Skip to main content
Open access
Research
9 December 2020

Inequalities in Public Water Arsenic Concentrations in Counties and Community Water Systems across the United States, 2006–2011

Publication: Environmental Health Perspectives
Volume 128, Issue 12
CID: 127001

Abstract

Background:

In the United States, nationwide estimates of public drinking water arsenic exposure are not readily available. We used the U.S. Environmental Protection Agency’s (EPA) Six-Year Review contaminant occurrence data set to estimate public water arsenic exposure. We compared community water system (CWS) arsenic concentrations during 2006–2008 vs. after 2009–2011, the initial monitoring period for compliance with the U.S. EPA’s 10μg/L arsenic maximum contaminant level (MCL).

Objective:

Our objective was to characterize potential inequalities in CWS arsenic exposure over time and across sociodemographic subgroups.

Methods:

We estimated 3-y average arsenic concentrations for 36,406 CWSs (98%) and 2,740 counties (87%) and compared differences in means and quantiles of water arsenic (via quantile regression) between both 3-y periods for U.S. regions and sociodemographic subgroups. We assigned CWSs and counties MCL compliance categories (High if above the MCL; Low if below) for each 3-y period.

Results:

From 2006–2008 to 2009–2011, mean and 95th percentile CWS arsenic (in micrograms per liter) declined by 10.3% (95% CI: 6.5%, 14.1%) and 11.5% (8.3%, 14.8%) nationwide, by 11.4% (4.7%, 18.1%) and 16.3% (8.1%, 24.5%) for the Southwest, and by 36.8% (7.4%, 66.1%) and 26.5% (12.1%, 40.8%) for New England, respectively. CWSs in the High/High compliance category (not MCL compliant) were more likely in the Southwest (61.1%), served by groundwater (94.7%), serving smaller populations (mean 1,102 persons), and serving Hispanic communities (38.3%).

Discussion:

Larger absolute declines in CWS arsenic concentrations at higher water arsenic quantiles indicate declines are related to MCL implementation. CWSs reliant on groundwater, serving smaller populations, located in the Southwest, and serving Hispanic communities were more likely to continue exceeding the arsenic MCL, raising environmental justice concerns. These estimates of public drinking water arsenic exposure can enable further surveillance and epidemiologic research, including assessing whether differential declines in water arsenic exposure resulted in differential declines in arsenic-associated disease. https://doi.org/10.1289/EHP7313

Introduction

Racial/ethnic, socioeconomic, and geographic inequalities in U.S. drinking water contaminant exposures reflect structural inequality ingrained in land-use patterns, regulatory policy, facility siting, underlying geologic processes, and municipal zoning decisions (Gibson et al. 2020; Morello-Frosch et al. 2002; Nigra 2020; Wilson et al. 2008). Yet, systematic studies of inequalities in public drinking water contaminant exposures are lacking. Previous studies have identified inequalities in public drinking water exposure estimates for nitrate (higher among Hispanic populations) and arsenic (higher among incarcerated populations in the Southwest) (Nigra and Navas-Acien 2020; Schaider et al. 2019). These studies highlight the critical need to systematically identify subgroups of the U.S. population with elevated public drinking water contaminant exposures in order to identify environmental justice concerns and to inform public health interventions and regulatory action needed to eliminate exposure inequalities.
Inorganic arsenic is a human carcinogen (IARC 2004), highly toxic metalloid (Kuo et al. 2015, 2017; Milton et al. 2017; Moody et al. 2018; Moon et al. 2017; Sanchez et al. 2018), and water contaminant present in many aquifers in the United States. In 2001, the U.S. Environmental Protection Agency (EPA) Final Arsenic Rule lowered the arsenic maximum contaminant level (MCL) in public water systems from 50 to 10μg/L (U.S. EPA 2001a, 2001b). The Final Arsenic Rule went into effect on 23 January 2006 and included an initial monitoring period for compliance with the new MCL (2006–2008). Reducing the MCL from 50 to 10μg/L has prevented an estimated 200–900 cancer cases per year (Nigra et al. 2017). Evaluating the extent to which MCL implementation reduced public water arsenic exposure, and identifying population subgroups disproportionately exposed by geography, race/ethnicity, and socioeconomic status (SES), remains critical.
Some evidence indicates that reductions in public water arsenic exposure were not uniform across all communities and that some public water systems remain in violation of the 10μg/L MCL, especially in the southwestern United States (Environmental Integrity Project 2016; Foster et al. 2019; Nigra et al. 2017; Welch et al. 2018). Systems serving smaller populations were more likely to be in violation of the new MCL than larger systems, likely because of the increased relative cost in treating, mixing, or switching source water per population-served size (Foster et al. 2019). Although racial/ethnic inequalities in urinary arsenic concentrations are well documented (with higher internal dose estimates for Hispanic and Asian-American subgroups), evaluating the relative contribution of diet vs. drinking water to total inorganic arsenic exposure remains challenging because no national-level estimates of public drinking water arsenic exposure are available (Awata et al. 2017a, 2017b; Cubadda et al. 2017; Jones et al. 2019; Kurzius-Spencer et al. 2014). Estimating public drinking water arsenic exposure for sociodemographic and geographic subgroups is needed to evaluate whether inequalities in arsenic exposure and compliance with the MCL persist across the United States, to inform future national- and state-level arsenic MCL deliberations, and to investigate whether inequalities in drinking water arsenic exposure by subgroup contributes to disparities in arsenic-related disease.
Our objective was to estimate public drinking water arsenic exposure in community water systems (CWSs) across the United States and to identify subgroups whose public water arsenic concentrations have remained above 10μg/L after the new arsenic MCL implementation and are, therefore, at disproportionate risk of arsenic-related adverse health outcomes. We estimated arsenic exposure at the CWS level (primary analysis) and at the county level (secondary analysis), comparing arsenic concentrations during 2006–2008 vs. after 2009–2011, the initial monitoring period for compliance with the MCL, using the national contaminant occurrence database supporting the U.S. EPA’s Third Six-Year Review of drinking water regulations. This database contains monitoring records for arsenic in public drinking water systems across the United States from 2006 to 2011. We considered the following subgroups in our analysis: a) U.S. region (CWS and county level); b) sociodemographic county cluster (CWS and county level); and c) population-served size and source water type (CWS level).

Materials and Methods

U.S. EPA Public Water Databases

The U.S. EPA compiles compliance monitoring data from public water supplies for regulated drinking water contaminants every 6 y as required by the Safe Drinking Water Act (SDWA) (U.S. EPA 2016a, 2016b). Data are obtained through the Information Collection Request process (voluntarily sent in from states, territories, and tribal authorities). The U.S. EPA works directly with agencies to collect records for a given monitoring period. We used arsenic monitoring data from the Third Six-Year Review (SYR3) period (2006–2011), which included approximately 13 million analytical records from 139,000 public water systems serving 290 million people annually. Data from 46 states, Washington DC, the Navajo Nation, and American Indian tribes from U.S. EPA Regions 1, 4, 5, 8, and 9, representing 95% of all public water systems and 92% of the total population served by public water systems nationally, were included (U.S. EPA 2016a, 2016b). Agencies with primacy for implementing the SDWA at public water systems in the states of Colorado, Delaware, Georgia, and Mississippi and U.S. EPA Regions 2, 6, 7, and 10 for water systems serving American Indian tribes did not submit data for SYR3 (U.S. EPA 2016a). The U.S. EPA conducted extensive quality assurance and quality control assessments prior to publishing the final SYR3 data (U.S. EPA 2016a, 2016b, 2016c). From an initial 297,354 arsenic monitoring records from 54,845 public water systems, we excluded 67,089 records from 17,747 systems categorized as transient or nontransient noncommunity water systems. Of the remaining 230,265 records, 104,730 reported arsenic detections and 125,535 reported nondetections.
We replaced arsenic values below the limit of detection (LOD) with the LOD divided by the square root of 2. Although the U.S. EPA established the maximum LOD of 0.5μg/L, many systems reported lower and higher LODs or did not report record-specific LODs. When records reported the laboratory LOD as <5μg/L, we replaced arsenic values below the LOD with the LOD divided by the square root of 2 (N=30,256 records); for 1,202 records from Wisconsin and West Virginia that likely reported LODs in incorrect units (milligrams per liter instead of micrograms per liter), we imputed the corrected LOD divided by the square root of 2 (West Virginia Department of Health and Human Services 2019; Wisconsin Department of Natural Resources 2019). We imputed the value of 0.35μg/L (the U.S. EPA LOD of 0.5μg/L divided by the square root of 2) for a) 88,640 records with arsenic levels reported below the LOD without record-specific LOD; and b) the remaining 5,437 records for which LODs were reported as 5μg/L because these were considered unreliable.
To assign counties-served for each CWS, we merged the SYR3 arsenic monitoring data with system inventory information extracted from the U.S. EPA Safe Drinking Water Information System (SDWIS), including counties served, number of people served, type of system, and source water type (U.S. EPA 2017). Because only county served was reliably reported in SDWIS for each CWS, we could not aggregate to smaller geographic scales (e.g., census-tract levels). Of the 37,098 CWSs, we successfully merged 36,372 to SDWIS. Of the remaining CWSs, we manually assigned county-served to 452 of these systems based on ZIP code (439 via U.S. Department of Housing and Urban Development Crosswalk files, and 13 via GoogleMaps) (Figure S1 presents a flowchart of the data cleaning, data merging, and inclusion/exclusion criteria). The final sample size included 36,406 CWSs (98% of CWSs in the SYR3) serving a total of 2,740 counties (87% of 3,141 U.S. counties and county-equivalents) across 47 states and serving 254,610,301 people.

