Rita: A Quasi

Apr 21, 2010 - Prior to Hurricanes Katrina and Rita (HKR), significant associations were noted between soil lead (SL) and blood lead (BL) in New Orlea...
0 downloads 0 Views 1MB Size
Environ. Sci. Technol. 2010, 44, 4433–4440

New Orleans before and after Hurricanes Katrina/Rita: A Quasi-Experiment of the Association between Soil Lead and Children’s Blood Lead S A M M Y Z A H R A N , †,# H O W A R D W . M I E L K E , * ,‡,§ CHRISTOPHER R. GONZALES,| ERIC T. POWELL,| AND STEPHAN WEILER⊥ School of Global Environmental Sustainability, Colorado State University, Fort Collins, Colorado 80523-1784, Center for Bioenvironmental Research, Tulane University, 1430 Tulane Avenue SL-3, New Orleans Louisiana 70112, Department of Chemistry, Tulane University, New Orleans, Louisiana 70118, Lead Lab, Inc., P.O. Box 791125, New Orleans, Louisiana 70179-1125, and Department of Economics, Colorado State University, Fort Collins, Colorado 80523

Received February 19, 2010. Revised manuscript received March 30, 2010. Accepted April 5, 2010.

Prior to Hurricanes Katrina and Rita (HKR), significant associations were noted between soil lead (SL) and blood lead (BL) in New Orleans. Engineering failure of New Orleans levees and canal walls after HKR set the stage for a quasiexperiment to evaluate BL responses by 13 306 children to reductions in SL. High density soil surveying conducted in 46 census tracts before HKR was repeated after the flood. Paired t test results show that SL decreased from 328.54 to 203.33 mg/ kg post-HKR (t ) 3.296, p e 0.01). Decreases in SL are associated with declines in children’s BL response (r ) 0.308, p e 0.05). When SL decreased at least 1%, median children’s BL declined 1.55 µg/dL. Declines in median BL are largest in census tracts with g50% decrease in SL. Also individual BL in children was predicted as a function of SL, adjusting for age, year of observation, and depth of flood waters. At the individual scale, BL decreased significantly in post-HKR as a function of SL, with BL decreases ranging from b ) -1.20 to -1.65 µg/dL, depending on the decline of SL and whether children were born in the post-HKR period. Our results support policy to improve soil conditions for children.

Introduction On August 29, 2005, Hurricane Katrina and its storm surge triggered a catastrophic engineering failure of the levee system * Corresponding author phone: 504 988-3889; fax: 504 988-6428; e-mail: [email protected] or [email protected] (preferred). † Center for Disaster and Risk Analysis, Department of Sociology, Colorado State University. ‡ Center for Bioenvironmental Research, Tulane University. § Department of Chemistry, Tulane University. | Lead Lab, Inc. ⊥ Department of Economics, Colorado State University. # Current address: Center for Disaster and Risk Analysis, Department of Sociology, Colorado State University, Fort Collins, Colorado 80523-1784. 10.1021/es100572s

 2010 American Chemical Society

Published on Web 04/21/2010

and walls of several outfall canals that inundated 80% of New Orleans, flooding some areas to a depth of more than 3 m, and the loss of over 1400 lives (1, 2). Still reeling from Katrina, Hurricane Rita made landfall on September 23, 2005, and rising waters again breached the already compromised levees and canals and inundated the Ninth Ward and surrounding Gentilly neighborhoods of New Orleans. Hurricanes Katrina and Rita (HKR) and the flooding of New Orleans set the stage for an unexpected chance to evaluate the relationship between environment and health. Census tracts (or neighborhoods) are delineated by boundaries which conform to recognizable physical, cultural, economic, and demographic characteristics and have a population range of 2500-8000 inhabitants (3). High density soil surveying and mapping were conducted at the census tract scale prior to HKR. Soils in metropolitan New Orleans were known to carry high concentrations of lead (Pb), and the maps revealed that 71 of 286 census tracts had median soil lead of g400 mg/kg and 10 census tracts had a median SL of g1000 mg/kg (4, 5). A pilot soil project, begun prior to HKR on 25 Pb-contaminated properties covered with clean Mississippi River sediments, noted relatively small changes on the same properties after HKR (6). Although multiple toxins were described after HKR (7), this article focuses on soil Pb. Studies indicated that soil Pb (SL) of a census tract is an important determinant of children’s blood Pb (BL), and SL concentrations appear to influence educational outcomes (8-12). The specific objective of this study is to analytically exploit pre- and post-HKR as a quasi-experiment for investigating the empirical relationship between SL and children’s BL. Our investigation is organized into four sections. First, we describe data collection and measurement of the key variables. Second, we describe the analytic logic of statistical tests performed to estimate the census tract change of SL and BL before and after HKR. Third, we present statistical results, beginning with aggregate analyses of SL and BL and ending with an estimation of the BL in individual children as a function of SL conditions. Finally, in the discussion we consider the environmental health relevance of our findings and suggest policy initiatives to anchor gains in the wellbeing of children.

