Policy Changes and Child Blood Lead Levels by Age 2 Years for Children Born in Illinois, 2001–2014

Abstract

Objectives. To evaluate how lowering the blood lead level (BLL) intervention threshold affects childhood lead testing policy.

Methods. We geocoded 4.19 million Illinois lead testing records (2001–2016) and linked to 2.37 million birth records (2001–2014), data on housing age, industrial emissions, and roads. We used multinomial logistic regression to determine predictors of BLLs of 10 micrograms per deciliter (µg/dL) or greater, 5 to 9 µg/dL, and 4 µg/dL.

Results. We found that 2.2% of children had BLLs of 10 µg/dL or greater, 8.9% had BLLs of 5 to 9 µg/dL, and 5.7% had BLLs of 4 µg/dL. Pre-1930 housing was associated with more than 2- to 4-fold increased relative risk of BLLs above all thresholds. Housing built in 1951 to 1978 was associated with increased relative risk of BLLs of 5 to 9 µg/dL (relative risk ratio [RRR] = 1.14; 95% confidence interval [CI] = 1.06, 1.21) but not with increased relative risk of BLLs of 10 µg/dL or greater (RRR = 0.99; 95% CI = 0.84, 1.16). At a given address, previous BLLs of 5 to 9 µg/dL or BLLs of 10 µg/dL or greater were associated with increased risk of BLLs of 5 to 9 µg/dL or BLLs of 10 µg/dL or greater among current occupants by 2.37- (95% CI = 2.20, 2.54) fold and 4.08- (95% CI = 3.69, 4.52) fold, respectively.

Conclusions. The relative importance of determinants of above-threshold BLLs changes with decreasing intervention thresholds.

Public Health Implications. States may need to update lead screening guidelines when decreasing the intervention threshold.

Lead exposure remains a significant threat to children’s health in the United States. In 2014 alone, there were 37 824 cases of confirmed blood lead levels (BLLs) of 5 micrograms per deciliter (µg/dL) or greater among the 30 states that reported data to the Centers for Disease Control and Prevention (CDC).¹ Lead is a neurotoxin that irreversibly damages the developing brain.² Exposure in childhood is associated with decreased IQ³ and educational attainment^4,5 and increased risk of delinquency.^4,6 The most commonly cited risk factor for lead exposure is the age of housing.⁷ Lead was used as an additive in paint until a federal ban in 1978; however, because of a series of voluntary industry standards and local public health campaigns, the popularity of lead paint began declining around the 1930s and the concentration of lead in paint decreased significantly after 1950.⁸ Another commonly cited source of lead exposure is drinking water, which may be contaminated by flowing through lead service lines and lead residential pipes which were in use throughout most of the 20th century.⁹ Lead was also used as an additive in gasoline until 1995¹⁰ and is still released during some industrial processes involving heavy metals.¹¹ Therefore, proximity to roadways and industrial sites may be associated with lead exposure. The relative impact of these different exposure sources is not well understood.¹²

Until 2012, the CDC designated 10 µg/dL as the “BLL of concern.” In 2012, the CDC began designating a “reference BLL” as a guide for policymaking, which was initially set at 5 µg/dL, based on the 97.5th percentile of BLLs in the National Health and Nutrition Examination Survey, but more recent data indicate that the 97.5th percentile of BLLs is now 3.5 µg/dL.¹³ While there is consensus that there is no safe level of lead,¹⁴ many states use this CDC reference level as the threshold that triggers interventions such as case management and environmental inspections.

Although federal guidelines mandate that all children on Medicaid must be screened for lead exposure at ages 1 and 2 years, state and local health departments set their own guidelines for children not on Medicaid. While some states have adopted universal testing,¹⁵ neither the CDC nor the US Preventive Services Task Force has recommended the practice.^16,17 Many states, including Illinois, mandate testing only in certain zip codes that are designated high risk; an understanding of exposure sources is necessary to accurately define risk. However, as the intervention threshold decreases, the distribution and relative importance of different predictors of lead exposure may also change.

As part of a comprehensive evaluation of lead screening policies in Illinois, we used geospatial analysis of lead testing and exposure data to investigate the relative importance of different sources of lead exposure in predicting above-threshold BLLs in the setting of changing intervention thresholds. We linked 16 years of lead tests administered to more than 1 million children to Illinois birth records and geocoded the test address, allowing us to integrate exposure data, including housing age, industrial emissions, and the location of major roads. Using multinomial logistic regression analysis, we identified demographic and exposure variables associated with 3 outcomes: BLLs above the intervention threshold in Illinois during our sample period (10 µg/dL), BLLs between 5 and 9 µg/dL, and BLLs at a potential future level of reference of 4 µg/dL. To the best of our knowledge, this article is the first to show the implications of decreasing BLL thresholds for lead screening policy.