Water Arsenic Exposure Estimates

We averaged available water arsenic monitoring records to 3-y periods (2006–2008 and 2009–2011) because these sampling periods differentiated between during (2006–2008) and after (2009–2011) the initial compliance monitoring period. We also averaged to 3-y periods because the number of records reported by CWSs was differential by prior MCL exceedances and source water type and varied substantially across CWSs (range: 1–1,506; mean: 4) (U.S. EPA 2004). Averaging available records to 3-y periods also minimized missing data and facilitated the comparison of arsenic exposure estimates across CWSs (for diagnostics regarding this analytical decision, see Tables S1 and S2). Few CWSs reported records of both raw and finished (i.e., treated) water samples within the same year (N=1,182). When the 3-y average of arsenic in finished water samples was lower than in raw concentrations (N=336 CWSs), we calculated the 3-y average with only the finished water samples.
At the county level (secondary analyses), we estimated weighted average 3-y (2006–2008 and 2009–2011) water arsenic concentrations, accounting for the number of people served by each CWS (population served). For each 3-y period, the population-served weight for each CWS was calculated as the population served by that CWS divided by the total population served by all CWSs serving that county. CWSs not reporting data for a particular 3-y period did not contribute to that county’s total population served or 3-y average (Equation S1). To avoid reporting a county-level average derived from CWSs that served only small populations relative to the entire county population, the average water arsenic concentration for a given county and 3-y period was estimated only if the CWSs serving that county reported serving at least 50% or more of the public water–reliant population in the entire county. We estimated the public water–reliant population for each county using the latest nationwide U.S. Census statistic on county-level household tap water source from the 1990 U.S. Census (Ruggles et al. 2019), which was also recently used by the U.S. Geologic Survey (USGS) (Ayotte et al. 2017). For descriptive purposes, we also calculated county-level 6-y average water arsenic concentrations as the average of the two 3-y period estimations. We mapped 3- and 6-y county-level estimates of water arsenic averages across the conterminous United States using the maps package in R version 3.5.3 (Becker and Wilks 2018).

Statistical Analysis: Water Arsenic Exposure Estimates

Primary analyses were conducted at the CWS level and secondary analyses at the county level. We calculated the distribution—including percentiles, arithmetic means, and geometric means—of 6- and 3-y average water arsenic concentrations at the CWS (N=36,406) and county level (N=2,740) for the entire United States. We also compared mean differences between the two time periods. Because the reductions in average water arsenic concentrations over time occurred at the highest end of the distribution, we also performed quantile regression at the 75th, 80th, 85th, 95th, and 99th percentiles to estimate the difference in water arsenic concentration over time at these quantiles. For CWS-level models, we evaluated both crude models and models adjusting for source water type and the size of the population served to determine whether differences in water arsenic concentrations over time were related to these variables.

Statistical Analysis: Analyses Stratified by Subgroup

To identify subgroups of the U.S. population whose estimated public drinking water arsenic exposures were relatively high, we stratified our analyses by the following population subgroups: region (CWS- and county-level analyses), sociodemographic county-cluster (CWS- and county-level analyses), source water type (CWS-level analyses), and size of the population served by CWSs (CWS-level analyses). Region groupings were based on USGS-identified areas with similar groundwater arsenic patterns: Pacific Northwest (Washington, Oregon, Montana, Wyoming, and Idaho), Southwest (California, Nevada, Utah, Colorado, Arizona, New Mexico, and Texas), Central Midwest (North Dakota, South Dakota, Nebraska, Kansas, and Missouri), Eastern Midwest (Wisconsin, Illinois, Indiana, Michigan, Ohio, Minnesota, and Iowa), Southeast (Oklahoma, Arizona, Louisiana, Mississippi, Alabama, Florida, Georgia, Tennessee, Kentucky, South Carolina, North Carolina, Virginia, and West Virginia), Mid-Atlantic (Pennsylvania, Maryland, District of Columbia, Delaware, New York, New Jersey, Connecticut, and Rhode Island), New England (Massachusetts, Vermont, New Hampshire, and Maine), and Alaska/Hawaii (Alaska and Hawaii) (Ayotte et al. 2017). Sociodemographic county-clusters (N=8 distinct clusters) were derived by Wallace et al. (2019) to enable the direct comparison of county-level health outcomes and behaviors while accounting for the sociodemographic makeup of a county’s population. Briefly, Wallace et al. (2019) used k-means analysis to identify groups of counties with similar sociodemographic profiles (e.g., race and ethnicity, age, educational attainment, employer status, health insurance status). In the present study, we stratified public water arsenic exposure estimates by these sociodemographic county-clusters to identify characteristics of population subgroups exposed to elevated public water arsenic exposure. These sociodemographic county-clusters are as follows: Semi-Urban, High SES; Semi-Urban, Mid/Low SES; Semi-Urban, Hispanic; Mostly Rural, Mid SES; Rural, Mid/Low SES; Young, Urban, Mid/High SES; Rural, American Indian; and Rural, High SES. We also stratified CWS-level analyses by source water type (groundwater vs. surface water, as reported in SDWIS) and by the size of the population served (500, 501–3,300, 3,301–10,000, and >10,000 persons) (these are not relevant at the county level).
We conducted sensitivity analyses for the comparison of water arsenic concentrations over the two 3-y periods. First, we replaced all arsenic concentration values below the LOD with 0.35μg/L (the standard U.S. EPA LOD divided by the square root of 2), regardless of the LOD reported in each record. Second, we excluded records with a reported LOD of 5μg/L. Third, we estimated county average water arsenic concentrations for a given 3-y period only if the CWSs serving that county reported serving at least 70% and 80% or more of the public water–reliant population in the entire county (instead of 50% or more as in our main analysis).

MCL Compliance Evaluation

We assigned all CWSs and counties to one of four MCL compliance categories using the 10μg/L MCL cut point based on the average water arsenic concentration in the first (2006–2008) and second (2009–2011) 3-y time periods: Low/Low (<10μg/L in both periods); High/Low (>10μg/L in 2006–2008, but <10μg/L in 2009–2011); Low/High (<10μg/L in 2006–2008 but >10μg/L in 2009–2011); and High/High (>10μg/L in both periods). We also categorized compliance categories using cutoff values of 5μg/L (the current MCL for New Hampshire and New Jersey) and 1μg/L (the MCL for the Netherlands and close to the U.S. EPA’s MCL goal of 0μg/L). We compared characteristics of CWSs and counties (region, sociodemographic county-cluster, source water type, and population served) by compliance category, and we mapped these county-level compliance categories for the conterminous United States. Finally, we identified all counties and CWSs that served counties classified into the High/Low (i.e., successful in reducing water arsenic below the new MCL) and High/High (i.e., not successful in reducing water arsenic below the new MCL) compliance categories to identify counties that were successful vs. unsuccessful in reducing water arsenic concentrations. We also assessed the odds ratio (OR) of average water arsenic concentration exceeding the 10μg/L MCL at either time period (2006–2008 and 2009–2011) for CWSs in a given subgroup, compared with all CWSs not in that subgroup (reference), using logistic regression models with generalized estimating equations to account for two 3-y averages per CWS.

Interactive Map and Publicly Available Data

We also created an online interactive map of county-level water arsenic concentrations for all three time periods (2006–2008, 2009–2011, and 2006–2011) and MCL compliance categories to improve accessibility and results dissemination (https://annenigra.github.io/ColumbiaArsenicMap.html). Data sets of the 3-y and 6-y average arsenic concentrations at the CWS and county level and a reproducible R archive of analyses and data sets are available in the Supplemental Material and via GitHub (https://github.com/annenigra/epa-public-water-arsenic).