Methods Data Collection and Variable Operations. Figure 1 shows the location (along with median water depth on August 31, 2005) of 46 census tracts where soil samples were collected and analyzed in two phases. Pre-HKR soil data are from the New Orleans Soil Lead Survey II completed in 2000 (4). The postHKR soil data are from the New Orleans Soil Lead Survey III, completed in 2006. Samples were collected at the rate of 19 samples per census tract from the top 2.5 cm of the soil surface within residential neighborhoods of metropolitan New Orleans (4). The extraction procedure involved room temperature leachate methods using 1 M nitric acid (HNO3), a scheme that correlates with total methods (13). The extraction procedure has the advantage of more closely resembling physiological conditions compared with extraction methods based on high temperatures and concentrated HNO3. The SL extraction protocol is the same for Survey III as for Survey II and requires mixing 0.4 g of dry and sieved (#10 USGS-2 mm) soils with 1 M HNO3 and agitated at slow speed on an Eberbach shaker for 2 h at room temperature (∼22 °C). The extract is then centrifuged (10 min at 1600 × g) and filtered using Fisherbrand P4 paper. The extract is stored in 20 mL of polypropylene scintillation vials until analyzed. A Spectro Analytical Instruments CIROS CCD Inductively VOL. 44, NO. 12, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4433

FIGURE 1. Map (with water depth on August 31, 2005) of the 46 census tracts (with identification number) collected post-Hurricanes Katrina and Rita (HKR). The same census tracts were matched with pre-HKR soil sampling results as illustrated in Figures 2 and 3.

FIGURE 2. Median soil Pb by sampled census tract in metropolitan New Orleans, pre- and post-HKR. The numbers on the horizontal axis refer to the census tracts as shown in Figure 1. The paired t test results are t ) 3.296, p e 0.01. Coupled Plasma Atomic Emission Spectrometer (ICP-AES) is used to analyze the metals in each sample. The ICP-AES is calibrated with NIST traceable standards, and a laboratory reference, at a rate of 1 per 15 samples, is analyzed during each run. Internal laboratory references included one low SL sample from New Orleans City Park and one high SL 4434

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 12, 2010

sample from the junction of Elysian Fields and Interstate 10 in the inner city of New Orleans. Duplicate extractions are included for every 15 samples. Sample collections were stratified using U.S. Census Bureau maps as a guide, and this approach provided the wealth of amassed public census tract information for each

FIGURE 3. Median blood Pb by sampled census tracts in metropolitan New Orleans, pre- and post-HKR. The numbers given on the horizontal axis refer to census tracts shown in Figure 1. The paired t test results are t ) 5.088, p e 0.001. of the sampled census tracts (14). Both phases of collection resulted in 874 (total n ) 1748) surface samples collected from 46 census tracts. Census tracts are assigned the median result of SL samples collected in each study period, expressed in mg/kg units. Our analyses of the relationship between SL and children’s BL focus on the 46 census tracts observed in both collection periods. The BL data set of children e6 years old was organized by the Louisiana Childhood Lead Poisoning Prevention Program (LACLPPP), Louisiana Office of Public Health. The LACLPPP program follows the Centers for Disease Control and Prevention (CDC) protocols for collection, preparation, and analysis of BL results (15). BL results were addressmatched and grouped by census tract. BL expressed in µg/ dL for 13 306 children are analyzed. A total of 11 191 children were measured for BL in the pre-HKR period, and 2115 children are measured in the post-HKR period. In addition to BL values, LACLPPP data contains information on the age (in months) of the child when the blood sample was taken. Age is a reasonable proxy for length of exposure to SL conditions, especially in New Orleans where the soils rarely freeze and outdoor play is possible throughout the year. We use age information as a control variable in statistical analyses. Descriptive statistics and variable definition statements on SL, BL, age in months, and other variables are presented in Table 1. Analytic Logic. Our investigation of the relationship between SL and BL begins with aggregate-level analyses of change in median child BL levels in New Orleans neighborhoods as a function of change in median SL. First, we analyze change in neighborhood SL before and after HKR using a paired t test to evaluate whether observed changes in neighborhood SL supersede statistical chance. The same logic is used to estimate change in median BL in children of the same neighborhoods. Information on both census tract SL and child BL data are presented below. Next, we correlate percent change in neighborhood SL (before and after HKR) with percent change in median BL of children (measured before and after HKR). Insofar as neighborhood SL affects children’s BL, and assuming HKR changed the SL in New Orleans, then we expect to see a