METHODS

We obtained birth records for all 2.37 million children born in Illinois between 2001 and 2014 from the Illinois Department of Public Health. Birth records included demographic information such as race, ethnicity, parental education level, and parental age. We also obtained records of all 4.19 million lead tests performed on these children between 2001 and 2016. In Illinois, there are 3 groups of children for whom doctors are required to complete blood lead tests: children who are on Medicaid, who comprise 41.6% of the population aged younger than 6 years according to the 2017 American Communities Survey; children who live in 1 of the 43.5% of zip codes designated high-risk by the Illinois Department of Public Health, including the entire city of Chicago; and children who screen positive on a risk assessment questionnaire provided by the Illinois Department of Public Health. The distribution of high-risk zip codes in Illinois is mapped in Figure 1a. The screening rate in our sample was 43.0%.

FIGURE 1—

Map of Illinois Zip Codes Showing (a) Illinois Department of Public Health Risk Designation and (b) Median Housing Age

Matching

Overall, we successfully matched 89.4% of all lead tests to a birth record and this rate improved steadily throughout the study period as shown in Figure A (available as a supplement to the online version of this article at http://www.ajph.org). We linked the lead testing and birth data sets by using a custom fuzzy matching algorithm based on the Jaro–Winkler string distance¹⁸ of first name, last name, and date of birth with manual review of optimal cutoffs. Each lead test record contained the name and address of the child, the date of the blood draw, the type of test (venous or capillary), the test result, and identifiers for the testing laboratory.

Certain laboratories reported all values below a certain threshold as the cutoff value, implying a minimum reporting limit. We determined the minimum reporting cutoffs for each laboratory test type–year combination by manual review of BLL histograms. The distribution of BLLs is skewed to the right; therefore, an absence of tests below a certain value for a given laboratory likely indicates that the laboratory has a minimum reporting limit. We recoded all test results that were reported at those limits to the average test result below the cutoff in that year–test type group among laboratories without cutoffs. We recoded all the corrected test results to values less than 3.8 µg/dL. We plotted a histogram of laboratory cutoffs in Figure B and we replicated our analysis with unadjusted test results in Table A (available as supplements to the online version of this article at http://www.ajph.org).

We only used tests before the age of 2 years because younger children spend more time in the home and are likely to be maximally exposed to lead hazards at home. For each child, we used the highest venous test if available. If children had only capillary tests, we used the highest confirmed capillary test—that is, a capillary test followed by another test within 3 months.¹⁹ Absent venous or confirmed capillary test, we used the highest unconfirmed capillary test. We replicated our analysis using only venous tests in Table B (available as a supplement to the online version of this article at http://www.ajph.org).

We standardized and geocoded all test addresses with ArcGIS (ESRI, Redlands, CA). We obtained data on housing age from parcel data provided by the Zillow Transaction and Assessment Database.²⁰ Among the 1 201 801 children with a lead test by age 2 years, we successfully geocoded testing addresses of 928 371 and linked parcel data of 640 347. The geocoding rate improved throughout the study period as shown in Figure C (available as a supplement to the online version of this article at http://www.ajph.org). Median housing age by zip code is mapped in Figure 1b. We obtained income data for each census block group from the 2015 American Community Survey²¹ and defined each census block group as low income if the median household income was less than $40 000.

We also collected Toxic Release Inventory (TRI) data maintained by the Environmental Protection Agency, which details industrial lead emissions, including geocodes for the point sources.²² These data are consistently available since 2001; hence, we started our sample that year. In addition, we obtained the location of major roadways, defined as state and interstate highways, from the Illinois Department of Transportation.²³ We calculated the distance from lead-emitting facilities and roadways to each child’s address. We performed data cleaning, linkage, analysis, and plotting in R version 3.4 (R Foundation, Vienna, Austria) and Stata version 15.1 (StataCorp LP, College Station, TX).