Results

Average CWS Arsenic Estimates Nationwide and by Region

Nationwide, the mean CWS arsenic concentrations [95% confidence intervals (CIs)] in 2006–2008 and 2009–2011 were 1.89μg/L (95% CI: 1.84, 1.94) and 1.7μg/L (95% CI: 1.64, 1.75) [mean difference=0.19 (95% CI: 0.27, 0.12); corresponding percentage decrease=10.1% (95% CI: 14.1%, 6.5%) (Table 1)]. By region, CWS arsenic concentrations for the first time period (2006–2008) were highest in the Southwest [mean=3.59μg/L (95% CI: 3.41, 3.76)], followed by Alaska/Hawaii [mean=2.17μg/L (95% CI: 1.73, 2.61)], the Pacific Northwest [mean=2.15μg/L (95% CI: 2.03, 2.27)], and the Eastern Midwest [mean=2.03μg/L (95% CI: 1.92, 2.14)] (Table 1; Figure 1). CWSs in the New England region experienced the greatest absolute mean decline in average water arsenic between 2006–2008 and 2009–2011 [mean difference=0.75μg/L (95% CI: 1.35, 0.15); corresponding percentage decrease=36.8% (95% CI: 7.4%, 66.1%)], followed by the Southwest [meandifference=0.41μg/L (95% CI: 0.65, 0.17); corresponding percentage decrease=11.4% (95% CI: 4.7%, 18.1%)]; and Eastern Midwest [mean difference=0.40μg/L (95% CI: 0.54, 0.25); corresponding percentage decrease=19.7% (95% CI: 12.2%, 26.8%)]. Detailed distributions of water arsenic concentrations in 2006–2008 and 2009–2011 stratified by region can be found in Excel Table S1. Distributions of average water arsenic from 2006–2008 were markedly bimodal in the Southwest and New England regions (Figure 1). Mean difference estimates did not change with adjustment for source water type and size of population served (Table S3).
Table 1 Arithmetic means (95% CIs), mean differences (95% CIs), and corresponding percentage differences (95% CIs) of 3-y average water arsenic concentrations (μg/L) in community water systems (CWSs) from 2006–2008 and 2009–2011, stratified by CWS subgroup (total N=36,406 CWSs).
CWS Categories2006–20082009–2011Mean difference (95% CI)Corresponding percentage difference (95% CI)
CWSs (N)Records (thousands) (N)Mean (95% CI)CWSs (N)Records (thousands) (N)Mean (95% CI)
All CWSs30,8202101.89 (1.84, 1.94)32,4812201.7 (1.64, 1.75)0.19 (0.27, 0.12)10.3 (14.1, 6.5)
Source water typea 150  160   
 Groundwater26,279602.02 (1.96, 2.08)27,572611.80 (1.75, 1.86)0.22 (0.30, 0.14)10.8 (14.9, 6.7)
 Surface water4,537 1.12 (1.04, 1.19)4,906 1.08 (1.01, 1.15)0.04 (0.14, 0.07)3.2 (12.5, 6.1)
Population size servedb
50018,001702.09 (2.01, 2.17)18,101711.89 (1.81, 1.97)0.20 (0.32, 0.09)9.6 (15.1, 4.2)
 501–3,3007,190431.78 (1.70, 1.86)8,048441.59 (1.52, 1.67)0.19 (0.30, 0.08)10.5 (16.8, 4.3)
 3,301–10,0002,803271.53 (1.43, 1.62)3,236271.36 (1.27, 1.45)0.17 (0.30, 0.03)11.1 (19.9, 2.2)
 10,001–100,0002,485501.27 (1.19, 1.36)2,730511.19 (1.10, 1.27)0.09 (0.21, 0.04)6.7 (16.5, 3)
>100,000341231.21 (1.05, 1.37)366231.22 (1.07, 1.38)0.01 (0.21, 0.23)0.9 (17.3, 19.1)
Region
 Alaska/Hawaii4141.82.17 (1.73, 2.61)4261.91.65 (1.22, 2.09)0.52 (1.14, 0.10)23.8 (52.3, 4.6)
 Central Midwest2,2389.61.94 (1.80, 2.08)2,520101.98 (1.85, 2.11)0.04 (0.15, 0.23)2.2 (7.5, 12)
 Eastern Midwest4,878262.03 (1.92, 2.14)5,476271.63 (1.53, 1.74)0.40 (0.54, 0.25)19.5 (26.8, 12.2)
 Mid-Atlantic4,695250.88 (0.82, 0.94)4,611250.81 (0.75, 0.87)0.07 (0.15, 0.02)7.5 (17.3, 2.3)
 New England1,587112.05 (1.62, 2.47)1,592111.29 (0.87, 1.72)0.75 (1.35, 0.15)36.8 (66.1, 7.4)
 Pacific Northwest3,986202.15 (2.03, 2.27)3,584192.11 (1.99, 2.24)0.04 (0.21, 0.13)1.8 (9.8, 6.2)
 Southeast6,686300.66 (0.63, 0.70)7,243320.65 (0.62, 0.68)0.01 (0.06, 0.03)2.1 (8.6, 4.4)
 Southwest6,336893.59 (3.41, 3.76)7,029903.18 (3.01, 3.34)0.41 (0.65, 0.17)11.4 (18.1, 4.7)
Sociodemographic county-clusterc
 Semi-Urban, High SES14,479781.54 (1.47, 1.61)14,479791.39 (1.32, 1.46)0.15 (0.25, 0.06)9.9 (16.2, 3.7)
 Semi-Urban, Mid/Low SES1,03460.63 (0.53, 0.72)1,3286.80.65 (0.57, 0.74)0.03 (0.10, 0.15)4.3 (16, 24.7)
 Semi-Urban, Hispanic3,537433.60 (3.38, 3.82)3,849443.4 (3.19, 3.62)0.19 (0.50, 0.11)5.3 (13.8, 3.2)
 Mostly Rural, Mid SES7,717381.44 (1.37, 1.52)8,120391.26 (1.18, 1.34)0.19 (0.30, 0.07)12.9 (20.5, 5.2)
 Rural, Mid/Low SES3851.71.22 (0.92, 1.52)47921.11 (0.84, 1.38)0.11 (0.51, 0.29)9.1 (42.1, 23.8)
 Young, Urban, Mid/High SES901262.89 (2.59, 3.19)912262.71 (2.41, 3.01)0.18 (0.60, 0.25)6.1 (20.9, 8.6)
 Rural, American Indian40422.95 (2.46, 3.44)4122.12.50 (2.02, 2.99)0.45 (1.13, 0.24)15.1 (38.4, 8.1)
 Rural, High SES4,423222.32 (2.14, 2.50)4,565221.98 (1.80, 2.16)0.34 (0.59, 0.09)14.7 (25.6, 3.8)
Note: Records indicates the total number of individual monitoring records contributing to a given estimate. CI, confidence interval; EPA, Environmental Protection Agency; SES, socioeconomic status.
a
Groundwater is considered CWSs served by surface water under the influence of groundwater and groundwater under the influence of surface water.
b
Categories of population served are standard U.S. EPA categories. Population served is adjusted total population served, which accounts for systems that sell or purchase water and avoids overcounting.
c
A total of 172 CWSs served more than one county; of these, approximately half served counties categorized to different sociodemographic county-clusters (e.g., NY7003493 serves New York, New York (Young, Urban, Mid/High SES) and Bronx, New York (Semi-Urban, Hispanic). These CWSs are represented for each county that they serve in the sociodemographic county-cluster analyses (N=36,674).
Figure 1. Distribution of average arsenic concentrations (μg/L) in community water systems (CWSs) stratified by region for the period of 2006–2008. Filled polygons represent density plots. Box plot upper, middle, and lower hinges correspond to the 25th, 50th, and 75th percentiles, respectively. Box plots for the Southeast and Mid-Atlantic are difficult to visualize because of an extreme right skew. The 3-y average arsenic concentration for each CWS is represented by a dot. Mean region-specific concentrations are indicated by the outlined white circle; mean (95% confidence intervals) are also listed for each region in text. The 10μg/L maximum contaminant level is indicated by the red dashed line. The x-axis is truncated at 54.60μg/L. The R code for this figure was adapted from Allen et al. (2019).
Figure 2 illustrates the change in CWS water arsenic concentrations at a given quantile over the two time periods, stratified by region. Regions were ordered by the mean arsenic concentration from 2006–2008. Changes in CWS water arsenic concentrations nationwide from 2006–2008 to 2009–2011 were larger at increasing quantiles of the water arsenic distribution; significant changes occurred at the 80th quantile [decline of 0.13μg/L (95% CI: 0.23, 0.03)] through the 99th quantile [decline of 4.20μg/L (95% CI: 5.56, 2.84)] (Figure 2; Table S4). When stratified by region, quantile regression models indicated larger declines at the 99th percentile for New England [decline of 7.99μg/L (95% CI: 16.00, 0.02); corresponding percentage decrease=45.3%], the Eastern Midwest [decline of 8.44μg/L (95% CI: 11.83, 5.05); corresponding percentage decrease=40.5%], the Southwest [decline of 3.46μg/L (95% CI: 9.00, 2.08); corresponding percentage decrease=11.5%], and the Mid-Atlantic [decline of 2.43μg/L (95% CI: 4.46, 0.40); corresponding percentage decrease=23.1%] (Figure 2; Table S5).
Figure 2. Change in water arsenic concentrations (μg/L) at a given quantile from 2006–2008 vs. 2009–2011, stratified by region (quantile regression results). Regions are ordered by increasing mean arsenic concentrations in 2006–2008. The x-axis indicates the 80th, 85th, 90th, 95th, 96th, 97th, 98th, and 99th percentiles for all CWSs within a region. Shaded areas indicate the quantile regression confidence interval. The red dashed line indicates zero change in water arsenic concentration across the two time periods. Alaska/Hawaii was not included in this figure because of small sample size and unstable estimates (see Table S4). Note: CWS, community water system.
At the county level, nationwide weighted average water arsenic concentrations were mean=1.45μg/L (95% CI: 1.34, 1.57) in 2006–2008 and mean=1.43μg/L (95% CI: 1.32, 1.54) in 2009–2011 (Table S6). County-level average arsenic estimates for 2006–2008 and 2009–2011 displayed similar regional patterns to those at the CWS level, with a general pattern of markedly higher water arsenic concentrations in western vs. eastern counties (Figure 3; Excel Table S2). When restricting our analysis to counties with CWSs serving at least 70% or 80% of the public water–reliant population, overall effect estimates remained similar, but sample size was reduced (Table S7).
Figure 3. County-level estimated weighted average water arsenic concentrations in community water systems (CWSs) from (A) 2006–2011, (B) 2006–2008, and (C) 2009–2011 (N=36,406 CWSs serving N=2,740 counties). Average arsenic concentrations from 2006–2011 are the averages of both 3-y time periods (2006–2008 and 2009–2011). CWS arsenic concentrations were weighted by the population served by each system to estimate the county-level weighted average CWS arsenic concentrations. Counties that were not represented by any CWSs in the SYR3 database are labeled as “No data available.” Counties with “Inadequate data” did not have CWS data representing at least 50% of the public water–reliant population for either time period (2006–2008 and 2009–2011). SYR, Six-Year Review.