significant positive association between median SL and median BL levels in children before and after HKR. Moreover, this expected positive association between ∆SL and ∆BL ought to be highest for pre- and post-HKR comparisons of children 36 months or less. Provided that flood waters from HKR significantly changed neighborhood SL, we expect a higher correlation between ∆SL and ∆BL for children under age 3 because such children, born in the postHKR period, if SL changed are exposed to different SL conditions. Likewise, we logically expect a less significant relationship between ∆SL and ∆BL for pre- and post-HKR comparisons of children older than 3 years of age because such children were exposed to different pre-HKR SL conditions during infancy. Next, to specify the relationship between neighborhood SL and median child BL more precisely, we conduct a series of ordinary least-squares regression (OLS) models. In addition to estimating the magnitude of change in neighborhood SL levels resulting from HKR, our regression models control for median age and depth of flood waters (that could plausibly account for variation in median child BL levels). We use a regression-based difference-in-differences (DD) procedure, comparing the median BL levels of children in neighborhoods experiencing a change in neighborhood SL before and after HKR with the median child BL levels in neighborhoods not experiencing a change in SL conditions. We identify census tracts that experienced more than a 1% and more than a 50% change in median SL over the two measurement periods. The logic of our DD analysis is straightforward. Let t ) 0 denote the pre-HKR period and t ) 1 denote the post-HKR measurement period, and yit denote the median child BL for neighborhood i in period t. A regression-based estimator is modeled as yit ) β0 + β1xi + β2πt + β3xi*πt + εit

(1)

where xi is a dummy variable that assumes a value of 1 if a census tract experienced a measurable decline in median SL (of 1%+ and 50%+) and 0 if the census tract did not experience a measurable change in SL condition over time, and πt is a dummy variable taking a value of 1 if the census VOL. 44, NO. 12, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4435

TABLE 1. Variable Operations and Descriptive Statistics variable

operation

median age age water depth year post-Katrina soil lead ∆ (1 ) ∆ > 1%)

soil lead ∆ (1 ) ∆ > 50%) DD estimator (1 ) post, ∆ > 1%) DD estimator (1 ) post, ∆ > 50%) percent change soil lead

percent change median blood lead median blood lead individual blood lead

Predictors midpoint age of observed children in a census tract age of child in months estimated neighborhood flood level after levee failure in meters as of 08/31/2005 values range from 2000 to 2008 1 ) observation in the post-Katrina and Rita period; 0 ) pre-Katrina and Rita observation 1) neighborhood experienced at least a 1% decline in SL composition from pre- to post-Katrina/Rita measurement periods; 0 ) did not 1 ) neighborhood experienced at least a 50% decline in SL composition from pre- to post-Katrina/Rita measurement periods; 0 ) did not 1 ) if observation is in the post-Katrina/Rita period in a census tract experiencing observed SL decline; 0 ) if not 1 ) if observation is in the post-Katrina/Rita period in a census tract experiencing at least a 50% decline in SL; 0 ) if not percentage change in SL before and after Hurricanes Katrina and Rita Response Variables percent change median BL of children in a census tract before and after Hurricanes Katrina and Rita midpoint BL of children in a census tract µg/dL units observed BL of child in µg/dL units

tract observation is in the post- HKR period and 0 if observed in the pre-HKR period. The DD estimator is β3 (the coefficient of interaction between xi and πt) that assumes a value of 1 only for neighborhoods that witnessed a measurable decline in median SL and were observed in the post-HKR measurement period. The DD estimator is the variable of analytic interest for our investigation. A statistically significant negative coefficient in our DD estimator will indicate that the SL altering effects of HKR produced measurable change in the median BL levels of children in sampled New Orleans neighborhoods. Such an observed effect provides evidence that neighborhood SL conditions partially explain median BL levels in children. As with our bivariate analyses of ∆SL and ∆BL, we render separate DD analyses by age group (over and under 36 months of age), and if a change occurred we expect to observe stronger effects for children born in the post-HKR period. Finally, we examine the SL altering effects of HKR on the BL levels of 13 306 children. This analysis is at the individual level, involving a random effects regression procedure that allows each census tract to have its own intercept that yields a weighted average of between and within census tract effects. A model of the BL yij of child i in census tract j is specified as yij ) β1 + β2x2ij + ... + βpxpij + εij

