Exploring indoor PM2.5 pollution characteristics in Xi'an city and its health implications using interpretable machine learning

Indoor ambient fine particulate matter (PM_2.5) exposure is harmful to human health, but its sources and health effects are not fully understood. This study conducted a 2-week monitoring campaign in Xi'an, China, assessing levels of indoor and outdoor PM_2.5 pollution, fractional exhaled nitric oxide, and cytokines (interleukin-6 (IL-6), tumor necrosis factor-α, and 8‑hydroxy-2′-deoxyguanosine) in 12 households. The results revealed an average indoor PM_2.5 concentration of 71.18 μg/m³, exceeding WHO (25 μg/m³) and China indoor air quality standards (50 μg/m³). A random forest regression model effectively predicted indoor PM_2.5 levels (R² = 0.84), while SHapley Additive exPlanations indicated that outdoor PM_2.5 was the primary contributor to indoor concentrations, contributing 38 %. Although average FeNO levels (14.68 ppb) did not exceed healthy thresholds or correlate with overall PM_2.5 concentrations, FeNO levels were notably more sensitive to PM_2.5 in children than in adults. Conversely, IL-6 levels were correlated strongly with indoor PM_2.5 concentrations (0.67, p < 0.05) and were particularly responsive to smoking (0.72, p < 0.05) and outdoor PM_2.5 (0.62, p < 0.05). This study highlights the critical effect of indoor PM_2.5 exposure on health and emphasizes the complex sources and implications of indoor pollution in Xi'an, providing a scientific basis for improving indoor air quality.