Average CWS Arsenic Estimates by Sociodemographic County-Cluster

CWSs in the Semi-Urban, Hispanic cluster had the highest mean of estimated water arsenic concentration [mean=3.60μg/L (95% CI: 3.38, 3.82)] in 2006–2008, followed by CWSs in the Rural, American Indian cluster [mean=2.95μg/L (95% CI: 2.46, 3.44)], the Rural, High SES cluster [mean=2.32μg/L (95% CI: 2.14, 2.50)], the Young, Urban, Mid/High SES cluster [mean=2.89μg/L (95% CI: 2.59, 3.19)], the Rural, Mid/Low SES cluster [mean=1.22μg/L (95% CI: 0.92, 1.52)], the Semi-Urban, High SES cluster [mean=1.54μg/L (95% CI: 1.47, 1.61)], the Mostly Rural, Mid SES cluster [mean=1.44μg/L (95% CI: 1.37, 1.52)], and the Semi-Urban, Mid/Low SES cluster [mean=0.63μg/L (95% CI: 0.53, 0.72)] (Table 1). The largest significant decline from 2006–2008 to 2009–2011 was observed for CWSs in the Rural, High SES cluster [decline of 0.34μg/L (95% CI: 0.59, 0.09); corresponding to a 10.5% percent decrease]. Detailed distributions of water arsenic concentrations in 2006–2008 and 2009–2011 stratified by sociodemographic cluster can be found in Excel Table S3. Results were similar for analysis at the county level (Excel Table S4).

Average CWS Arsenic Estimates by Source Water Type and Size of Population Served

CWSs served by groundwater had higher mean water arsenic levels compared with those served by surface water both in 2006–2008 (mean=2.02μg/L vs. 1.12μg/L) and 2009–2011 (mean=1.80μg/L vs. 1.08μg/L) (Table 1). The average decline in water arsenic from 2006–2008 to 2009–2011 was greater for CWSs served by groundwater [mean difference=0.22μg/L (95% CI: 0.30, 0.14); corresponding to a 10.8% decrease] compared with those served by surface water [mean difference=0.04μg/L (95% CI: 0.14, 0.07); corresponding to a 3.6% decrease]. In analyses stratified by the size of the population served by CWSs, water arsenic concentrations decreased for CWSs serving 500 persons [mean difference=0.20μg/L (95% CI: 0.32, 0.09); corresponding to a 9.6% decrease], CWSs serving 501–3,300 persons [mean difference=0.19μg/L (95% CI: 0.30, 0.08); corresponding to a 10.7% decrease], and CWSs serving 3,301–10,000 persons [mean difference=0.17μg/L (95% CI: 0.30, 0.03); corresponding to an 11.1% decrease]. Further adjustment for source water type did not change mean difference estimates in CWS arsenic concentrations over time by region, sociodemographic county-cluster, or size of population served (Table S3).

Spatial Patterns and ORs of MCL Exceedances

The percentage of CWSs with average concentrations of arsenic above the 10μg/L MCL was 3.2% in 2006–2008 vs. 2.3% in 2009–2011 (Table 2; Fischer exact test p<0.001). Across both time periods, the adjusted odds of MCL exceedance were significantly greater for CWSs in the Southwest [OR=3.98 (95% CI: 3.61, 4.40)], Alaska/Hawaii [OR=2.03 (95% CI: 1.47, 2.80)], and the Pacific Northwest [OR=1.20 (95% CI: 1.04, 1.37)], compared with all other CWSs. The odds of MCL exceedance across both time periods were also significantly greater for CWSs in the Semi-Urban, Hispanic county-cluster [OR=3.36 (95% CI: 3.02, 3.73)], the Young, Urban, Mid/High SES county-cluster [OR=2.02 (95% CI: 1.60, 2.54)], and the Rural, American Indian county-cluster [OR=2.64 (95% CI: 1.99, 3.51)], compared with all other CWSs.
Table 2 Number (%) of 3-y average arsenic concentrations (μg/L) equal to or exceeding the maximum contaminant level (MCL, 10μg/L), and ORs (95% CIs) of MCL exceedance (>10μg/L) in either time period (2006–2008 and 2009–2011; ORs compare CWSs in a given subgroup vs. all other CWSs). N=36,406 total CWSs.
CWS Categories2006–20082009–2011OR (95% CI) of MCL exceedance (>10μg/L)
CWSs (N)N(%)MCLCWSs (N)N(%)MCLCrudeAdjusted for source water type
All CWSs30,820995 (3.2)32,481738 (2.3)NANA
Source water typea
 Groundwater26,279936 (3.6)27,572699 (2.5)2.85 (2.31, 3.51)NA
 Surface water4,53759 (1.3)4,90639 (0.8)0.35 (0.29, 0.43)NA
Population size served (persons)b
50018,001682 (3.8)18,101522 (2.9)1.62 (1.46, 1.80)1.38 (1.24, 1.54)
 501–3,3007,190227 (3.2)8,048165 (2.1)0.95 (0.84, 1.06)0.95 (0.85, 1.07)
 3,301–10,0002,80353 (1.9)3,23635 (1.1)0.53 (0.43, 0.66)0.63 (0.50, 0.78)
 10,001–100,0002,48533 (1.3)2,73015 (0.5)0.35 (0.26, 0.46)0.48 (0.35, 0.65)
>100,0003410 (0)3661 (0.3)0.06 (0.01, 0.4)0.11 (0.01, 0.76)
Region
 Alaska/Hawaii41425 (6)42616 (3.8)1.87 (1.36, 2.58)2.03 (1.47, 2.80)
 Central Midwest2,23838 (1.7)2,52036 (1.4)0.55 (0.43, 0.70)0.53 (0.41, 0.67)
 Eastern Midwest4,878187 (3.8)5,47693 (1.7)0.93 (0.82, 1.07)0.88 (0.77, 1.01)
 Mid-Atlantic4,69551 (1.1)4,61124 (0.5)0.30 (0.23, 0.37)0.30 (0.24, 0.38)
 New England1,58750 (3.2)1,59211 (0.7)0.75 (0.58, 0.97)0.74 (0.57, 0.95)
 Pacific Northwest3,986149 (3.7)3,584117 (3.3)1.24 (1.08, 1.43)1.20 (1.04, 1.37)
 Southeast6,68623 (0.3)7,24323 (0.3)0.10 (0.07, 0.13)0.10 (0.08, 0.14)
 Southwest6,336472 (7.4)7,029418 (5.9)3.81 (3.46, 4.21)3.98 (3.61, 4.40)
Sociodemographic county-clusterc
 Semi-Urban, High SES14,479287 (2)14,479177 (1.2)0.54 (0.49, 0.61)0.53 (0.48, 0.60)
 Semi-Urban, Mid/Low SES1,0349 (0.9)1,32810 (0.8)0.29 (0.18, 0.45)0.28 (0.18, 0.45)
 Semi-Urban, Hispanic3,537266 (7.5)3,849270 (7)3.36 (3.03, 3.74)3.36 (3.02, 3.73)
 Mostly Rural, Mid SES7,717200 (2.6)8,120123 (1.5)0.67 (0.60, 0.76)0.70 (0.61, 0.79)
 Rural, Mid/Low SES3855 (1.3)4794 (0.8)0.36 (0.19, 0.70)0.36 (0.19, 0.69)
 Young, Urban, Mid/High SES90148 (5.3)91233 (3.6)1.74 (1.39, 2.20)2.02 (1.60, 2.54)
 Rural, American Indian40431 (7.7)41224 (5.8)2.43 (1.83, 3.23)2.64 (1.99, 3.51)
 Rural, High SES4,423155 (3.5)4,565104 (2.3)1.04 (0.91, 1.19)0.99 (0.87, 1.14)
Note: CI, confidence interval; CWS, community water system; EPA, Environmental Protection Agency; NA, not applicable; OR, odds ratio; SES, socioeconomic status.
a
Groundwater is considered CWSs served by surface water under the influence of groundwater and groundwater under the influence of surface water.
b
Categories of population served are standard U.S. EPA categories. Population served is adjusted total population served, which accounts for systems that sell or purchase water and avoids overcounting.
c
A total of 172 CWSs served more than one county; of these, approximately half served counties categorized to different sociodemographic county-clusters (e.g., NY7003493 serves New York, New York (Young, Urban, Mid/High SES) and Bronx, New York (Semi-Urban, Hispanic). These CWSs are represented for each county that they serve in the sociodemographic county-cluster analyses (N=36,674).