(2)

where x2ij through xpij are parameters and εij is a residual. The random effects model divides the residual term into two componentssa census tract-specific error component (ζj), which is constant across blood lead levels, and a childspecific error component (ij) which varies between children and census tract: εij ) ζj + ij 4436

9

mean

(3)

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 12, 2010

SD

min

max

27.739

6.974

14

55

30.687

16.599

6

72

1.087

1.081

0

2.308

2002.76

3.66

2000

2008

0.50

0.50

0

1

0.630

0.485

0

1

0.282

0.453

0

1

0.315

0.467

0

1

0.141

0.350

0

1

-11.772

47.873

-78.962

86.099

-31.077

21.359

-74.0

16.0

4.292

1.885

1.3

6.010

5.229

0

11.0 112

Substituting for εij into regression model (1), we obtain the random intercept model with parameters yij ) β1 + β2x2ij + ... + βpxpij + ζj + ij ) (β1 + ζj) + β2x2ij + ... + βpxpij + ij

(4)

The census tract-specific error component ζj can be thought of as the combined effects of omitted neighborhood characteristics or unobserved heterogeneity (16). That is, the error component usefully captures information on variables not observed in our model, allowing us to more confidently estimate the SL altering effects of HKR on the BL levels of individual children. Our individual-level modeling efforts also examine children by age group, with our logical expectation being that more sizable BL declines will be observed in children 36 months of age (exposed to higher neighborhood SL conditions in the pre-HKR period).

Results Figure 2 shows median SL for 46 census tracts in New Orleans, measured before and after HKR. In 29 of 46 neighborhoods examined, we observe decline in median SL. In the postHKR period, 6 of 46 census tracts have SL levels that exceed the EPA regulatory standard of 400 mg/kg, as compared to 15 of 46 neighborhoods exceeding this standard in the preHKR period. Across sampled census tracts the average median SL declined 45.59%, going from 328.54 mg/kg (SD ) 373.70) to 203.33 mg/kg (SD ) 178.38). Paired t test results indicate that the decline in the median SL is statistically significant (t ) 3.296, p e 0.01). Figure 3 shows median BL levels in children for sampled census tracts before and after HKR. Like Figure 2, we plot census tracts on the horizontal axis and median BL levels of

FIGURE 4. Change in median census tract soil Pb and blood Pb by age group pre- and post-HKR. Panel A includes all children r ) 0.308, p e 0.05. Panel B includes only children 36 months or less, r ) 0.335, p e 0.05. Panel C includes children older than 36 months, r ) -0.11, p e 0.10.

children on the vertical axis. As expected, results show that median child BL levels declined in the majority of census tracts (37 of 46) in the post-HKR period. In fact, average median BL levels in children declined 32.85% (5.14-3.45 µg/ dL), approximate to the percentage decline observed in neighborhood SL. Paired t tests, comparing pre- and postHKR median BL levels in children, show that observed neighborhood differences through time are statistically significant (where t ) 5.09, p e 0.001). Figure 4 combines observations from Figures 2 and 3. Three panels are presented, corresponding to different age groups compared through time. In each graph, ∆SL is plotted on the x-axis and ∆BL on the y-axis. Panel A is for all children. The scattering of dots (with best-fit quadratic) shows that as the percent change in median SL in a neighborhood decreases, so too does the percent change in neighborhood median BL of children (r ) 0.308, p e 0.05). In panel B, the percent change in median BL for children 36 months. As hypothesized, the slope of the association between ∆SL and ∆BL (for children >36 months) is indistinguishable from zero (r ) -0.11, p g 0.10). The logic of this corroborated

