Aggregate-level Lead Exposure and Child Maltreatment

Abstract

The purpose of this study was to examine the possible ecological association between aggregate blood lead levels (BLL) and rates of child maltreatment. To this end, we employed an ecologic study design, analyzing results from 59,645 child BLL tests between the years 1996 and 2007, and 6,640 substantiated maltreatment investigations from 2006 to 2016 in a large Midwest city. Separate Bayesian spatial Poisson conditional autoregressive (CAR) and Bayesian spatial zero-inflated Poisson CAR models were used to predict the occurrence of maltreatment.

Bivariate results showed that aggregate rates of maltreatment increased as aggregate BLL increased. Multivariate results showed that medium-exposure BLL census tracts (OR = 1.38) and high-exposure BLL tracts (OR = 1.38) had increased odds of substantiated investigations for any maltreatment compared to low BLL census tracts even after controlling for crime rates, age of the housing stock, and concentrated disadvantage. Our findings, considered with prior research, continue to reveal a confluence of deleterious outcomes in areas where exposure to lead seems elevated. In this case, child maltreatment also appears to represent a macro-level correlate of aggregate lead exposure. Yet our results preclude any causal inference, and further research on the intersection of child maltreatment with environmental toxins is needed to determine if contaminant abatement should be considered as a possible maltreatment prevention strategy.

Keywords

prevention of child abuse child abuse community violence sexual abuse physical abuse

Introduction

Within a public health framework, the etiological factors precipitating child abuse and neglect are not only found within parent–child dyads or family homes, but also from the community environment those homes are embedded within (Herrenkohl et al., 2016). Previous research has suggested, for example, that there is a link between neighborhood factors and child abuse, neglect, and fatalities (Farrell et al., 2017). Specifically, various aggregate variables that correlate with involvement with Child Protection Services (CPS) have included poverty (Frioux et al., 2014), racial composition (Maguire-Jack et al., 2020), housing instability (Coulton et al., 2018), and concentration of alcohol outlets (Morton et al., 2014). While far from representing durable causes of maltreatment, these connections between environment and maltreatment are nonetheless important. Understanding risk factors within neighborhoods can help to guide financial resources to probe causal pathways and ultimately inform primary and tertiary prevention efforts (Commission to Eliminate Child Abuse and Neglect Fatalities, 2016).

When viewing this topic through the lens of public health, however, one notable neighborhood characteristic missing is exposure to environmental contamination, including hazardous wastes, air and water pollutants, and heavy metal toxins. Families living in low-income neighborhoods are disproportionately exposed to environmental hazards, and this exposure can have far-reaching negative physical and developmental consequences (Evans & Kantrowitz, 2002). In prior decades, one prominent toxin often found in paint, piping, water, and soil that has received considerable research attention in other fields is lead.

Young children seem particularly sensitive to lead burden and appear vulnerable to various deleterious impacts of elevated BLL (Winter & Sampson, 2017). Lead exposure in children may disrupt normative neurological development (Cecil et al., 2008), and adversely affect executive functioning, impulsivity, and emotional regulation (Bellinger, 2008). Lead exposure has been longitudinally correlated in human studies with a host of adverse adult outcomes, including poor health, lower intelligence, crime, and violence (Bellinger, 2017; Patrick, 2006). And prior research, some using the data included herein, has revealed correlations at more macro-levels between higher aggregate lead levels and both violent and non-violent crime (Boutwell et al., 2016; Stretesky & Lynch, 2004).

Should a correlation emerge between child maltreatment and lead exposure, one possible reason for such an association may involve the neurotoxic effects of lead burden in young children (Jaffee, 2019). More specifically, the increased rates of child maltreatment could owe their existence to the neurobehavioral changes that, over time, contribute to increased aggression of parents (see, Jaffee, 2019). However, no study to which we are aware has examined the possible connection between elevated BLL and child maltreatment. We hypothesize that environments with higher lead exposure levels among residents would also exhibit more instances of child abuse and neglect irrespective of crime, concentrated disadvantage, and age of housing stock. In this study, we spatially analyzed aggregate lead levels across a community to assess possible associations with child maltreatment rates. Because CPS investigator assessment of risk may be biased due to race, which inflates rates of substantiations (Dettlaff et al., 2011), we controlled for racial composition as part of our concentrated disadvantage variable.

