Disentangling childhood asthma risk structure via Bayesian network topology

Asthma is a complex chronic respiratory disease that involves intricate interactions among indoor air pollution, allergen exposure, and multiple comorbidities. Traditional statistical methods often fail to capture dependencies among risk factors in nonlinear, multifactorial relationships. To address this, we employ a Bayesian network model from a network science perspective to represent probabilistic dependencies via a directed acyclic graph and conditional probability tables, enabling comprehensive analysis of dependency patterns and community structures. Community detection reveals four distinct modules, which yield a modularity of 0.59. Additionally, we develop a novel quantitative assessment system to rank environmental factors based on their relative contributions to asthma risk. Among individuals with allergic tendencies, changes in cleaning habits were linked to the highest increase in predicted asthma risk, reaching 7.64%. Paradoxically, this behavioral change appears to elevate asthma risk. Furthermore, the combined effect of cleaning habit modification and pet avoidance demonstrates significant synergy, providing valuable insights for developing early warning systems and targeted preventive strategies.