Geospatial assessment of racial/ethnic composition, social vulnerability, and lead service lines in New York City

INTRODUCTION

Ongoing racial/ethnic disparities in water lead concentrations, lead exposure from household dust and paint, and child blood lead levels (a measure of internal dose) in the US are well documented and are of great public health concern.^1-4 Substantial epidemiologic and toxicologic evidence supports that there is no safe level of lead exposure, especially for children.^5-8 Large scale infrastructural investments are needed to further reduce disparities and to decrease overall exposure to historical sources of lead, many of which are still present in the built environment.

Lead enters public drinking water via corrosion of distribution and plumbing system materials that contain lead, including solder, fixtures, faucets, and lead service lines that connect homes to public drinking water distribution systems (acidity, mineral content, and other source water characteristics also influence corrosivity).⁹ Although the US EPA Lead and Copper Rule (LCR) requires public water systems to conduct routine compliance monitoring for lead at taps throughout the distribution system and to intervene to reduce lead corrosion, current regulations are inadequate to reduce water lead exposure and eliminate racial/ethnic disparities in water lead exposure. For example, the current LCR fails to establish a health-based standard, enables the biased selection of taps and homes selected for sampling (resulting in documented racial disparities in reported compliance monitoring sample measurements), and does not require the reporting of absolute measured concentrations at all monitoring points (only higher percentile values are reported when measured above the action level) even though water lead concentrations are highly variable across season and within water distribution systems.^1,10-14 Large scale replacement of lead water service lines, which are significant sources of lead to drinking water even when source water contains no lead, remains a critically needed intervention to reduce public drinking water lead exposures and disparities.⁹

In the fall of 2021, the US Congress passed the Infrastructure Investment and Jobs Act, which allocates $55 billion to improving drinking water quality across the US.¹⁵ New York state (NYS) expects to receive $115 million in 2022 alone to support the replacement of lead water service lines.¹⁶ Identifying communities with higher prevalence of lead service lines is critically needed to help inform equitable prioritization of lead service line replacement throughout New York City (NYC) and NYS. Under NYS’s current lead service line replacement program, municipalities are prioritized for funding if those municipalities are low income (median household income <150% of regional average), are likely to have a high number of lead services lines (500 or more homes constructed before 1939), and have a high percentage of children with elevated blood lead levels (>0.5% children with blood level >5 μg/dL, although the Centers for Disease Control and Prevention’s (CDC) level is 3.5 μg/dL).¹⁷ Although the location of lead service lines is largely unknown for most utilities, the NYC Department of Environmental Protection (NYCDEP) recently published a database of water service line material for city buildings (water service line material is recorded as Potential Lead, Unknown, Not Lead, and Not Applicable).¹⁸ The database provides a unique and powerful opportunity to evaluate neighborhood level sociodemographic characteristics with potential lead service lines. Integrating this environmental information with community racial/ethnic composition and other domains of social vulnerability are needed to prioritize activities within NYC and other large municipalities.

Our objectives were to a) to characterize the number and percentage of Potential Lead and Unknown service lines in NYC overall and across political jurisdictions, and b) to evaluate the association between racial/ethnic composition and reported water service line material (specifically Potential Lead and Unknown service lines, as we are specifically interested in lead line replacement) across NYC at the census tract level using geospatial models. In secondary analyses, we also evaluated the association between two constructed measures of social vulnerability (housing vulnerability and child lead exposure vulnerability) and service line material. In this paper, we evaluate neighborhood racial/ethnic composition to inform equitable interventions and consider that differences in service line material by racial/ethnic composition reflects structural racism. Structural racism “encompasses racist practices and systems that are imbedded into institutions in the US,” and is the “totality of ways in which societies foster racial discrimination through mutually reinforcing systems.^19,20” The leading framework explaining how racial/ethnic and other sociodemographic disparities in US drinking water exposures are created and reinforced theorizes that inequities in the natural (e.g. hydrogeology, climate), built (e.g. infrastructure, technical and financial capacity, economies of scale), and sociopolitical (e.g. social and political capital, housing discrimination) environments underlie these disparities.²¹ For drinking water lead exposure, inequities in water infrastructure, housing discrimination, and inadequacies in federal drinking water regulations are particularly relevant.