Methods

We assessed the association between aggregate elevated BLLs and instances of substantiated investigations of child maltreatment and neglect across census tracts (n = 106) in Saint Louis, Missouri using an ecologic study design. Based on prior literature, crime rates, age of the housing stock, and concentrated disadvantage were considered important covariates and included in the analyses (Boutwell et al., 2016). Four datasets were merged for this study. Maltreatment data were obtained from the Missouri Children’s Division. Lead data were obtained from the Missouri Department of Health and Senior Services Health Strategic Architecture and Information Cooperative (MOHSAIC). Crime data were obtained from the Saint Louis Metro Police Department for the years 2010–2012. Finally, data on housing stock and socioeconomic factors were obtained from the U.S. Census’s 2008–2012 American Community Survey (ACS). The Institutional Review Board at Saint Louis University and the Privacy Review Board of the Department of Social Services Children’s Division of Missouri approved this study.

Measures

Maltreatment.

The locations of all substantiated investigations of child maltreatment within Saint Louis city limits from January 2006 to December 2016 were analyzed. Of the total 7,106 substantiated investigations within urban city limits, 6,927 (97%) were geocoded. Addresses that could not be geocoded (n = 79), or were located outside of the study area (the boundary lines of the city; n = 287), were omitted, rendering 6,640 substantiated investigations for analysis. Substantiated investigations were then aggregated to the census tract level by counting the number of occurrences within the corresponding census tract boundary. Four primary outcomes were tabulated: (a) total investigations, (b) physical abuse investigations, including allegations of physical injury inflicted on a child other than by accidental means, (c) sexual abuse investigations, including allegations of any sexual or sexualized interaction with a child, and (d) emotional, medical, or physical child neglect investigations, including failure to provide care necessary for a child’s basic wellbeing.

Blood lead levels.

We aggregated all BLL tests of 59,645 children <72 months in age and who had BLL tests performed between 1996 and 2007, to the individual’s residential census tract and then calculated the proportion of BLL tests that registered ≥5 µg/DL (micrograms per deciliter) within each census tract. Elevated BLL proportions at the census tract were then categorized into an ordinal scale based on proportions ≥5 µg/DL. The corresponding values were low (7.5–5.1%), medium (35.2–54.2%), and high (54.3–73.8%) exposure.

Crimes.

We also geocoded the locations of all 90,433 crimes (95% of all crimes), both violent and non-violent, reported to the police for the years 2010–2012. Crimes were aggregated to the census tract level and the mean crime rate per census tract was calculated by also utilizing data from the 5-year population estimate from the 2008–2012 ACS (United States Census Bureau, 2015). Census tract crime rates were log-transformed to achieve a normal distribution for regression analysis.

Concentrated disadvantage index.

Consistent with the measure proposed by Sampson et al. (2008), we used principal components analysis of eight variables from the 2008–2012 ACS without rotation to estimate the concentrated disadvantage index. The index was comprised of the proportion of households below the federal poverty line, population that was African-American, households unemployed during the past year, household use of public health insurance programs, households with children <18 years, female-headed households, educational attainment less than high school graduation, and the inverse of median household income (mean-centered). The constructed component explained 67.0% of the total variance of the 8 variables.

Housing age.

The mean age of housing units per census tract was drawn from the 2008–2012 ACS in order to account for a persistent lingering source of lead exposure (primarily via lead-based paints and dust containing lead particles in older housing stock).

Analysis

Moran’s I tests of residuals identified that spatial autocorrelation was present for all outcomes (p < .001) indicating the need for spatially-adjusted statistical models. Therefore, we fit separate Bayesian spatial Poisson CAR models using Integrated Nested Laplace Approximation (INLA; Martins et al., 2013) to predict the occurrence of (a) total, (b) physical abuse, and (c) sexual abuse investigations at the census tract level. For the emotional, medical, or physical child neglect investigation outcome, we fit a Bayesian spatial zero-inflated Poisson (ZIP) CAR model to adjust for excessive zero counts across census tracts (Manh et al., 2011). INLA models, via Gaussian Markov random fields, reduce computation time while also accounting for the spatial adjacency of census tracts (Rue & Martino, 2007). Spatial adjacency of census tracts was defined using the queen contiguity approach. All models were fit with an offset, where the offset for the ith census tract was the natural log of the total population of the ith census tract. All models employed non-informative zero-mean normal priors using the default and recommended settings for the precision. We used the R-INLA package in version 3.4.0 of R for all analyses. All models were adjusted for the aforementioned relevant covariates. We then exponentiated the model coefficients (i.e., exp[β]) in order to present odds ratios (OR) with corresponding 95% credible intervals.