Data Analysis

We investigated correlates of above-threshold BLLs at different thresholds by estimating the following multinomial logistic regression model:

where BLL_iyzt = x is a categorical variable that is 1 if child i had a BLL of 4 µg/dL, 2 if the child had a BLL of 5 to 9 µg/dL, and 3 if the child had a BLL of 10 µg/dL or greater. Although some have suggested 3.5 µg/dL as a future reference level, we used 4 µg/dL as a hypothetical cutoff because during our study period, the Illinois lead program stored test results truncated to the nearest integer. The subscripts y, z, and t denote the birth cohort, zip code, and the year of test, respectively. Vintage_i is a vector of indicators for decade of construction of housing at testing address for child i, which was initially stratified by decade to generate Figure 2, and subsequently stratified as 1930 or before, 1931 to 1950, 1951 to 1977, and 1978 or after to generate the tables; TRI_iy is a vector of cumulative industrial lead emissions at given distances from the child up to birth year y; Road_i is a vector of indicators for major roads at given distances from the child’s testing address; and X_i is a vector of child and mother characteristics, including indicators for previous instances of above-threshold BLLs at the child’s testing address, race, ethnicity, age, Medicaid status, education and age of the mother as recorded in the birth certificate, and median income in the child’s census block group.

FIGURE 2—

Relationship Between Housing Construction Decade and Relative Risk Ratio of Above-Threshold Blood Lead Levels Among Children Aged 0 to 2 Years: Illinois, 2001–2014

Note. RRR = relative risk ratio. The reference category is housing built in 2000 or later. Benjamini–Hochberg–corrected P values of between-group χ2 testing for equality of coefficients are indicated with asterisks at the bottom of the graph.

*P < .05; **P < .01; ***P < .001.

We included vectors of indicators for the child’s birth cohort (ξ_y), the combination of testing laboratory and year of test (λ_t), and month of test (τ_t). By controlling for the combination of testing laboratory–year fixed effects, we partialled out systematic measurement or reporting bias. We report relative risk ratios (RRRs) and 95% confidence intervals (CIs) for all coefficients. We tested for equality of the coefficients for the 3 outcomes by using the χ² test and used the Benjamini–Hochberg procedure to account for multiple testing.²⁴ Z_z is a vector of zip code–level means of all the other regressors included in the model. This modeling choice of random effects has been proposed to solve the incidental parameters problem induced by including fixed effects in nonlinear regressions.^25,26 By controlling for zip code random effects, we partialled out average housing conditions in each neighborhood, reducing the risk that our results were affected by selection bias and sorting. We report a version of this model that does not include zip code random effects in Table C, without adjustment of laboratory reporting thresholds in Table A, with only venous measurement in Table B, and with data only from the second half of the sample period (2009–2016) in Table D (available as supplements to the online version of this article at http://www.ajph.org).

RESULTS

Table 1 shows summary statistics of relevant variables in our study group. Based on mother’s race, children in our sample were 66.8% White, 21.8% Black, and 3.8% Asian. More than a third of our sample (36.6%) lived in census block group with a median income less than $40 000, and almost half of the children (47.9%) had mothers with a high-school diploma or less education. Many children lived in old housing, with 45.6% living in housing built before 1930, 8.3% living in housing built between 1931 and 1950, and 30.2% living in housing built between 1951 and 1978. Figure 1b shows a map of median housing age by zip code, showing that the oldest housing is concentrated within the city of Chicago and in the western part of the state, while the Chicago and St Louis, Missouri, suburbs have predominantly newer housing. The current definition of high-risk zip codes for the purposes of lead screening appears to correlate well with the oldest housing. Only 3.4% of children lived within 30 meters of a major road and even fewer (0.3%) lived within 250 meters of a facility with any lead emissions.

TABLE 1—

Summary Statistics of Study Group: Children Aged 0 to 2 Years Tested for Lead Exposure, Illinois, 2001–2014

	Mean ±SD or No. (%)
Lead exposure
BLL by 2 y	2.44 ±2.80
BLL ≥ 10 µg/dL	26 781 (2.2)
BLL 5–9 µg/dL	106 431 (8.9)
BLL 4 µg/dL	68 647 (5.7)
Demographics
Race
White	800 920 (66.8)
Black	261 831 (21.8)
Asian	45 560 (3.8)
Male	612 880 (51.1)
Hispanic	341 492 (28.4)
On Medicaid at time of test	310 195 (25.8)
Low income in block group	333 010 (36.6)
Mother characteristics
Mother aged < 20 y at birth	132 251 (12.1)
Mother ≤ high school education	474 554 (47.9)
Housing at test address
Previous BLL
5–9 µg/dL	50 502 (7.9)
≥ 10 µg/dL	34 673 (5.4)
Year housing built
≤1930	292 250 (45.6)
1931–1950	53 347 (8.3)
1951–1977	193 585 (30.2)
≥ 1978	101 162 (15.8)
TRI emissions
Any TRI emissions within 250 m of residence	2 742 (0.3)
TRI emission > 1 ton within 250 m of residence	1 007 (0.08)
TRI emissions (tons) within 250 m of residence	0.07 ±4.42
TRI emissions (tons) within 500 m of residence	0.29 ±8.36
Distance to major road
< 30 m from address	31 391 (3.4)
30–100 m from address	45 913 (4.9)