expectation is that children older than 3 years of age were exposed to higher neighborhood SL levels in infancy. Next, we render a series of OLS regression models predicting median BL in children as a function of HRKinduced change in neighborhood SL, controlling for median age and estimated depth of water inundation. Water depth is measured as neighborhood flood level after levee failure in meters as of August 31, 2006 as indicated in Figure 1. This parameter functions as a proxy for the destruction experienced by a neighborhood. Also, insofar as destruction experienced is related to habitability of a neighborhood, the water depth variable (together with Katrina/Rita period binary variable) provides added statistical control for unmeasured out-migration dynamics that may weaken comparability of child populations before and after HKR. In Table 2, six models are shown, corresponding to different age groupings and levels of measured decline in census tract SL. Column 1 is an OLS model for all children. The relevant test statistic in column 1 is the DD estimator (1 ) post, >1% decline). Results show that median BL is 1.55 µg/dL lower (p e 0.05) in children measured in the postHKR period residing in neighborhoods with a 1%+ decline in SL levels. Columns 2 and 3 subdivide child population by age. As hypothesized, HKR-induced decline in SL levels produced noticeably different BL outcomes for children 36 months. Median VOL. 44, NO. 12, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4437

TABLE 2. Ordinary Least-Squares Regression Coefficients Predicting Median Neighborhood BL in Children, before and after Hurricanes Katrina and Rita

Katrina/Rita period (1 ) post) soil lead ∆ (1 ) ∆ > 1%) soil lead ∆ (1 ) ∆ > 50%) DD estimator (1 ) post, ∆ > 1%) DD estimator (1 ) post, ∆ > 50%) median age (months) water depth (m) constant F R2 root MSE N

1. median BL all children

2. median BL < 36 months

-0.799 (0.577)

-0.652 (0.580)

-1.854a (0.468) -1.127a (0.367) -1.128a (0.346) -1.560a (0.413)

1.280a (0.477)

0.962b (0.415)

1.056c (0.587)

-

3. median BL > 36 months

-

4. median BL all children

-

-

-

-

-

0.109a (0.030) -0.443a (0.139) 1.811a (0.686) 13.87 0.417 1.410 90

0.071b (0.0347) -0.321b (0.157) 3.761a (0.944) 14.27 0.359 1.718 89

Robust standard errors in parentheses. p < 0.01. errors in parentheses. p < 0.1.

b

-

a

2.063 (0.464)

0.0742b (0.029) -0.436a (0.144) 2.789a (0.735) 13.47 0.392 1.512 92

6. median BL > 36 months

-

a

-1.548b (0.686) -1.493b (0.675) -0.136 (0.681)

a

5. median BL < 36 months

b

1.923 (0.424)

-

1.470 (0.700)

-

-

-2.278 (0.620) -1.633 (0.546) -1.411 (0.874) a

a

0.070a (0.0263) -0.404a (0.135) 3.091a (0.695) 27.25 0.457 1.428 92

0.093a (0.0247) -0.363a (0.126) 2.216a (0.581) 22.98 0.489 1.320 90

Robust standard errors in parentheses. p < 0.05.

c

0.082b (0.032) -0.357b (0.208) 3.165a (0.918) 9.27 0.362 1.713 89 Robust standard

TABLE 3. Random Effects Generalized Least-Squares Regression Coefficients Predicting BL in Children, before and after Hurricane Katrina

Katrina/Rita period (1 ) post) soil lead ∆ (1 ) ∆ > 1%) soil lead ∆ (1 ) ∆ > 50%) DD estimator (1 ) post, ∆ > 1%) DD estimator (1 ) post, ∆ > 50%) age (months) water depth (m) year constant Wald χ2 R2within R2between R2overall N

1. BL all children

2. BL < 36 months

3. BL > 36 months

4. BL all children

5. BL < 36 months

6. BL > 36 months

-0.871a (0.196)

-0.719a (0.232)

-1.212a (0.352)

-1.249a (0.175)

-1.062a (0.208)

-1.559a (0.300)

1.554a (0.420)

1.523a (0.446)

1.468a (0.352)

-

-

-

-

-

-

-1.195a (0.169)

-1.132a (0.201)

-1.089a (0.305)

-

-

-

-

-

-

-1.649a (0.219)

-1.637a (0.275)

-1.548b (0.343)

0.0336a (0.002) -0.344b (0.174) -0.145a (0.0313) 294.25a (62.73) 289.3 0.045 0.319 0.070 13 306

0.137a (0.006) -0.345c (0.188) -0.138a (0.0379) 278.26a (75.94) 907.79 0.083 0.310 0.102 8400

-0.027a (0.008) -0.322 (0.217) -0.184a (0.053) 375.61a (105.90) 460.04 0.047 0.254 0.061 4906