Results

As seen in Table 1, rates of maltreatment increased in step with level of elevated BLL per census tract, with rates (per 1,000) of any types of maltreatment of 29 for low BLL, 59 for medium BLL, and 72 for high BLL. Medium BLL tracts (OR = 1.38; 95% CI = 1.08–1.76) and high BLL tracts (OR = 1.38; 95% CI = 1.08–1.76) were significantly associated with an increased odds of substantiated investigations for any maltreatment compared to low BLL tracts even after controlling for crime, housing age, concentrated disadvantage, and spatial autocorrelation. Likewise, elevated BLLs were significantly associated with physical abuse (medium BLL OR = 1.68; 95% CI = 1.35–2.11; high BLL OR = 1.37, 95% CI = 1.07–1.75), and sexual abuse (medium BLL OR = 1.36; 95% CI = .99–1.85; high BLL OR = 1.58; 95% CI = 1.15–2.21). Concentrated disadvantage (OR = 1.22; 95% CI = 1.08–1.38) and crime (OR = 1.69; 95% CI = 1.37–2.10) were also associated with increased odds of total substantiated investigations for any abuse/neglect.

Discussion

Our results suggest that elevated BLLs at the level of census tracts are correlated with not only crime as has been previously found (Boutwell et al., 2016; Stretesky & Lynch, 2004), but also with child abuse and neglect. Even after adjusting for important community and spatial correlates, lead exposure was positively related to all types of substantiated investigations of maltreatment, including physical and sexual abuse. Further research on the intersection of child maltreatment with environmental toxins is needed, however, to adequately determine if lead abatement might be considered as a maltreatment prevention strategy. Equally relevant, moreover, is that a public health model to prevent child abuse and neglect endorses a collaborative agency approach, bringing together child welfare, law enforcement, health care, and mental health providers (Commission to Eliminate Child Abuse and Neglect Fatalities, 2016). This public health collaboration may also benefit from an expansion to such agencies as the Department of Housing and Urban Development Office of Lead Hazard Control and Healthy Homes or others at the forefront of lead abatement.

A key limitation of this study was the inability to infer causality, given the ecological nature of the data. Exemplifying this limitation was the inability to know if individuals who maltreated their children were exposed to lead at any point in their development, or if maltreated children had ever been exposed. Exposures were aggregated and used as a general barometer for lead burden in a given census tract, thus assessing time order between exposure and outcome was ultimately not possible in the current article. Generalizability was limited because our analysis was conducted only in Saint Louis City, which has a high proportion of African-American citizens living in economically poor neighborhoods with older housing, yet lacks ethnic diversity in other respects. It is a strength of our analysis that we could closely examine a minority group known to be at risk for environmental contagions in the broader United States. However, this limits the generalizability of our findings to other more ethnically diverse populations and rural communities. That said, we intended this as a preliminary assessment regarding the correlation of lead exposure and child maltreatment at the aggregate-level. As such, we argue that this represents a launch pad for future research—at both the individual and aggregate levels—not the landing pad for causal inference or firm policy recommendations.

Prior to the 1980s, lead was a common additive found in paint, gas, water pipes, and other household and industrial products. Half (52%) of all homes built before 1978 have some lead-based paint, according to the 2011 American Healthy Homes Survey (U.S. Department of Housing and Urban Development, 2011). In fact, more than half of the neighborhoods in Saint Louis city have a median housing age of 100 years or more (Nguyen & O’Dea, 2019). While BLLs have certainly decreased over time in the United States (Wheeler & Brown, 2013), there remains upwards of three million homes that have lead-based paint where children under 6 years of age live (Dewalt et al., 2015). High profile cases, such as relatively recent incidents in Flint, Michigan, continue to demonstrate that lead can pose a lingering health problem, especially for the most disadvantaged families in the United States.

Table 1.

Crude Rates and Regression Estimates by Type of Abuse Allegations Across Census Tracts (n = 106) in St. Louis, Missouri, 2006–2016.

	Substantiated Allegations		Physical Allegations		Sexual Allegations		Emotional, Medical, or Physical Neglect Allegations
Crude rates per 1,000 people	Mean ± SD	95% CI	Mean ± SD	95% CI	Mean ± SD	95% CI	Mean ± SD	95% CI
Blood lead levels
Low	29.3 ± 19.5	(22.7, 35.9)	5.2 ± 3.8	(3.9, 6.5)	3.9 ± 3.7	(2.7, 5.2)	1.1 ± 1.4	(.6, 1.6)
Medium	59.2 ± 25.0	(50.7, 67.6)	13.8 ± 7.5	(11.3, 16.4)	6.2 ± 4.5	(4.6, 7.7)	3.0 ± 2.7	(1.9, 4.1)
High	71.5 ± 27.9	(61.8, 81.3)	16.8 ± 8.3	(13.9, 19.6)	9.9 ± 5.3	(8.1, 11.8)	3.9 ± 3.4	(2.7, 5.0)
Total	53.0 ± 30.3	(47.2, 58.8)	11.8 ± 8.5	(10.2, 13.4)	6.6 ± 5.2	(5.3, 7.6)	2.6 ± 3.0	(2.1, 3.2)