Characteristics of CWSs by MCL Compliance Categories

When evaluating compliance categories applying the 10μg/L cut point, the large majority of CWSs were categorized into the Low/Low category (N=25,846, 96.1%), followed by the High/High category (N=493, 1.8%), the High/Low category (N=397, 1.5%), and the Low/High category (N=159, <1%) (Table 3). The mean change from 2006–2008 to 2009–2011 (95% CI) in CWS arsenic concentration was 0.01μg/L (95% CI: 0.02, 0.01) in the Low/Low category, 12.84μg/L (95% CI: 15.36, 10.32) in the High/Low category, 6.81μg/L (95% CI: 5.17, 8.46) in the Low/High category, and 0.97μg/L (95% CI: 1.62, 0.33) in the High/High category. CWSs in the High/High and Low/High categories were more likely to be in the Southwest, to serve smaller populations, to serve Semi-Urban, Hispanic counties, and to be served by groundwater compared with those in Low/Low and High/Low categories. CWSs in the Low/Low category were more likely to rely on surface water, to serve a larger population size, to serve Semi-Urban, High SES counties, and to be located in the Southeast and Mid-Atlantic, and less likely to be located in the Southwest and Pacific Northwest compared with CWSs in other categories. CWSs in the High/Low category (those who successfully reduced water arsenic below the 10μg/L MCL) were more likely to rely on groundwater, to serve a smaller population size, to be tribal systems, to serve Semi-Urban Hispanic counties, and to be located in the New England, Eastern Midwest, Pacific Northwest, and Southwest regions. When applying the 5μg/L cut point (MCL for New Jersey and New Hampshire), 7.4% of CWSs (N=1,993) were in the High/High category. The majority of these systems were in the Southwest (N=874,43.9%), Pacific Northwest (N=317,15.9%), or Eastern Midwest (N=308,15.5%) regions (Excel Table S5). Patterns were similar for CWSs in the High/High category applying the 1μg/L cut point (MCL for the Netherlands).
Table 3 Characteristics of community water systems (CWSs) by compliance category applying 10μg/L cut point for categorization (N=26,895 CWSs with average arsenic estimates from both 3-y time periods available).
CWS CategoriesAll CWSsCategory 1 (Low/Low)Category 2 (High/Low)Category 3 (Low/High)Category 4 (High/High)
Mean arsenic change (95% CI) (μg/L)0.17 (0.13, 0.22)0.01 (0.01, 0.02)12.84 (10.32, 15.36)6.81 (8.46, 5.17)0.97 (0.33, 1.62)
N (%)26,89525,846 (96.1)397 (1.5)159 (<1)493 (1.8)
Source water type [N (%)]
 Groundwater22,733 (84.5)21,745 (84.1)370 (93.2)151 (95.0)467 (94.7)
 Surface water4,162 (15.5)4,101 (15.9)27 (6.8)8 (5.0)26 (5.3)
Population served [mean (SE)]8,064 (510)8,326 (530)2,417 (407)1,152 (227)1,102 (148)
Region [N (%)]
 Alaska/Hawaii358 (1.3)336 (1.3)10 (2.5)3 (1.9)9 (1.8)
 Central Midwest2,110 (7.8)2,068 (8.0)12 (3)11 (6.9)19 (3.9)
 Eastern Midwest4,291 (16.0)4,116 (15.9)96 (24.2)24 (15.1)55 (11.2)
 Mid-Atlantic4,427 (16.5)4,375 (16.9)29 (7.3)5 (3.1)18 (3.7)
 New England1,450 (5.4)1,400 (5.4)39 (9.8)5 (3.1)6 (1.2)
 Pacific Northwest3,142 (11.7)2,982 (11.5)60 (15.1)26 (16.4)74 (15)
 Southeast6,134 (22.8)6,108 (23.6)7 (1.8)8 (5)11 (2.2)
 Southwest4,983 (18.5)4,461 (17.3)144 (36.3)77 (48.4)301 (61.1)
Sociodemographic county-clustera
 Semi-Urban, High SES11,205 (41.4)10,910 (41.9)141 (35.5)46 (28.9)108 (21.6)
 Semi-Urban, Mid/Low SES973 (3.6)964 (3.7)1 (0.3)1 (0.6)7 (1.4)
 Semi-Urban, Hispanic2,846 (10.5)2,545 (9.8)59 (14.9)48 (30.2)194 (38.8)
 Mostly Rural, Mid SES6,779 (25)6,580 (25.3)91 (22.9)26 (16.4)82 (16.4)
 Rural, Mid/Low SES332 (1.2)327 (1.3)2 (0.5)1 (0.6)2 (0.4)
 Young, Urban, Mid/High SES786 (2.9)738 (2.8)19 (4.8)3 (1.9)26 (5.2)
 Rural, American Indian325 (1.2)294 (1.1)10 (2.5)4 (2.5)17 (3.4)
 Rural, High SES3,830 (14.1)3,662 (14.1)74 (18.6)30 (18.9)64 (12.8)
Tribal water systems [N (%)]291 (1.1)241 (<1)11 (2.8)8 (5)31 (6.3)
Top states represented (N)CA (1,987)CA (1,798)AZ (59)CA (24)CA (119)
 PA (1,810)PA (1,786)MI (54)TX (23)TX (80)
 NY (1,588)NY (1,564)CA (46)AZ (16)AZ (51)
Note: Compliance categories were assigned by the estimated 3-y average arsenic concentration during the first and second 3-y period. Low refers to estimated 3-y average arsenic concentrations 10μg/L. High refers to estimated 3-y average arsenic concentrations >10μg/L. States included in geologic regions are as follows: Alaska/Hawaii (Alaska and Hawaii), Central Midwest (North Dakota, South Dakota, Nebraska, Kansas, and Missouri), Eastern Midwest (Wisconsin, Illinois, Indiana, Michigan, Ohio, Minnesota, and Iowa), Mid-Atlantic (Pennsylvania, Maryland, District of Columbia, Delaware, New York, New Jersey, Connecticut, and Rhode Island), New England (Massachusetts, Vermont, New Hampshire, and Maine), Pacific Northwest (Washington, Oregon, Montana, Wyoming, and Idaho), Southeast (Oklahoma, Arizona, Louisiana, Mississippi, Alabama, Florida, Georgia, Tennessee, Kentucky, South Carolina, North Carolina, Virginia, and West Virginia), and Southwest (California, Nevada, Utah, Colorado, Arizona, New Mexico, and Texas). Population served is adjusted total population served, which accounts for systems that sell or purchase water to avoid overcounting. CI, confidence interval; SE, standard error; SES, socioeconomic status.
a
A total of 172 CWSs served more than one county; of these, approximately half served counties categorized to different sociodemographic county-clusters (e.g., NY7003493 serves New York, New York (Young, Urban, Mid/High SES) and Bronx, New York (Semi-Urban, Hispanic). These CWSs are represented for each county that they serve in the sociodemographic county-cluster analyses (N=27,076 with average arsenic estimates from both time periods available).
At the county level, counties were most likely categorized into the Low/Low category (N=2,228,98.5%), followed by the High/High category (N=18,<1%), the High/Low category (N=12,<1%), and the Low/High category when applying the 10μg/L cut point (N=4,<1%) (Figure 4; Excel Table S6). When evaluating county compliance categories applying the 5- and 1μg/L cut points, the majority of counties in the Central Midwest and Southwest United States had weighted average public drinking water arsenic concentrations exceeding these thresholds (Figure 4). In addition to the static maps presented here, an interactive map of county-level water arsenic concentrations and MCL compliance categories is available at https://annenigra.github.io/ColumbiaArsenicMap.html. Characteristics of counties by compliance category applying the 10-, 5-, and 1μg/L cut points were similar to findings at the CWS level (Excel Table S6) and support that the new MCL likely reduced water arsenic exposure for many communities, with water arsenic levels from 2006–2008 being well below the 10μg/L standard.
Figure 4. County-level compliance categories of estimated water arsenic concentrations in community water systems (CWSs) from 2006–2008 through 2009–2011 using (A) 10μg/L, (B) 5μg/L, and (C) 1μg/L cut points (N=26,895 CWSs serving N=2,262 counties with average arsenic estimates from both 3-y time periods available). Compliance categories were assigned based on 3-y average water arsenic concentrations from the two time periods of interest (2006–2008 vs. 2009–2011) using a cutoff of 10μg/L. Low refers to estimated 3-y average arsenic concentrations 10μg/L. High refers to estimated 3-y average arsenic concentrations >10μg/L. Counties identified as “missing one data period” had only one 3-y average water arsenic concentration for either 2006–2008 or 2009–2011; Counties identified as “no data available” did not have adequate data in the SYR3 for either time period. Note: SYR, Six-Year Review.
A total of 18 counties (served by a total of 81 CWSs) were categorized into the High/High compliance category (water arsenic >10μg/L during both time periods) (Excel Table S7). The counties were Payette, Idaho; Rawlins, Kansas; Lapeer, Michigan; Dundy and Polk, Nebraska; Esmeralda, Eureka, and Lander, Nevada; Lincoln, North Carolina; Socorro, New Mexico; Ramsey, North Dakota; and Andrews, Crane, Duval, Falls, Gaines, Hudspeth, and Jim Hogg, Texas. Excel Table S8 identifies the 12 counties categorized into the High/Low category (2006–2008 water arsenic >10μg/L; 2009–2011 water arsenic10μg/L) and the CWSs serving each county. These counties were Yavapai, Arizona; Scott, Kansas; Kalamazoo and Ogemaw, Michigan; Box Butte and Deuel, Nebraska; Lyon, Nevada; Sandoval and Valencia, New Mexico; Borden, Texas; and Millard and San Juan, Utah.