METHODS

RESULTS

Out of 854,672 service lines, 489,535 (57.3%) were reported as Not Lead, 136,891 (16.0%) as Potential Lead, 227,443 (26.6%) as Unknown, and 803 (<1%) as Not Applicable (Supplemental Table 5). The largest absolute number of Potential Lead and Unknown service lines were in Queens (N= 66,508) and Brooklyn (N= 91,377), respectively (Supplemental Table 6). Relative to other boroughs, a larger proportion of service lines were Potential Lead in the Bronx (21.2%) and Queens (20.5%), and Brooklyn had the largest proportion of service lines that were Unknown (33.2%). Buildings with multiple dwelling units had a similar proportion of Potential Lead and Unknown services lines compared to other buildings (17.7% versus 15.6%, and 25.4% versus 26.9%, respectively) (Supplemental Table 7), while buildings owned by NYC had a higher proportion of Unknown service lines compared to other buildings (62.2% versus 26.0%) (Supplemental Table 8). Tables of the absolute number and proportion of service line types by zip code, community district, state senate district, and state assembly district overall and further stratified by buildings with multiple dwelling units or those which are owned by NYC are in Supplemental Tables 9-22.

Figure 1 displays census tract proportions of Unknown and Potential Lead service lines. Census tracts with the highest proportion of Unknown service lines were located in Staten Island and southern Queens, while the highest proportion of Potential Lead service lines were located in eastern Queens and eastern Bronx (these areas generally have higher proportions of non-Hispanic Asian and Hispanic/Latino/non-Hispanic Black residents, respectively, see Supplemental Figure 3). The mean number of service line records within census tracts was 409 (minimum N=1, maximum N= 3,252), and the proportion of Potential Lead service lines ranged from 0% (N= 140 tracts) to 70% (N= 1 tract) (Table 1). Census tracts in the highest quartile of the proportion of Potential Lead service lines (range >23% - 70%) had lower mean total number and proportions of non-Hispanic Black and non-Hispanic White residents, higher mean total number and proportions of Hispanic/Latino residents, and higher scores of the CDC’s socioeconomic vulnerability index (Table 1).

• Download figure
• Open in new tab

Figure 1. Percentage of building water service lines recorded as Potential Lead (Panel A) and Unknown (Panel B) by census tract across New York City (N= 2,083).

Census tracts shown in gray were not included in the analysis. For Potential Lead service lines, local indicators of spatial association (LISA) clusters categorized as “High/High” were identified in northern and eastern Queens and eastern Bronx (Supplemental Figure 1).

Table 2 presents the RR (95% CrIs) of Potential Lead and Unknown service lines per 20% higher proportion of residents belonging to a given racial/ethnic group at the census tract level (N= 2,083). In fully adjusted models (Model 2), the RR (95% CrI) of Potential Lead service lines was 0.92 (0.89, 0.96) for non-Hispanic White residents, 1.15 (1.11, 1.21) for Hispanic/Latino residents, and null for non-Hispanic Black and non-Hispanic Asian residents. For Unknown service lines, the RR (95% CrI) was 1.03 (1.01, 1.06) for non-Hispanic Black residents, 1.04 (1.01, 1.07) for non-Hispanic White residents, 0.94 (0.91, 0.97) for non-Hispanic Asian residents, and 0.95 (0.92, 0.98) for Hispanic/Latino residents. Flexible cubic spline models indicated that most associations were non-linear, especially for Unknown lines and Potential Lead lines for Hispanic/Latino and non-Hispanic Asian residents (Figure 2). In “leave-one-out” models, associations for non-Hispanic Black and non-Hispanic White residents and Unknown lines were driven by lower proportions of Hispanic/Latino residents (GMR 1.07, 95% CI 1.03, 1.11, and GMR 1.08, 95% CI 1.04, 1.12, respectively); associations for Hispanic/Latino and non-Hispanic Asian residents were driven by lower proportions of non-Hispanic White residents (GMR 1.16, 95% CI 1.10, 1.25 and GMR 1.08, 95% CI 1.02, 1.14, respectively).

