METHODS: Overall spatial clustering of cases compared to controls was assessed using a difference of K-Function test. Adaptive bandwidth kernel density estimation (KDE) was calculated for cases and controls, and output from the two KDE functions was used to calculate a relative risk surface. Statistical significance of global clustering and local clusters of relative risk was evaluated using Monte Carlo simulation. The results were mapped along with neighborhood level housing characteristics to visually inspect the degree to which older housing accounts for exposure.
RESULTS: Spatially smoothed relative risk values for elevated blood lead vary between 0.13 and 6.88 within our study area. We found evidence of statistically significant global clustering of elevated blood lead cases with respect to controls drawn from the at-risk population. We also found evidence of statistically significant local clusters of elevated blood lead cases. Visualization of these results showed that areas of highest relative risk correspond to neighborhoods closest to Denver’s freeway corridors with greater than 75% older housing. The area of highest relative risk was located just east of the I-76/I-70 freeway interchange radiating north and south of I-70.
CONCLUSIONS: Relative risk of elevated blood lead is not evenly distributed across Denver’s communities at-risk for lead exposure. These methods allow for identification of risk at finer spatial scales compared to aggregating cases to spatial units such as census tracts. Using these methods can assist other public health entities interested in avoiding the pitfalls of aggregation when examining point level data valuable for use in targeting and tailoring public health intervention strategies.