Discussion

Community water arsenic concentrations decreased significantly in the New England, Eastern Midwest, and Southwest regions of the United States from 2006 through 2011. Declines in CWS arsenic concentrations were largest at the higher end of the distribution and generally nonsignificant at quantiles below the 80th percentile, indicating that the decline occurred in places with elevated water arsenic concentrations, likely because of specific interventions to comply with the 10μg/L MCL. Significant declines also occurred for CWSs that were reliant on groundwater, serving smaller populations, and serving counties classified as Mostly Rural, Mid SES, Semi-Urban, High SES, and Rural, High SES. Despite the observed decline in CWS arsenic concentrations, several sociodemographic subgroups remained more likely exposed to arsenic concentrations in public drinking water exceeding the arsenic 10μg/L MCL. CWSs more likely to continue exceeding the arsenic MCL were those serving Hispanic communities, located in the Southwestern United States, reliant on groundwater, and serving smaller populations. These results could have important implications for future efforts aimed at reducing the arsenic MCL further and for the regulation of similar drinking water contaminants.
We previously estimated a 17% reduction in public water arsenic exposure nationwide (corresponding to a decline in water arsenic exposure of 1.42μg/L) from 2003 to 2014, using urinary arsenic concentrations measured in the National Health and Nutrition Examination Survey (NHANES) (corresponding to an estimated 200–900 avoided excess cases of lung and bladder cancer per year), but we were not able to assess differences by geographic subgroups and were limited by self-reported dietary and water source data (Nigra et al. 2017). In the present study, the average change in CWS arsenic concentration differed across MCL compliance categories using the 10μg/L cut point, from no change in Low/Low and High/High CWSs to 12.84μg/L in High/Low CWSs, indicating that CWSs in the High/Low compliance category reported relatively large reductions in accordance with MCL implementation compared with those in the High/High compliance category. At both the CWS and the county level, our findings indicate that a large majority of CWSs and U.S. counties (especially those in the western United States) had average public drinking water arsenic concentrations exceeding the more health protective regulatory standards of 5μg/L (the MCL for New Jersey and New Hampshire) and 1μg/L (the MCL for the Netherlands).
The present study indicates mean public drinking water arsenic concentrations decreased from 2006 to 2011 by 10% nationwide, with significant declines ranging from 11.4% in the Southwest to 36.6% in New England, and from 9.7% in Semi-Urban, High SES counties to 14.7% in Rural, High SES counties. These direct estimates of arsenic concentrations as reported in CWS monitoring records are not subject to potential confounding by diet and other arsenic exposure sources and further affirm that public health benefits resulting from the new arsenic MCL implementation were not uniform across the United States. Thus, any subsequent declines in arsenic-associated adverse health outcomes, which have not been adequately characterized, are likely differential by region and sociodemographic subgroup. Moreover, subgroups with the highest exposure from 2006 to 2008 (e.g., the Southwest and Semi-Urban Hispanic counties) could see greater reductions in arsenic-associated disease, even with smaller absolute reductions in water arsenic exposure, compared with subgroups with relatively low arsenic exposure from 2006 to 2008. Future work should evaluate whether MCL implementation has reduced arsenic-associated disease separately for these relevant subgroups.
The present study revealed substantial environmental justice concerns for largely Hispanic communities in the Southwest region that are served by CWSs that rely on groundwater and serve smaller populations. Of 18 counties categorized into the High/High compliance category using the 10μg/L cut point, 7 of these were Semi-Urban, Hispanic counties (38.9%) and 12 were in the Southwest region (66.7%). In addition, a recent study also found that Southwestern CWSs that exclusively serve correctional facilities reported arsenic concentrations that were twice as high as other Southwestern CWSs from 2006 to 2011 (Nigra and Navas-Acien 2020). Together, these findings highlight significant environmental justice concerns for drinking water arsenic exposure in the United States, even among communities reliant on public water systems. Additional financial and technical resources and interventions are needed to ensure that communities served by these CWSs are protected from elevated drinking water arsenic exposure. The present analysis is likely not representative of tribal community water systems or of counties in the Rural, American Indian sociodemographic county-cluster given that less than 50% of the more than 700 tribal systems currently operating were included in our analysis, limiting our ability to draw conclusions about whether tribal CWSs in the United States remain at risk for elevated water arsenic exposure. The present study highlights the importance of collecting tribal public water systems data in the future SYR so that analyses can evaluate tribal water systems and Rural, American Indian counties, especially because the establishment of CWSs in some tribal communities has likely been critical for reducing water arsenic exposure (Navas-Acien et al. 2009; Thomas et al. 2019). We also found higher estimated water arsenic concentrations for CWSs and counties in the Rural, High SES sociodemographic cluster, which could reflect that these counties are predominately located in the Central Midwest where water arsenic concentrations were high (see Wallace et al. 2019).
We present CWS- and county-level estimates of arsenic exposure in public drinking water across the United States for 2006–2011, leveraging two of the largest U.S. EPA databases of public water available: SYR3 and SDWIS. The time period analyzed in this study (2006–2011) coincides with the U.S. EPA’s MCL regulatory change, which was enacted in 2001 and went into effect in 2006. Although not formalized in the Final Arsenic Rule, training documents indicate that an initial monitoring period gave systems time to comply with the new MCL (U.S. EPA 2002). Depending on whether the primary source of water was surface or groundwater, systems had until December 2006 or 2007 to collect initial monitoring samples, respectively. To allow time to address noncompliance (typically through installing treatment systems or switching or mixing water sources), an additional year of sampling was allowed for systems whose initial monitoring samples exceeded the MCL (U.S. EPA 2002, 2004). SYR data beyond 2011 is not yet available. The U.S. EPA estimates that data from the Fourth Six-Year Review (SYR4) will be made available in 2023 and future work should analyze these data to evaluate continued compliance with the 10μg/L MCL.
Because of the additional cost of installing and operating arsenic treatment systems, small CWSs were anticipated to experience difficulty reducing water arsenic concentrations below the 10μg/L MCL. Although we found the largest decreases in 3-y average CWS arsenic concentrations among CWSs that served the smallest population sizes, we also found that CWSs in the High/High and Low/High compliance categories were more likely to serve smaller populations than CWSs in other compliance categories. Therefore, although the largest decreases in average arsenic concentrations have occurred among smaller CWSs, numerous smaller CWSs were still struggling to reduce water arsenic concentrations to <10μg/L as of 2009–2011. These findings are consistent with the focus of federal and state programs (Indian Health Service, U.S. EPA, U.S. Department of Agriculture, New Mexico, California) that provide financial and technical assistance to small public water systems. Federal and state efforts are still needed to support those systems not in compliance.
Data quality limitations of the SYR3 must be considered. First, due to the voluntary data submission, states and U.S. EPA Regional offices that did not submit monitoring records for the SYR3 (e.g., Colorado and Mississippi; U.S. EPA for tribal public water systems) had significantly higher arsenic violation rates reported in SDWIS compared with states and agencies that did contribute monitoring records (U.S. EPA 2016a). Second, reported LODs differed substantially across CWSs and were often missing from monitoring records; moreover, we found several reported LODs that were likely reported with incorrect units. We assigned county-served based on reported ZIP code for 418 CWSs that could not be matched automatically to our SDWIS database; however, these ZIP codes could represent administrative addresses and may not reflect the distribution system and the population served. Still, the quality assurance and quality control measures implemented by the U.S. EPA prior to the publication of the SYR3 were extensive and addressed many known data quality-related challenges (U.S. EPA 2016b). We categorized CWS source water type using reported source water type in SDWIS (categorized as surface vs. groundwater). It is possible that some CWSs may have changed source water type over time, which could influence our findings, which are stratified by source water type.
Unlike epidemiologic research on air quality, which often relies on national monitoring networks for exposure assessments (e.g., the National Air Toxics Assessment), epidemiologic research on drinking water contaminants remains limited by the lack of a centralized, national monitoring network of public drinking water. The merging of SDWIS with SYR3 enables nationwide epidemiologic studies assigning estimated water arsenic exposures to populations reliant on public drinking water. CWS-level estimates of exposure may be more relevant for populations residing in counties with significant heterogeneity of public water arsenic concentrations. The association between estimates of CWS arsenic concentrations and biomarkers of inorganic arsenic exposure can also be evaluated in epidemiologic cohorts with biomarker and dietary recall data available, including national surveys (e.g., NHANES) and other studies. The quality of the SYR databases could be improved and used for this purpose by including higher-resolution geographical distribution data beyond county-served (e.g., census-tract served) and implementing additional efforts and compensation for obtaining a more complete reporting of regulated contaminant monitoring records and rigorous oversight of data entry and quality. Some monitoring records were excluded from the present analysis because of missing data, data quality concerns, and difficulties in merging a small number of records in SYR3 with SDWIS.
The present study indicates that the U.S. EPA’s Arsenic Rule and MCL change was successful in reducing public drinking water arsenic exposure for most subgroups of the U.S. population and provides environmental exposure estimates for further epidemiologic investigation, although certain subgroups remain more highly exposed. Future work should validate these CWS- and county-level public water arsenic exposures with biomarkers of inorganic arsenic exposure in epidemiologic cohorts and should evaluate whether communities and counties with decreased estimated public water arsenic concentrations experienced a subsequent reduction in arsenic-associated health outcomes (e.g., cardiovascular disease, related cancers, adverse birth outcomes). Stronger regulations, targeted compliance enforcement, and continued state and federal funding for infrastructure and technical assistance support for public water systems is needed to reduce inequalities and to further protect numerous communities in the United States affected by elevated drinking water arsenic exposure.