• Download figure
• Open in new tab

Figure 2. Flexible natural cubic spline models for relative risks (RR) and 95% credible intervals (CrIs) of service line type (Potential Lead in purple and Unknown in green) associated with a 20% higher proportion of residents in a given racial/ethnic subgroup within New York City at the census tract level (N=2,083).

Lines represent the RR of Potential Lead and Unknown service lines by proportion of residents in a given racial/ethnic group (effect estimate values are given on the left y-axis). Green and blue shaded areas (outlined by dotted lines) represent 95% CrIs. Shaded gray bars represent the distribution of the proportion of residents in a given racial/ethnic group at the census tract level across NYC (values are given on the x-axis and the percent of census tracts with a given proportion of residents of each racial/ethnic group is given on the right y-axis). Models are globally smooth conditional autoregressive (CAR) Poisson models via the S.CARbym function in the “CARBayes” R package, and include knots at the 33^rd and 66^th percentiles, with the reference set to the 10^th percentile of the distribution. Model offset is set as the log census tract total number of buildings with service line material record available. The total number of burn-in iterations was set to 20,000, 5,000 iterations were retained, and a thinning factor of 5 was applied. Spatial neighbors were identified via a queen contiguity matrix for neighbors (i=1) versus non-neighbors (i=0). Models are adjusted for population density, median building construction year, and the constructed child lead exposure vulnerability index.

Per increase in vulnerability score corresponding to the interquartile range, higher child lead exposure vulnerability score was associated with both Potential Lead and Unknown service lines, while higher housing vulnerability score was associated with Unknown service lines only (Table 2; similar findings across quartiles, see Supplemental Table 4). In analyses stratified by borough, Potential Lead service lines were associated with higher proportions of non-Hispanic White and non-Hispanic Asian residents in the Bronx and Manhattan, and with higher proportions of non-Hispanic Black residents in Queens (Supplemental Table 23). For models evaluating racial/ethnic composition and vulnerability indices, further adjustment for median building construction year attenuated effect estimates for Potential Lead but not Unknown lines, reflecting that Unknown lines are more evenly distributed throughout the city (Figure 1).

DISCUSSION

We found a high number and proportion of Potential Lead (16%) and Unknown (26.6%) service lines in residential buildings throughout NYC and significant inequities in Potential Lead service line prevalence, underscoring the high overall presence of lead service lines within NYC and the importance of prioritizing lead service line replacement for the most impacted communities. Higher proportions of Hispanic/Latino residents and higher child lead exposure vulnerability index scores were significantly associated with Potential Lead service lines. These findings suggest that NYC communities with children who are already highly vulnerable to lead exposures from numerous sources (paint, dust) have, in addition, a higher proportion of lead service lines in their communities. The positive association between child lead exposure vulnerability and Potential Lead service lines is likely related to variables that were not included in the housing vulnerability index (the number of residents <6 years old, median household income, the proportion of residential units that were vacant, and the proportion of adults without health insurance), or other unmeasured and correlated variables. While positive associations for Hispanic/Latino residents and non-Hispanic Asian residents with Potential Lead service lines was driven by lower proportions of non-Hispanic White residents, associations between non-Hispanic Black and non-Hispanic White residents and Unknown service lines were driven by lower proportions of Hispanic/Latino residents. Prioritization for service line replacement should also consider neighborhoods with higher proportions of Hispanic/Latino residents and neighborhoods with a higher proportion and absolute number of Potential Lead lines, including but not limited to eastern Queens and eastern Bronx, and NYC overall. An undetermined proportion of Unknown services lines are lead; further evaluation of these service lines is warranted throughout NYC.