Note. BLL = blood lead level; TRI = Toxic Release Inventory; µg/dL = micrograms per deciliter. The sample size was n = 1 201 801.

Among tested children, 2.2% had their highest BLL at 10 µg/dL or greater, 8.9% had it at 5 to 9 µg/dL, and 5.7% had it at 4 µg/dL. Table 2 presents our regression coefficients expressed in RRRs. Several demographic factors were associated with increased relative risk of above-threshold BLLs. Being Black was associated with increased relative risk of above-threshold BLLs. By contrast, Hispanic ethnicity was associated with lower risk of above-threshold BLLs, and being Asian had no association with relative risk of above-threshold BLLs.

TABLE 2—

Relative Risk of Exposure Sources and Demographic Characteristics on Different Blood Lead Levels by Age 2 Years for Children Born in Illinois: 2001–2014

Variable	4 µg/dL, RRR (95% CI)	5–9 µg/dL, RRR (95% CI)	≥ 10 µg/dL, RRR (95% CI)
Demographics
Male (Ref: female)	1.07 (1.04, 1.1)	1.13 (1.11, 1.15)	1.16 (1.12, 1.21)
Race/ethnicity (Ref: non-Hispanic White)
Black	1.36 (1.28, 1.44)	1.37 (1.29, 1.47)	1.34 (1.20, 1.49)
Hispanic	0.93 (0.88, 0.98)	0.87 (0.83, 0.92)	0.7 (0.64, 0.76)
Asian	0.97 (0.88, 1.06)	0.97 (0.88, 1.08)	1.06 (0.88, 1.27)
Child on Medicaid at time of test (Ref = non-Medicaid)	1.19 (1.11, 1.28)	1.19 (1.11, 1.27)	1.72 (1.43, 2.06)
Median census block group income < $40 000 (Ref: ≥ $40 000)	1.1 (1.06, 1.14)	1.08 (1.04, 1.13)	1.11 (1.03, 1.19)
Mother characteristics
Mother aged < 20 y at birth (Ref: aged ≥ 20 y)	1.11 (1.07, 1.16)	1.08 (1.05, 1.12)	0.98 (0.91, 1.04)
Mother ≤ high-school education (Ref: education > high school)	1.22 (1.19, 1.26)	1.27 (1.24, 1.31)	1.35 (1.27, 1.42)
Housing characteristics
Year housing built (Ref: ≥ 1978)
≤ 1930	2.28 (2.13, 2.45)	2.65 (2.44, 2.89)	4.06 (3.49, 4.71)
1931–1950	1.65 (1.53, 1.77)	1.67 (1.54, 1.82)	2 (1.70, 2.36)
1951–1977	1.26 (1.19, 1.34)	1.14 (1.06, 1.21)	0.99 (0.84, 1.16)
Previous BLL at address (Ref: no previous BLL ≥ 5 µg/dL)
5–9 µg/dL	1.27 (1.20, 1.34)	2.37 (2.20, 2.54)	1.83 (1.66, 2.02)
≥10 µg/dL	1.32 (1.23, 1.42)	1.79 (1.68, 1.90)	4.08 (3.69, 4.52)
TRI emissions and roadways
Cumulative past TRI emissions from air (Ref: no TRI emissions within 500 meters)
Within 250 meters (tons)	1.001 (1.0004, 1.0017)	1.0009 (1.0001, 1.0016)	1.0027 (1.0016, 1.0037)
Within 500 meters (tons)	0.9997 (0.999, 1.0005)	0.9996 (0.9991, 1.0001)	0.9987 (0.9979, 0.9996)
Distance from major roadways (Ref: > 100 meters)
Within 30–100 meters	1.01 (0.93, 1.09)	1.05 (0.99, 1.11)	0.98 (0.87, 1.09)
Within < 30 meters	1.09 (0.99, 1.19)	1.08 (0.97, 1.21)	1.10 (0.94, 1.29)

Note. BLL = blood lead level; CI = confidence interval; RRR = relative risk ratio; TRI = Toxic Release Inventory; µg/dL = micrograms per deciliter. Pearson goodness of fit: 8.00E-23. The model accounts for random effects in the child’s zip code and fixed effects in the child’s birth cohort, the combination of testing laboratory–year of test, and month of test.