Acknowledgments

This study was supported by National Institutes of Health/National Institute of Environmental Health Sciences (NIEHS) grants P42ES010349, P30ES009089, R01ES028758, and R21ES029668. A.E.N. was also supported by NIEHS grants 5T32ES007322 and F31ES029799.

Article Notes

The authors declare they have no actual or potential competing financial interests.

Supplementary Material

File (ehp7313.smcontents.508.pdf)
File (ehp7313.s001.acco.pdf)
File (ehp7313.s002.codeanddata.acco.zip)

References

Allen M, Poggiali D, Whitaker K, Marshall TR, Kievit R. 2019. Raincloud plots: a multi-platform tool for robust data visualization. Wellcome Open Res 4:63. https://pubmed.ncbi.nlm.nih.gov/31069261/, https://doi.org/10.12688/wellcomeopenres.15191.1.
Awata H, Linder S, Mitchell LE, Delclos GL. 2017a. Association of dietary intake and biomarker levels of arsenic, cadmium, lead, and mercury among Asian populations in the United States: NHANES 2011–2012. Environ Health Perspect 125(3):314–323. https://pubmed.ncbi.nlm.nih.gov/27586241/, https://doi.org/10.1289/EHP28.
Awata H, Linder S, Mitchell LE, Delclos GL. 2017b. Biomarker levels of toxic metals among Asian populations in the United States: NHANES 2011–2012. Environ Health Perspect 125(3):306–313. https://pubmed.ncbi.nlm.nih.gov/27517362/, https://doi.org/10.1289/EHP27.
Ayotte JD, Medalie L, Qi SL, Backer LC, Nolan BT. 2017. Estimating the high-arsenic domestic-well population in the conterminous United States. Environ Sci Technol 51(21):12443–12454. https://pubmed.ncbi.nlm.nih.gov/29043784/, https://doi.org/10.1021/acs.est.7b02881.
Becker RA, Wilks AR. 2018. maps: draw geographical maps. Version 3.3.0. https://CRAN.R-project.org/package=maps [accessed 9 November 2020].
Cubadda F, Jackson BP, Cottingham KL, Van Horne YO, Kurzius-Spencer M. 2017. Human exposure to dietary inorganic arsenic and other arsenic species: state of knowledge, gaps and uncertainties. Sci Total Environ 579:1228–1239. https://pubmed.ncbi.nlm.nih.gov/27914647/, https://doi.org/10.1016/j.scitotenv.2016.11.108.
Environmental Integrity Project. 2016. Arsenic in California drinking water. https://environmentalintegrity.org/wp-content/uploads/CA-Arsenic-Report.pdf [accessed 9 November 2020].
Foster SA, Pennino MJ, Compton JE, Leibowitz SG, Kile ML. 2019. Arsenic drinking water violations decreased across the United States following revision of the maximum contaminant level. Environ Sci Technol 53(19):11478–11485. https://pubmed.ncbi.nlm.nih.gov/31502444/, https://doi.org/10.1021/acs.est.9b02358.
Gibson JM, Fisher M, Clonch A, MacDonald JM, Cook PJ. 2020. Children drinking private well water have higher blood lead than those with city water. Proc Natl Acad Sci USA 117(29):16898–16907. https://pubmed.ncbi.nlm.nih.gov/32631989/, https://doi.org/10.1073/pnas.2002729117.
IARC (International Agency for Research on Cancer). 2004. Some Drinking water Disinfectants and Contaminants, including Arsenic. https://publications.iarc.fr/Book-And-Report-Series/Iarc-Monographs-On-The-Identification-Of-Carcinogenic-Hazards-To-Humans/Some-Drinking-Water-Disinfectants-And-Contaminants-Including-Arsenic-2004 [accessed 9 November 2020].
Jones MR, Tellez-Plaza M, Vaidya D, Grau-Perez M, Post WS, Kaufman JD, et al. 2019. Ethnic, geographic and dietary differences in arsenic exposure in the Multi-Ethnic Study of Atherosclerosis (MESA). J Expo Sci Environ Epidemiol 29(3):310–322. https://pubmed.ncbi.nlm.nih.gov/29795237/, https://doi.org/10.1038/s41370-018-0042-0.
Kuo CC, Howard BV, Umans JG, Gribble MO, Best LG, Francesconi KA, et al. 2015. Arsenic exposure, arsenic metabolism, and incident diabetes in the Strong Heart Study. Diabetes Care 38(4):620–627. https://pubmed.ncbi.nlm.nih.gov/25583752/, https://doi.org/10.2337/dc14-1641.
Kuo CC, Moon KA, Wang SL, Silbergeld E, Navas-Acien A. 2017. The association of arsenic metabolism with cancer, cardiovascular disease, and diabetes: a systematic review of the epidemiological evidence. Environ Health Perspect 125(8):087001. https://pubmed.ncbi.nlm.nih.gov/28796632/, https://doi.org/10.1289/EHP577.
Kurzius-Spencer M, Burgess JL, Harris RB, Hartz V, Roberge J, Huang S, et al. 2014. Contribution of diet to aggregate arsenic exposures: an analysis across populations. J Expo Sci Environ Epidemiol 24(2):156–162. https://pubmed.ncbi.nlm.nih.gov/23860400/, https://doi.org/10.1038/jes.2013.37.
Milton AH, Hussain S, Akter S, Rahman M, Mouly TA, Mitchell K. 2017. A review of the effects of chronic arsenic exposure on adverse pregnancy outcomes. Int J Environ Res Public Health 14(6):556. https://pubmed.ncbi.nlm.nih.gov/28545256/, https://doi.org/10.3390/ijerph14060556.
Moody EC, Coca SG, Sanders AP. 2018. Toxic metals and chronic kidney disease: a systematic review of recent literature. Curr Environ Health Rep 5(4):453–463. https://pubmed.ncbi.nlm.nih.gov/30338443/, https://doi.org/10.1007/s40572-018-0212-1.
Moon KA, Oberoi S, Barchowsky A, Chen Y, Guallar E, Nachman KE, et al. 2017. A dose-response meta-analysis of chronic arsenic exposure and incident cardiovascular disease. Int J Epidemiol 46(6):1924–1939. https://pubmed.ncbi.nlm.nih.gov/29040626/, https://doi.org/10.1093/ije/dyx202.
Morello-Frosch R, Pastor M Jr, Porras C, Sadd J. 2002. Environmental justice and regional inequality in southern California: implications for future research. Environ Health Perspect 110(suppl 2):149–154. https://pubmed.ncbi.nlm.nih.gov/11929723/, https://doi.org/10.1289/ehp.02110s2149.
Navas-Acien A, Umans JG, Howard BV, Goessler W, Francesconi KA, Crainiceanu CM, et al. 2009. Urine arsenic concentrations and species excretion patterns in American Indian communities over a 10-year period: the Strong Heart Study. Environ Health Perspect 117(9):1428–1433. https://pubmed.ncbi.nlm.nih.gov/19750109/, https://doi.org/10.1289/ehp.0800509.
Nigra AE. 2020. Environmental racism and the need for private well protections. Proc Natl Acad Sci USA 117(30):17476–17478. https://pubmed.ncbi.nlm.nih.gov/32641505/, https://doi.org/10.1073/pnas.2011547117.
Nigra AE, Navas-Acien A. 2020. Arsenic in US correctional facility drinking water, 2006–2011. Environ Res 188:109768. https://pubmed.ncbi.nlm.nih.gov/32585331/, https://doi.org/10.1016/j.envres.2020.109768.
Nigra AE, Sanchez TR, Nachman KE, Harvey D, Chillrud SN, Graziano JH, et al. 2017. The effect of the Environmental Protection Agency maximum contaminant level on arsenic exposure in the USA from 2003 to 2014: an analysis of the National Health and Nutrition Examination Survey (NHANES). Lancet Public Health 2(11):e513–e521. https://pubmed.ncbi.nlm.nih.gov/29250608/, https://doi.org/10.1016/S2468-2667(17)30195-0.
Ruggles S, Flood S, Goeken R, Grover J, Meyer E, Pacas J, et al. 2019. Integrated Public Use Microdata Series: Version 9.0 [dataset]. https://doi.org/10.18128/D010.V9.0 [accessed 9 November 2020].
Sanchez TR, Powers M, Perzanowski M, George CM, Graziano JH, Navas-Acien A. 2018. A meta-analysis of arsenic exposure and lung function: is there evidence of restrictive or obstructive lung disease? Curr Environ Health Rpt 5(2):244–254. https://pubmed.ncbi.nlm.nih.gov/29637476/, https://doi.org/10.1007/s40572-018-0192-1.
Schaider LA, Swetschinski L, Campbell C, Rudel RA. 2019. Environmental justice and drinking water quality: are there socioeconomic disparities in nitrate levels in U.S. drinking water? Environ Health 18(1):3. https://pubmed.ncbi.nlm.nih.gov/30651108/, https://doi.org/10.1186/s12940-018-0442-6.
Thomas ED, Gittelsohn J, Yracheta J, Powers M, O’Leary M, Harvey DE, et al. 2019. The Strong Heart Water Study: informing and designing a multi-level intervention to reduce arsenic exposure among private well users in Great Plains Indian Nations. Sci Total Environ 650(pt 2):3120–3133. https://pubmed.ncbi.nlm.nih.gov/30373089/, https://doi.org/10.1016/j.scitotenv.2018.09.204.
U.S. EPA (U.S. Environmental Protection Agency). 2001a. Arsenic and Clarifications to Compliance and New Source Monitoring Rule: a Quick Reference Guide. https://denr.sd.gov/des/dw/PDF/Quick%20Reference%20Guides/quickguide.pdf [accessed 9 November 2020].
U.S. EPA. 2001b. National primary drinking water regulations: arsenic and clarifications to compliance and new source contaminants monitoring. Fed Reg 66(14):6976–7066. https://www.govinfo.gov/content/pkg/FR-2001-01-22/pdf/FR-2001-01-22.pdf [accessed 9 November 2020].
U.S. EPA. 2002. The Arsenic Rule. Compliance, Reporting, and Enforcement Issues. https://www.epa.gov/sites/production/files/2015-09/documents/train4-compliance.pdf [accessed 9 November 2020].
U.S. EPA. 2004. The Standardized Monitoring Framework: a Quick Reference Guide. https://nepis.epa.gov/Exe/ZyPDF.cgi/3000667K.PDF?Dockey=3000667K.PDF [accessed 20 July 2020].
U.S. EPA. 2016a. The Analysis of Regulated Contaminant Occurrence Data from Public Water Systems in Support of the Third Six-Year Review of National Primary Drinking Water Regulations: Chemical Phase Rules and Radionuclides Rules. EPA-810-R-16-014. https://www.epa.gov/sites/production/files/2016-12/documents/810r16014.pdf [accessed 20 July 2020].
U.S. EPA. 2016b. The Data Management and Quality Assurance/Quality Control Process for the Third Six-Year Review Information Collection Rule Dataset. EPA-810-R-16-015. https://www.epa.gov/sites/production/files/2016-12/documents/810r16015_0.pdf [accessed 20 July 2020].
U.S. EPA. 2016c. User Guide to Downloading and Using SYR3 Data from EPA’s Website. https://www.epa.gov/sites/production/files/2016-12/documents/user_guide_to_obtaining_and_using_syr3_data.pdf [accessed 9 November 2020].
U.S. EPA. 2017. Safe Drinking Water Information System (SDWIS) Federal Reporting Services. https://www.epa.gov/ground-water-and-drinking-water/safe-drinking-water-information-system-sdwis-federal-reporting [accessed 9 November 2020].
Wallace M, Sharfstein JM, Kaminsky J, Lessler J. 2019. Comparison of US county-level public health performance rankings with county cluster and national rankings: assessment based on prevalence rates of smoking and obesity and motor vehicle crash death rates. JAMA Netw Open 2(1):e186816. https://pubmed.ncbi.nlm.nih.gov/30646196/, https://doi.org/10.1001/jamanetworkopen.2018.6816.
Welch B, Smit E, Cardenas A, Hystad P, Kile ML. 2018. Trends in urinary arsenic among the U.S. population by drinking water source: results from the National Health and Nutritional Examinations Survey 2003–2014. Environ Res 162:8–17. https://pubmed.ncbi.nlm.nih.gov/29272814/, https://doi.org/10.1016/j.envres.2017.12.012.
West Virginia Department of Health and Human Services. 2019. Drinking Water Watch Application. http://129.71.204.189:1977/DWWpublic/ [accessed 9 November 2020].
Wilson SM, Heaney CD, Cooper J, Wilson O. 2008. Built environment issues in unserved and underserved African-American neighborhoods in North Carolina. Environ Justice 1(2):63–72, https://doi.org/10.1089/env.2008.0509.
Wisconsin Department of Natural Resources. 2019. Public Drinking Water System Data Viewer. https://dnr.wi.gov/dwsviewer/ [accessed 9 November 2020].