0.034a (0.00235) -0.298c (0.160) -0.141a (0.0314) 286.46a (62.82) 977.83 0.046 0.355 0.076 13 306

0.137a (0.00576) -0.289c (0.165) -0.134a (0.0380) 271.26a (76.15) 914.61 0.084 0.363 0.110 8400

0.028a (0.0301) -0.273 (0.208) -0.178a (0.053) 364.97a (105.76) 455.42 0.048 0.266 0.069 4906

a Robust standard errors in parentheses. p < 0.01. errors in parentheses. p < 0.1.

b

9

2.184a (0.512)

Robust standard errors in parentheses. p < 0.05.

neighborhood BL decreased 1.49 µg/dL for children measured after HKR and living in census tracts with measured decline in SL levels (p e 0.05). Columns 4, 5, and 6 (in Table 2) report coefficients predicting median child BL for census tracts registering a 50%+ decline in SL conditions. Again, the key test parameter is the DD estimator (1 ) post, >50% decline). As expected, median BL levels are significantly lower in high SL declining census tracts. Results indicate that median BL levels declined by 2.28 µg/dL (95% CI, 1.04-3.51) for children in neighborhoods witnessing a HKR-induced 50%+ decline in SL. The marginal effect of a HKR-induced 50%+ decline in SL is a 53.06% (95% CI, 24.73-81.40%) decline in median child BL. Before moving ahead, the behavior of our median age control variable is also worth noting. Across all models executed, we find that median child BL increases significantly with unit increases in the median age of children in a neighborhood. This is sensible because age, in a city like New Orleans with 4438

2.116a (0.497)

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 12, 2010

1.978a (0.624)

c

Robust standard

notoriously high concentrations of SL, functions as a proxy for length of exposure. In the final step of our analysis, we analyze the neighborhood SL altering effects of HKR on the BL of 13 306 individual children. As in Table 2, we report six random effects regression models that modulate age structure and the magnitude of SL decline. In these models we also adjust for the year a child’s BL is observed, allowing us to control for known secular declines in child BL in our specification of the relationship between individual child BL and HKR-induced change in neighborhood SL. In Table 3 we report random effects coefficients predicting individual BL levels in children. In column 4, for example, results show that children residing in neighborhoods experiencing a HKR-generated 50%+ decline in SL had their BL levels reduced by 1.65 µg/dL (95% CI, 1.12-2.18 µg/dL). The estimated reduction in child BL of 1.65 µg/dL controls for child age and time (year), both factors of which are known covariates of child BL. A similarly

significant decline in BL is observed for children under 36 months of age (b ) -1.64, 95% CI, -2.30 to 0.98). Interestingly, at the individual level, an HKR-caused 50%+ decline in neighborhood SL also causes a statistically significant reduction in children over 36 months of age (b ) -1.55, p e 0.05). This result suggests that the health benefits of a substantial decline in neighborhood SL extend even to children born in the pre-HKR period.

(23). Fortunately, natural resources are readily available in New Orleans in the form of clean (5 mg Pb/kg) alluvium derived from soils eroding from the Mississippi River Watershed and flowing through the city at an average rate of ∼272 t (300 U.S. tons) per min (5, 24). In addition to the pilot project to emplace clean soils on private properties (5, 6), a project is underway in New Orleans to establish lead-safe soils in outdoor play areas at childcare centers (25).