The association between housing construction decade and RRR of above-threshold BLLs is plotted in Figure 2. Table 2 shows that, relative to housing built after 1978, housing built between 1950 and 1978 was not associated with increased relative risk of BLLs of 10 µg/dL or greater (RRR = 0.99; 95% CI = 0.84, 1.16) but was associated with increased relative risk of BLLs of 5–9 µg/dL (RRR 1.14; 95% CI = 1.06, 1.21) and BLLs of 4 µg/dL (RRR = 1.26; 95% CI = 1.19, 1.34). The association between housing age and risk of above-threshold BLLs was strongest for housing built before 1930, which was associated with increased risk of BLLs of 10 µg/dL or greater (RRR = 4.06; 95% CI = 3.49, 4.71), 5 µg/dL or greater (RRR = 2.65; 95% CI = 2.44, 2.89), and 4 µg/dL or greater (RRR = 2.28; 95% CI = 2.13, 2.45). We also found that previous instances of above-threshold BLLs at a given address were significantly associated with above-threshold BLLs among later residents. A previous resident having a BLL of 5 to 9 µg/dL increased the relative risk of current resident having a BLL of 5 to 9 µg/dL by 2.37- (95% CI = 2.2, 2.54) fold while a previous BLL of 10 µg/dL or greater increased the relative risk of a current resident having BLL of 10 or greater by 4.08- (95% CI = 3.69, 4.52) fold.

Moreover, proximity to state and interstate highways did not significantly increase the relative risk of above-threshold BLLs. Finally, an increase of cumulative air lead emissions within 250 meters of a child’s home was associated with higher relative risk of above-threshold BLLs at all thresholds (RRR for BLL ≥ 10 µg/dL = 1.0027; 95% CI = 1.0016, 1.0037 per ton of emissions).

Our results were stable in models using data without adjustment for laboratory reporting thresholds (Table A) and with venous tests only (Table B). Using a model that did not account for zip code random effects (Table C) produced larger RRRs for demographic variables and housing age. Using data from the second half of the sample period (Table D) generated consistent RRRs compared with the entire sample period, with the exception that being Black was not associated with increased relative risk of BLL of 10 µg/dL or greater, but continued to be associated with increased relative risk of BLLs of 4 µg/dL or greater.

DISCUSSION

As childhood lead levels continue to decrease around the nation, the CDC reference level is likely to be decreased,¹³ prompting state and local health departments to evaluate whether to lower the intervention threshold that triggers environmental inspection and case management.

Our exposure analysis revealed that the relative importance of exposure sources also changes as the exposure threshold decreases. We found that housing built between 1950 and 1978 did not appear to significantly increase the relative risk of BLLs of 10 µg/dL or greater but did increase the relative risk of BLLs of 5 to 9 µg/dL and 4 µg/dL. Therefore, states may want to take into account the prevalence of any pre-1978 housing when determining which zip codes are high risk under a lower intervention threshold. Focusing solely on housing built before 1950, as recommended by the current CDC guidelines,¹⁷ may result in missing children with BLLs above the lower intervention thresholds. Regardless of the intervention threshold, housing built before 1930 is associated with exceptionally high risk of above-threshold BLLs compared with housing built after 1978. The dramatic increase in relative risk of above-threshold BLLs for housing built before 1930 is consistent with surveys of the housing stock that indicate that lead hazards are much more prevalent in the oldest housing.²⁷ Thus, policymakers may need to develop nuanced programs aimed at the most at-risk children under any threshold. For example, housing built before 1930 could be targeted for preventive remediation, which was recently shown to reduce mean BLLs only in the highest-risk children,²⁸ while housing built between 1930 and 1978 could be targeted for blood lead screening.

We also showed that previous instances of both BLLs of 5 to 9 µg/dL and 10 µg/dL or greater at an address were associated with increased relative risk of above-threshold BLLs of the current occupants. During the study period, a BLL of 10 µg/dL would have triggered visits from case management, an environmental inspection, and mandated remediation of lead hazards if found. In reality, remediation of lead hazards often does not occur or may take many months to complete.²⁹ This lack of compliance may explain why children living in housing with previous BLLs of 10 µg/dL or greater remain at risk for exposure. Illinois has recently decreased the intervention threshold from 10 to 5 µg/dL, and our findings suggest that homes where BLLs of 5 to 9 µg/dL were previously detected may continue to put children at increased risk for BLLs above the new threshold if they are not remediated in a timely fashion.

