BACKGROUND: While rates of elevated blood lead levels (EBLLs) have declined nationally, there remain many local areas with high rates. Identifying and targeting resources to these areas is necessary for the effective elimination of lead exposure. Crude prevalence estimates from blood lead testing data are often limited by small sample sizes. An alternative is estimating the prevalence based on area-level risk factors. We sought to model the associations between EBLL prevalence and these risk factors in Minnesota.
METHODS: A cross section of blood lead levels for children aged 0–6 years in Minnesota from 2011–2015 was constructed from blood lead surveillance data. EBLL cases were confirmed blood lead tests of at least 5 μg/dL. An ecologic generalized linear model was fit to estimate the prevalence of EBLLs in a tract from its region, neighborhood deprivation index, and housing age. Tract EBLL prevalences estimated by the model were compared to crude estimates by comparing the point estimates and relative standard errors (RSEs) with generalized linear models.
RESULTS: Among 306,528 children tested, 1.1% had elevated blood lead levels. By tract, the crude prevalence ranged from 0.0% to 11.9% with a median of 0.5% (1332 tracts). Tracts in the highest quintile of deprivation had an EBLL prevalence 2.9 times higher than tracts in the lowest (p<0.01). Tracts in the highest quintile of percent of houses built before 1950 had an EBLL prevalence 2.6 times higher than tracts in the lowest (p<0.01). There was no difference between the crude and modeled point estimates (p=1.00). The Pearson correlation coefficient was 0.70. However, the modeled estimates were more precise than the crude estimates (p<0.01). The median RSE of the crude estimates was 15.1% (Interquartile range (IQR): 12.1–18.6%) while that of the modeled estimates was 0.8% (IQR: 0.8–1.0%). Modeled estimates were significantly higher than the crude estimates in 14 tracts and significantly lower in 100 tracts.
CONCLUSIONS: In Minnesota, census tracts with older housing stock and greater deprivation tended to have higher prevalences of EBLLs. The tract EBLL prevalence estimated by an ecologic model can be used as a substitute for crude prevalence estimates when the crude estimates are imprecise. The difference between crude and modeled estimates can identify areas for public health intervention: identifying areas that could benefit from improved targeted testing and areas that may have sources of lead exposure beyond those associated with deprived areas and older housing.