Discussion

Acknowledgments

Hurricanes Katrina and Rita storm surges and subsequent flooding of New Orleans had the measurable effect of significantly altering Pb contaminated soil surfaces. Our results show that, on average, median census tract SL declined by about 46%. Moreover, we find that the decline in census tract SL correlates significantly with observed decreases in median census tract BL in children. Ordinary least-squares regression results show that the relationship between median SL and median BL holds with the addition of statistical controls. As hypothesized, we find that the main beneficiaries of declining SL are children born in post-HKR compared to children exposed to significantly higher SL conditions preHKR. Results from individual-level analyses of BL of 13 306 children corroborate neighborhood-level analyses, and furthermore, HKR-induced change in neighborhood SL decreased BL levels in children by a remarkable 1.65 µg/dL. Overall, our results corroborate and extend existing environmental health literature which indicated a strong association between SL and BL (8, 9, 11, 17). Furthermore, the results indicate that while the standard regulatory model of abating lead-based paint may gradually reduce BL levels in children, such policy efforts may nevertheless be too little and too slow because young children exposed from all sources of Pb dust accumulated in soil suffer lasting cognitive damage that undermines education and future economic wellbeing (10, 12). These stark facts underscore the policy implications of these findings. Disasters with such clear human and economic costs nevertheless have silver linings precisely because previous policy regimes poorly addressed environmental risks to vulnerable populations. The policy irony is that the HKRs’ flood waters had a measurably positive effect on neighborhood SL conditions, in literal waves that no other policy intervention could probably match. This irony highlights not only the undeniable interaction between poor economic circumstances and consequent poor health outcomes, but also how this interaction could result in a self-reinforcing cycle of vulnerable socioeconomic conditions, lead-tainted environments, and poor scholastic achievement (12, 18). The apparent response of children’s BL to changes in SL indicates the need for decreasing U.S. Pb standards for residential soils. The current SL standards are 400 mg/kg for bare soil at play sites and 1200 mg/kg for nonplay areas (19). Previously, we noted a BL change of 1.4 µg/dL per 100 mg/kg change in SL below 100 mg/kg, and a BL change of 0.32 µg/dL per 100 mg/kg change in SL above 300 mg/kg (11). Assuming the CDC BL guideline g10 µg/dL, a SL of 80 mg/kg was described as protective for most New Orleans children (9). Current research indicates BL levels as low as 2 µg/dL have adverse health effects to multiple organ systems (i.e., cardiovascular, kidney, and nervous systems) (20-22). Lowering the BL guidelines and including a margin of safety suggests a SL standard substantially lower than 80 mg/kg to protect children living in residential neighborhoods. The results support an alternative policy to systematically improve soils within children’s play areas. Precedence for such a policy has been established by Norway’s national program to test and renovate soils at all childcare centers, elementary schools, and public parks in the 10 largest cities

Grants: A U.S. HUD grant, Lead Technical Study LAHBC000203 to Xavier University of Louisiana funded the post-Katrina “Survey III” soil collection conducted from April 4 through June 5, 2006 at Xavier University of Louisiana’s College of Pharmacy as part of the Environmental Health and Toxicology program. After Hurricanes Katrina/Rita, the program was terminated and the laboratory was moved to Tulane University where chemical analysis was funded from July 24 through September 6, 2006 by a Cooperative Agreement between the Association of Minority Health Professions Schools and ATSDR for the Environmental Health, Health Services, and Toxicology Research Program grant 03040 to Xavier University. A Greater New Orleans FoundationEnvironmental Fund grant to Lead Lab, Inc. funded the LeadSafe Soil at Childcare Centers Project. Blood Lead Data: Special thanks are given to Ngoc Huynh, Louisiana Childhood Lead Poisoning Prevention Program (LACLPPP), and this publication was thus also supported by Cooperative Agreement EH0660205CONT10 from the Centers for Disease Control and Prevention (CDC). The views are the authors’ and not necessarily those of the funding agencies.

Literature Cited (1) Heerden, I.; Bryan, M. The Storm: What Went Wrong and Why During Hurricane Katrina-the Inside Story from One Louisiana Scientist; Viking Press: New York, 2006. (2) Louisiana Department of Health and Hospitals. Reports of Missing and Deceased. 2006. http://www.dhh.louisiana.gov/ offices/page.asp?ID)192&Detail)5248. (3) U.S. Census Bureau, Geography Division. Census Tracts and Block Numbering Areas. http://www.census.gov/geo/www/ cen_tract.html. (4) Mielke, H. W.; Gonzales, C.; Powell, E.; Mielke, P. W., Jr. Changes of multiple metal accumulation (MMA) in New Orleans soil: Preliminary evaluation of differences between survey I (1992) and survey II (2000). Int. J. Environ. Res. Public Health 2005, 2 (2), 308–313, doi:10.3390/ijerph2005020016. (5) Mielke, H. W.; Powell, E. T.; Gonzales, C. R.; Mielke, P. W., Jr.; Ottesen, R. T.; Langedal, M. New Orleans soil lead (Pb) cleanup using Mississippi River alluvium: Need, feasibility and cost. Environ. Sci. Technol. 2006, 40 (08), 2784–2789. (6) Mielke, H. W.; Powell, E. T.; Gonzales, C. R.; Mielke, P. W., Jr. Hurricane Katrina’s impact on New Orleans soils treated with low lead Mississippi River alluvium. Environ. Sci. Technol. 2006, 40 (24), 7623–7628. (7) Abel, M.; Presley, S.; Rainwater, T.; Austin, G.; McDaniel, L.; Marsland, E.; Leftwich, B.; Anderson, T.; Kendall, R.; Cobb, G. Spatial and temporal evaluation of metal concentrations in soils and sediments from New Orleans, Louisiana, USA, following Hurricanes Katrina and Rita. Environ. Toxicol. Chem. 2007, 26 (10), 2108–2114. (8) Mielke, H. W.; Dugas, D.; Mielke, P. W.; Smith, K. S.; Smith, S. L.; Gonzales, C. R. Associations between lead dust contaminated soil and childhood blood lead: A case study of urban New Orleans and rural Lafourche Parish, Louisiana, USA. Environ. Health Perspect. 1997, 105 (9), 950–954. (9) Mielke, H. W.; Smith, M. K.; Gonzales, C. R.; Mielke, P. W., Jr. The urban environment and children’s health: Soils as an integrator of lead, zinc and cadmium in New Orleans, Louisiana, USA. Environ. Res. 1999, 80 (2), 117–129. (10) Mielke, H. W.; Berry, K. J.; Mielke, P. W., Jr.; Powell, E. T.; Gonzales, C. R. Multiple metal accumulation as a factor in learning achievement within various New Orleans communities. Environ. Res. 2005, 97 (1), 67–75. VOL. 44, NO. 12, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4439