We found that being Black, along with several other demographic variables such as having a mother with low education levels and living in a low-income area, was associated with increased relative risk of above-threshold BLLs, which is consistent with the existing literature.^13,30 We found that these demographic effects on BLLs persisted after we controlled for neighborhood characteristics and housing age, suggesting that even within a given zip code and for a given housing age, families from disadvantaged backgrounds may be living in the most dangerous housing. In the version of our model run on the subsample of children born between 2009 and 2016 (Table D) we found that being Black appeared to no longer be associated with BLL of 10 µg/dL or greater but continued to be associated with BLLs of 4 µg/dL or greater. This finding implies that while disparities in severe lead poisoning may have shrunk in recent years,¹³ disparities persist in rates of moderately elevated BLLs.

Our results indicate that current air lead emissions from industrial sources did not significantly contribute to BLLs except for children living very close to lead-emitting facilities. Because of the high weight of lead particles, the amount of lead deposited in soil from air emissions decays rapidly within 100 meters of the source.³¹ Therefore, most children in our sample may not live close enough to facilities for these effects to become significant. Roadway emissions constituted another important source of lead in soil because gasoline contained lead until 1996. We found that that proximity to state and interstate highways was not associated with increased risk of above-threshold BLLs in our sample. While significant lead hazards remain in urban soil,³² a more detailed model including data on historical traffic density may be required to analyze these effects of roadways on lead exposure. While older studies have found a strong correlation between proximity to roadways and BLL,³³ a recent analysis found that the association between proximity to roadways and BLLs is weakening over time,⁵ which could be related to a range of factors including decreasing soil lead concentrations, changes in the built environment, or changes in behavior. In addition, our analysis does not include data on historical traffic density, which may modify the association between above-threshold BLLs and proximity to state and interstate highways.

Limitations

This study is one of the largest population-level analyses of the importance of different lead exposure sources to date. However, there are several limitations to our work in addition to the lack of local road and traffic data. First, we did not examine the effects of lead in drinking water, a source of exposure that gained significant attention during the crisis in Flint, Michigan.³⁴ Lead in water may originate from lead in mains, service lines, or residential pipes,⁹ but there are little available data on each source. Given historical patterns of lead pipe use, we believe that the inclusion of housing vintage in our model partially accounts for the effects of water lead levels.³⁵ We performed a separate analysis including only children born in Chicago, where lead service lines were mandated until 1986,³⁵ and did not find a significant decrease in risk of BLLs of 5 µg/dL or greater for housing built after 1986 relative to housing built in 1978 (Figure E, available as a supplement to the online version of this article at http://www.ajph.org), suggesting that lead in these service lines did not make a significant contribution to children’s risk of above-threshold BLLs. This is consistent with previous studies that have found that lead in water accounted for less than 10% of total exposure among children with high lead levels.¹² Second, we conducted our analysis by using the address at the time of test, but not all of children’s exposure occurs in the home. We attempted to minimize this limitation by focusing on younger children aged 0 to 2 years. Third, while children’s Medicaid enrollment status was reported on the lead test, this variable may be incompletely reported given that 41.6% of Illinois children are on Medicaid, compared with only 25.8% in our sample.

Public Health Implications

Our results suggest that states may need to re-evaluate their screening guidelines for non–Medicaid-eligible children when they lower the intervention threshold because the determinants of lead exposure may change. When determining which zip codes are high-risk, focusing on pre-1950 housing is reasonable when the threshold is 10 µg/dL, but when the threshold decreases to 5 µg/dL, all children living in pre-1978 housing appear to be at risk for above-threshold BLLs. Regardless of the intervention threshold, children living in pre-1930 housing are at an exceptionally elevated risk for above-threshold BLLs, suggesting that policymakers should consider devoting additional resources to families living in these homes.

Using individual data geocoded at the test address and controlling for a variety of random and fixed effects, we show that even after controlling for the age of housing and the neighborhood they live in, Black children, children with low parental education level, and children living in low-income neighborhoods are at increased risk for above-threshold BLLs. Although there has been progress toward reducing racial disparities in above-threshold BLLs, our findings indicate that significant disparities in risk of high BLLs remain. This result highlights the continued need for considering race when allocating resources for childhood lead poisoning prevention.