI 1.92-2.17, P < .001) and high (OR 2.13, CI 1.95-2.34, P < .001) socioeconomic status, a previous diagnosis of contact dermatitis (OR 2.57, CI 2.37-2.78, P < .001), presence of siblings with an AD diagnosis (OR 2.21, CI 2.04-2.40, P < .001), and the percentage of siblings with any atopic condition (OR 2.58, CI 2.09-3.17, P < .001) drove risk for AD in a logistic regression model. A random forest prediction model with a sensitivity of 61% and a specificity of 84% was developed. Generalized mixed models accounting for the random effect of familial relationships boasted an area under the curve of 0.98.

CONCLUSION

Predictive modeling using noninvasive and accessible inputs is a powerful tool to stratify risk for developing AD.