These findings are supported by a growing body of literature characterizing inequities in public water system contaminant exposures across racial/ethnic and socioeconomic subgroups in the US, including lead.^4,30,31 Inequities in public drinking water contaminant exposures result from complex, multilevel interactions between the built (water infrastructure, land use patterns), sociopolitical (social vulnerability, regulatory policies and enforcement), and natural (geologic, hydrologic conditions) environments.²¹ Racism and classism operating through the built (priorities for repairing public water infrastructure, housing age and plumbing components) and sociopolitical environments (housing discrimination, lead corrosion control policies, federal LCR regulatory policies and enforcement) are likely avenues creating and reinforcing inequities in water lead exposure. Future analyses should consider a more intersectional approach by evaluating socioeconomic vulnerability, racial/ethnic composition, and other interacting dimensions of marginalization at the neighborhood-level. Future work should also evaluate how changes in neighborhood demographic composition has changed over time in relation to the replacement of lead service lines, including assessing residential segregation.

Our findings complement known geographic, socioeconomic, and racial/ethnic inequalities in NYC child blood lead levels (which integrate all lead exposure sources including drinking water, dust, and paint) published by the NYC Department of Health and Mental Hygiene.^32-34 For example, in 2017 the highest percentage of children with blood lead levels ≥5 μg/dL lived in Brooklyn (44%) and Queens (28%).³² To our knowledge, the NYC Department of Health and Mental Hygiene only makes aggregate child blood lead data publicly available at the United Healthcare Fund neighborhood level (N= 42 neighborhoods across the entire city).³⁵ Deidentified, individual-level records linking child blood lead levels to housing characteristics and service line material would enable valuable analyses quantifying the absolute and relative contribution of lead sources (e.g. drinking water, diet, dust, and paint) across racial/ethnic groups in NYC.^36,37

There are several limitations to our analyses. We relied on census tracts as our geographic unit of aggregation because the sociodemographic variables of interest for this analysis were mostly available only at the census tract and county level. Although reliance on administrative boundaries in geospatial studies can bias findings, our approach is directed at identifying community characteristics based on administrative boundaries specifically because these boundaries are likely the most useful for public health infrastructure investments and the prioritization of service line replacement, even though residents may self-define neighborhoods differently. We were unable to repeat our analysis at alternative geographic resolutions (e.g. census block, zip code) and therefore findings may differ across analyses repeated at alternative administrative units. We also relied on Census bridged race/ethnicity categories, which do not capture several important domains of race/ethnicity and limited our analysis to four crude groupings.

Our analysis is likely subject to non-trivial outcome misclassification because the NYCDEP Lead Service Line Location Coordinates dataset has not been validated and is based largely on historical records updated semi-annually. Outcome misclassification may be differential by neighborhood sociodemographic characteristics (e.g. single-family home owners in high-income neighborhoods may be more likely to contract with licensed plumbers, resulting in more accurate records). While non-differential exposure misclassification can bias findings in either direction, we expect that lead service lines were more likely replaced in buildings and neighborhoods with a higher proportion of non-Hispanic White residents, and that these records are more likely to be correctly updated in NYCDEP’s Lead Service Line Location Coordinates database. We suspect that observed associations for Potential Lead service lines might be underestimated for Hispanic/Latino, non-Hispanic Black, and non-Hispanic Asian groups because a higher proportion of Unknown lines may be lead. Future studies should assess the validity of reported service line material in a targeted, representative sample of NYC buildings.

There is no safe level of lead exposure for children.^5,38,39 Even at lower levels of exposure, lead is associated with impaired cognitive function, attention-related behavioral problems, and diminished academic performance, with many studies indicating that the steepest declines in neurocognitive outcome measures occur at the lowest levels of blood lead.^6,40,41 Numerous studies have demonstrated racial/ethnic and socioeconomic disparities in public drinking water exposures across the US, and have called for justice-oriented interventions and policies that protect the most highly exposed communities. While advancing environmental justice necessitates that all lead service lines should be identified and replaced within NYC and NYS, equitable efforts should prioritize the most impacted communities. Improvements to public health demands attention to all potential lead sources, including water, soil, and paint.