Information & Authors

Information

Published In

Environmental Health Perspectives
Volume 128Issue 12December 2020
PubMed: 33295795

History

Received: 24 April 2020
Revision received: 2 November 2020
Accepted: 3 November 2020
Published online: 9 December 2020

Authors

Affiliations

Anne E. Nigra
Department of Environmental Health Sciences, Columbia University Mailman School of Public Health, New York, New York, USA
Qixuan Chen
Department of Biostatistics, Columbia University Mailman School of Public Health, New York, New York, USA
Steven N. Chillrud
Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York, USA
Lili Wang
Office of Water, U.S. Environmental Protection Agency, Washington, DC, USA
David Harvey
United States Public Health Service, Rockville, Maryland, USA
Brian Mailloux
Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York, USA
Barnard College, Columbia University, New York, New York, USA
Pam Factor-Litvak
Department of Environmental Health Sciences, Columbia University Mailman School of Public Health, New York, New York, USA
Department of Epidemiology, Columbia University Mailman School of Public Health, New York, New York, USA
Ana Navas-Acien
Department of Environmental Health Sciences, Columbia University Mailman School of Public Health, New York, New York, USA

Notes

Address correspondence to Anne E. Nigra, Department of Environmental Health Sciences, Columbia University Mailman School of Public Health, 722 W. 168th St., 11th Floor, Room 1107, New York, NY 10032 USA. Email: [email protected]

Metrics & Citations

Metrics

About Article Metrics


Citations

Download citation

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click DOWNLOAD.

Cited by

  • Quantifying Disparities in Per- and Polyfluoroalkyl Substances (PFAS) Levels in Drinking Water from Overburdened Communities in New Jersey, 2019–2021, Environmental Health Perspectives, 10.1289/EHP12787, 132, 4, (2024).
  • Effect of an Arsenic Mitigation Program on Arsenic Exposure in American Indian Communities: A Cluster Randomized Controlled Trial of the Community-Led Strong Heart Water Study Program, Environmental Health Perspectives, 10.1289/EHP12548, 132, 3, (2024).
  • Potential health risk assessment of mixtures of heavy metals in drinking water, Groundwater for Sustainable Development, 10.1016/j.gsd.2024.101147, 25, (101147), (2024).
  • Endoplasmic reticulum stress responses and epigenetic alterations in arsenic carcinogenesis, Environmental Pollution, 10.1016/j.envpol.2024.123565, 347, (123565), (2024).
  • Prioritizing water availability study settings to address geogenic contaminants and related societal factors, Environmental Monitoring and Assessment, 10.1007/s10661-024-12362-2, 196, 3, (2024).
  • The Influence of Arsenic Co-Exposure in a Model of Alcohol-Induced Neurodegeneration in C57BL/6J Mice, Brain Sciences, 10.3390/brainsci13121633, 13, 12, (1633), (2023).
  • Geospatial Assessment of Racial/Ethnic Composition, Social Vulnerability, and Lead Water Service Lines in New York City, Environmental Health Perspectives, 10.1289/EHP12276, 131, 8, (2023).
  • Arsenic Exposure and Epigenetic Aging: The Association with Cardiovascular Disease and All-Cause Mortality in the Strong Heart Study, Environmental Health Perspectives, 10.1289/EHP11981, 131, 12, (2023).
  • Invited Perspective: Toxic Metals and Hypertensive Disorders of Pregnancy, Environmental Health Perspectives, 10.1289/EHP11963, 131, 4, (2023).
  • Paradox Lost? The Waning Health Advantage among the U.S. Hispanic Population, Environmental Health Perspectives, 10.1289/EHP11618, 131, 1, (2023).
  • See more

View Options

View options

PDF

View PDF

Get Access

Restore your content access

Enter your email address to restore your content access:

Note: This functionality works only for purchases done as a guest. If you already have an account, log in to access the content to which you are entitled.

Media

Figures

Other

Tables

Share

Share

Copy the content Link

Share on social media