How can household dampness-related exposure and its related health outcomes be predicted?

Abstract

Dampness-related exposures have been widely discovered in household, which are associated with childhood health. This study employed the Back-Propagation (BP) neural network to develop prediction models of overall dampness evaluation based on the on-site dampness inspection in 454 families. The trained models were then used to predict the overall dampness exposure in 15,266 families with five parent-reported dampness-related indicators. Two indices, including the overall-dampness- exposure-based (ODEB) score and combination-of parent-reported- based (CPRB) score, were introduced to reflect the household dampness exposure in Shanghai, and were used to build the relationship with different childhood respiratory diseases and allergies. The results indicated that the trained model of BP neural network had a reasonable accuracy with the coefficient of determination R2 greater than 0.5, and the standard deviation ratios were approaching 1.0. Both the ODEB score predicted by trained model and CPRB score were developed based on parent reported dampness indicators in the cross-sectional survey, and were significantly associated with childhood respiratory diseases after adjusting the confounders (AOR >1.3, p < 0.01, SDS factor >0). Exposure-response relationships were also found between the dampness exposure indices and health outcomes. A continuous dampness evaluation index with a reasonable accuracy was proposed based on BP neural network prediction of on-site dampness scoring, which was consistent with the CPRB score, showing an exposure-response relationship with childhood diseases. These findings provide us an approach to predict the household dampness and its related adverse health outcomes.