Childhood lead exposure remains a widespread, yet preventable public health problem despite community-level interventions (Campbell et al., 2013; Raymond & Brown, 2017; Rogers et al., 2014). The Advisory Committee on Childhood Lead Poisoning Prevention (2012) recognized that no level of lead exposure is safe. For a point of clinical reference, however, the Centers for Disease Control and Prevention (CDC) established the reference range for an elevated blood lead level (BLL) in children to be [greater than or equal to]5 [micro]g/dL (CDC, 2015). It is well documented that lead exposure in children, even at BLL concentrations
As a result of more stringent government policy, the severity and scope of the problem has significantly decreased in recent decades in the U.S. (Bridbord & Hanson, 2009). In the late 1970s, the national average BLL of children 1-5 years was 15 [micro]g/dL, while in 2010, it had declined to 1.3 [micro]g/dL (CDC, 2013). In Philadelphia, lead elevation has declined significantly since the early 2000s; however, the percentage of children with a newly identified elevated BLL remained stable from 2011 through 2015 (Philadelphia Department of Public Health [PDPH], 2017).
Lead-based paint and its dust in pre-1950 housing are known to be the most common sources of lead exposure in homes (Chandran & Cataldo, 2010). Other sources of lead exposure include contaminated soil, food, water, jewelry, toys, and folk medicine (Chandran & Cataldo, 2010). Pre-1950 housing, poverty, education, and race have been consistently reported in previous studies as common risk factors (Akkus & Ozdenerol, 2014; Brown & Longoria, 2010; Jones et al., 2009).
The use of GIS has become a valuable tool in identifying areas with high rates of elevated BLLs, assessing geographical risk factors, and monitoring screening rates. Although other studies have addressed community-level risk factors for BLLs [greater than or equal to]10 [micro]g/dL (the recommended reference range prior to 2012), a large-scale risk factor analysis has not been carried out recently in Philadelphia (Akkus & Ozdenerol, 2014).
We conducted a census tract-level analysis of elevated BLLs in children in Philadelphia to 1) determine the associations of community-level housing characteristics and socioeconomic status as risk factors and 2) develop a risk score for each census tract. This study will contribute to the literature from the perspective of a large urban environment with high poverty and old housing by using 5 [micro]g/dL to define lead elevation. The results can help guide public health officials to judiciously allocate resources in urban areas with a high prevalence of risk factors.
Blood Lead Level Data and Study Population
In Pennsylvania, all BLL test results for children
Our study population consisted of a cohort of children born in 2008, 2009, and 2010 who had a Philadelphia home address and at least one previous venous BLL test. We excluded from our analysis children with only capillary and unknown specimen sources due to the increased likelihood of false positives (Parsons, Reilly, & Esernio-Jenssen, 1997). Children with at least one BLL [greater than or equal to]5 [micro]g/dL were classified as having lead elevation. Children with all test results
To determine how representative the study sample was of the total birth cohort population, we calculated the total number of children in each birth cohort using birth certificates and then divided the sample population by the total birth cohort population.
Census Tract Characteristics
Census tracts were used to represent geographic areas throughout Philadelphia. The U.S. Census Bureau conducts the American Community Survey (ACS) annually to collect a wide range of information including, but not limited to, household demographics, occupations, and education levels. Data from the 2011-2015 ACS 5-year estimates were extracted for each Philadelphia census tract (U.S. Census Bureau 2015a, b, c, d, e).
The covariates selected for the model were based on findings of previous studies on risk factors of elevated BLLs in children (Akkus & Ozdenerol, 2014). Census tract parameters included median age of housing, percent of housing units that were rental properties, percent of properties that were vacant, percent of people [greater than or equal to]25 years with a high school diploma, percent of people living below the poverty line, and percent of Black residents. Census tract-level median age of housing was categorized as either before or after 1950. Pre-1950 housing and vacant properties were intended to capture potential exposure to lead-based paint in the home and indicators of overall housing quality. For all covariates, the continuous variables (i.e., percentages) were categorized into quartiles.
Geocoding and Data Selection
Each child's home address was geocoded and assigned to its respective census tract. Children were excluded from the analysis if their addresses were missing or unable to be geocoded. Only census tracts with at least one child tested were included. The geocoded dataset of individual children was merged with the aggregate census tract-level data from ACS based on census tract number.
Pearson's chi-square tests and mixed effects logistic regression models were used to assess bivariate associations between individual covariates and binary outcome of lead elevation. We used a mixed effects multivariable logistic regression based on census tract characteristics to estimate adjusted associations between covariates and the binary outcome of lead elevation. The models included random intercepts for census tracts to model differences between census tracts not explained...