Exploring spatially varying relationships between children's lead poisoning and environmental factors

Abstract

Children's lead poisoning continues to compromise children's health and development, particularly in the inner cities of the United States. We applied a global Poisson model, a Poisson with random effects model, and a geographically weighted Poisson regression (GWPR) model to deal with the spatial dependence and heterogeneity of the number of children's lead poisoning cases in Syracuse, New York. We used three environmental factors-the building year (i.e., the year of construction) of houses, the town taxable value of houses, and the soil lead concentration-averaged at the census block level to explore the spatially varying relationships between children's lead poisoning and environmental factors. The results indicated that GWPR not only produced better model fitting and reduced the spatial dependence and heterogeneity in the model residuals but also improved the model predictions for the spatial clusters, or hot spots, of children's lead poisoning across inner city neighborhoods. Furthermore, the spatially varying model coefficients and their associated statistical tests were visualized using geographical information system maps to show the high-risk areas for the impacts of the environmental factors on the response variable. This information can provide valuable insights for public health agencies to make better decisions on lead hazard intervention, mitigation, and control programs.

Global and Geographically Weighted Quantile Regression for Modeling the Incident Rate of Children’s Lead Poisoning in Syracuse, NY, USA

“…The regression relationships at the “average” level and at different quantiles are quite different. Shao et al (2017) concluded that more recently built houses with higher average town taxable values would have lower likelihood of lead poisoning, while the children would have a higher chance of exposure to lead if the surrounding of the house had higher soil lead concentration for the “average” relationship via Poisson models [ 11 ]. In this study, the global quantile models confirmed that influential directions extending to both lower and upper tails.…”

“…During the last decades the U.S. federal and state governments have devoted considerable efforts for lead hazard screening, mitigation, and control via intervention regulations and education programs. For example, the Lead Hazard Control Program of New York have assisted in reducing lead paint hazards in houses and provided lead inspection for rental properties and homes built before 1978 in the city of Syracuse (NY, USA), [ 11 ]. Although the national surveys and local surveillance data indicate that children’s blood lead level (BLL) (e.g., ≥10 g/dL) has declined over time [ 12 , 13 , 14 , 15 ], the incident of lead poisoning remains relatively high in some inner cities of USA.…”

“…On the other hand, the spatially varying relationships may provide us opportunities to search and identify the key and specific areas across a large geographical study region. Particularly, a localized regression method, geographically weighted regression (GWR), has become popular in different disciplines in recent years (e.g., [ 11 , 27 , 28 , 29 , 30 , 31 , 32 , 33 ]). The applications of GWR in epidemiological research helped to identify the “hot spots” of the inner cities where the disease rate is more severe than other areas, which may be more important and useful to policy makers for the purpose of planning and management [ 24 , 34 ].…”

Variability in the spatial density of vacant properties contributes to background lead (Pb) exposure in children

“…National figures of decreases in elevated BLL, ≥5 µg/dL, (Kennedy et al, 2014;Tsoi et al, 2016) obscure the fact that the spatial distribution of exposure is clustered in low-income areas. Studies have found a number of factors associated with BLL that cluster at the censustract level including: mean age of housing, mean value of housing, median housing income, and proportion of vacant housing (Reissman et al, 2001;Sargent et al, 1997;Shao et al, 2017b;Stewart et al, 2014); more so, individual BLL have been found to be spatially auto-correlated (Berg et al, 2017;Griffith et al, 1998). In other words, measurements of BLL are correlated with each other across space, suggesting that an underlying spatial process is influencing levels of exposure (Shao et al, 2017b).…”

“…Studies have found a number of factors associated with BLL that cluster at the censustract level including: mean age of housing, mean value of housing, median housing income, and proportion of vacant housing (Reissman et al, 2001;Sargent et al, 1997;Shao et al, 2017b;Stewart et al, 2014); more so, individual BLL have been found to be spatially auto-correlated (Berg et al, 2017;Griffith et al, 1998). In other words, measurements of BLL are correlated with each other across space, suggesting that an underlying spatial process is influencing levels of exposure (Shao et al, 2017b). This spatial process is not likely to be natural, but instead a function of the built environment (Krieger et al, 2003;Miranda et al, 2002).…”