Investigating Asthma Disparities in Hispanic Communities Using Machine Learning Algorithms on the All of Us Researcher Workbench

<b>Purpose</b>: This study aims to examine factors associated with asthma prevalence among Hispanic participants in the United States, focusing on access barriers, socioeconomic indicators such as education and income, and BMI. Data from the <i>All of Us</i> Research Program were analyzed using both traditional statistical models and interpretable machine learning algorithms. <b>Methods</b>: We analyzed data from the <i>All of Us</i> Research Program, comparing individuals with and without asthma. Logistic regression models and interpretable machine learning algorithms, including MARS (Multivariate Adaptive Regression Splines) and CIT (Conditional Inference Trees), were used to identify factors associated with asthma prevalence and their interactions. <b>Results</b>: The logistic regression analysis identified several variables associated with higher odds of asthma, including older age, female sex, greater access barriers, higher BMI, lower income, and higher education levels. Hispanic participants with greater access barriers had 26.3% higher odds of asthma prevalence (aPOR = 1.263, 95% CI: 1.114-1.433) compared to those without such barriers, and each unit increase in BMI was associated with a 2.9% increase in the odds of having asthma (aPOR = 1.029, 95% CI: 1.023-1.035). The MARS algorithm captured nonlinear relationships and interactions, highlighting BMI, age, sex, access barriers, income, and education as key predictors associated with asthma prevalence. Among participants younger than 60.6 years, younger age was linked with higher asthma prevalence. An interaction between age (above 21.5) and male sex indicated that the odds of asthma slightly decreased with age among males. Additionally, low-income and high BMI together were associated with elevated asthma prevalence, suggesting compounding vulnerabilities. The CIT identified BMI as the most influential variable and further stratified asthma prevalence by age, sex, education, income, and access barriers. Higher asthma prevalence was consistently observed among older females with high BMI, lower income, and greater access barriers. <b>Conclusions</b>: Among Hispanic participants in the <i>All of Us</i> Research Program, lower income combined with higher BMI and greater access barriers were significantly associated with increased odds of asthma. Males had lower odds of asthma, while older individuals showed higher asthma prevalence. These findings highlight important associations rather than causal relationships and may inform public health efforts to address asthma disparities related to weight and healthcare accessibility among Hispanic populations.