(11) Mielke, H. W.; Gonzales, C. R.; Powell, E.; Jartun, M.; Mielke, P. W., Jr. Nonlinear association between soil lead and blood lead of children in metropolitan New Orleans, Louisiana: 20002005. Sci. Total Environ. 2007, 388, 43–53. (12) Zahran, S.; Mielke, H. W.; Weiler, S.; Berry, K.; Gonzalez, C. Children’s blood lead and standardized test performance response as indicators of neurotoxicity in metropolitan New Orleans elementary schools. NeuroToxicol. 2009, 30, 888–897. (13) U.S. EPA. Urban Soil Lead Abatement Demonstration Project; Volume I: Integrated Report. EPA/600/P-93/00aF. Research Triangle Park, National Center for Environmental Assessment, 1996; Chapter 3, pp 4-6. (14) U.S. Census Tracts and Block Numbering Areas, New Orleans, LA. Table 32, selected structural characteristics of housing units: 1990 and summary tape file 3A, Louisiana 040, Lafourche Parish 050, Census tracts 140. U.S. Department of Commerce, Economics and Statistics Administration, Bureau of Census, Washington, DC, 1993; p 565-615. (15) U.S. CDC. Preventing Lead Poisoning in Young Children; Centers for Disease Control and Prevention: Atlanta, GA, 1991. (16) Rabe-Hesketh, S.; Skrondal, A. Multi-level and longitudinal modeling using Stata; Stata Press: College Station, TX, 2008. (17) Johnson, D. L.; Bretsch, J. K. Soil lead and children’s blood lead levels in Syracuse, NY, USA. Environ. Geochem. Health 2002, 24, 375–385. (18) Campanella, R.; Mielke, H. W. Human geography of New Orleans’ urban soil lead contaminated geochemical setting. Environ. Geochem. Health 2008, 30 (6), 531–540, doi: 10.1007/s10653008-9190-9. (19) Agency for Toxic Substances and Disease Registry. Case Studies in Environmental Medicine (CSEM) Lead Toxicity. What Are the

4440

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 12, 2010

(20)

(21)

(22)

(23)

(24)

(25)

U.S. Standards for Lead Levels? p 23. http://www.atsdr.cdc.gov/ csem/lead/docs/lead.pdf. Navas-Acien, A.; Guallar, E.; Silbergeld, E. K.; Rothenberg, S. J. Lead exposure and cardiovascular diseasesA systematic review. Environ. Health Perspect. 2007, 115, 472–482, doi:10.1289/ ehp.9785. Fadrowski, J. J.; Navas-Acien, A.; Tellez-Plaza, M.; Guallar, E.; Weaver, V. M.; Furth, S. L. Blood lead level and kidney function in U.S. adolescents: The third national health and nutrition examination survey. Arch. Intern. Med. 2010, 170 (1), 75–82. Jusko, T. A.; Henderson, C. R., Jr.; Lanphear, B. P.; Cory-Slechta, D. A.; Parsons, P. J.; Canfield, R. L. Blood lead concentrations