Developing a predictive model for fine particulate matter concentrations in low socio-economic households in Durban, South Africa

作者：Shezi B.^*; Jafta N.; Sartorius B.^*; Naidoo R. N.

来源：Indoor Air, 2018, 28(2): 228-237.

DOI：10.1111/ina.12432

• cross-validation
• factor analysis
• indoor air pollution
• model validation
• multivariate regression model
• particulate matter

In low-resource settings, there is a need to develop models that can address contributions of household and outdoor sources to population exposures. The aim of the study was to model indoor PM2.5 using household characteristics, activities, and outdoor sources. Households belonging to participants in the Mother and Child in the Environment (MACE) birth cohort, in Durban, South Africa, were randomly selected. A structured walk-through identified variables likely to generate PM2.5. MiniVol samplers were used to monitor PM2.5 for a period of 24 hours, followed by a post-activity questionnaire. Factor analysis was used as a variable reduction tool. Levels of PM2.5 in the south were higher than in the north of the city (P < .05); crowding and dwelling type, household emissions (incense, candles, cooking), and household smoking practices were factors associated with an increase in PM2.5 levels (P < .05), while room magnitude and natural ventilation factors were associated with a decrease in the PM2.5 levels (P < .05). A reasonably robust PM2.5 predictive model was obtained with model R-2 of 50%. Recognizing the challenges in characterizing exposure in environmental epidemiological studies, particularly in resource-constrained settings, modeling provides an opportunity to reasonably estimate indoor pollutant levels in unmeasured homes.

• 出版日期2018-3