Spatial regression results	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Blood lead levels
Low	1.0	reference	1.0	reference	1.0	reference	1.0	reference
Medium	1.38	(1.08, 1.76)	1.68	(1.35, 2.11)	1.36	(.99, 1.85)	1.07	(.70, 1.66)
High	1.43	(1.08, 1.91)	1.37	(1.07, 1.75)	1.58	(1.15, 2.21)	.72	(.46, 1.17)
Housing age	1.01	(.99, 1.01)	1.02	(1.01, 1.03)	1.02	(1.01, 1.03)	1.02	(1.01, 1.04)
Concentrated disadvantage	1.22	(1.08, 1.38)	1.34	(1.19, 1.50)	1.25	(1.06, 1.47)	1.73	1.34, 2.24)
Log(Crimes)	1.69	(1.37, 2.10)	1.52	(1.21, 1.91)	1.30	(.95, 1.80)	1.78	(1.10, 2.84)

Note. SD = Standard deviation; CI = Confidence interval for rates and credible interval for regression estimates; OR = Odds ratio.

Footnotes

Author’s Note

Brian B. Boutwell is also affiliated with University of Mississippi, Oxford MS. Jisuk Seon is now affiliated with Kyungnam University, Republic of Korea.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Jesse J. Helton

Jisuk Seon

Author Biographies

Jesse J. Helton, PhD, is an associate professor in the School of Social Work and College of Public Health and Social Justice at Saint Louis University. His research focuses on the overall safety and wellbeing of maltreated children, particularly those with disabilities and those struggling with food insecurity.

Erik J. Nelson, PhD, is an assistant professor in the Department of Epidemiology and Biostatistics, School of Public Health at Indiana University. His research combines epidemiologic, statistical, and geospatial modeling techniques to identify where and why health disparities occur so that appropriate interventions can be developed to help overcome them.

Brian B. Boutwell, PhD, is an associate professor of Legal Studies and Population Health Science at the University of Mississippi. He specializes in quantitative genetics, with a focus on environmental and psychological risk factors for antisocial and violent behavior. His research interests also include the intersection of health and disease risk factors with criminological outcomes.

Roger D. Lewis, PhD, is professor of Environmental and Occupational Health in the College for Public Health and Social Justice at Saint Louis University. His research focuses on environmental health interventions such as the development of chemi-fluorescent lead sensors, the removal of allergens from residential surfaces, and the use of mathematical models for targeting lead remediation.

Richard Rosenfeld, PhD, is professor and Chair of the Department of Criminology and Criminal Justice at the University of Missouri Saint Louis. His research interests include the social sources of criminal violence, crime statistics, and crime control policy. His recent publications address the social determinants of support for the death penalty, social capital and homicide, and racial profiling in police-citizen encounters.

Jisuk Seon, PhD, is an assistant professor in the Department of Social Welfare at Kyungnam University in South Korea. Her research focuses on the etiology of child maltreatment and the experiences of immigrant families who have been involved in child welfare. In addition, her research interests include social services and policy to prevent childhood adversities or alleviate its impact and improve developmental and health outcomes of vulnerable children.

References

Bellinger

D. C.

(2008). Neurological and behavioral consequences of childhood lead exposure. PLoS Medicine , 5(5), e115.

Bellinger

D. C.

(2017). Childhood lead exposure and adult outcomes. Jama , 317(12), 1219–1220.

Boutwell

B. B.

, Nelson

E. J.

, Emo

, Vaughn

M. G.

, Schootman

, Rosenfeld

, & Lewis

(2016). The intersection of aggregate-level lead exposure and crime. Environmental Research , 148, 79–85.

Cecil

K. M.

, Brubaker

C. J.

, Adler

C. M.

, Dietrich

K. N.

, Altaye

, Egelhoff

J. C.

, Wessel

, Elangovan

, Hornung

, Jarvis

, & Lanphear

B. P.

(2008). Decreased brain volume in adults with childhood lead exposure. PLoS Med , 5(5), e112.

Commission to Eliminate Child Abuse and Neglect Fatalities. (2016). Within our reach: A national strategy to eliminate child abuse and neglect fatalities . Government Printing Office.

Coulton

C. J.

, Richter

F. G. C.

, Korbin

, Crampton

, & Spilsbury

J. C.

(2018). Understanding trends in neighborhood child maltreatment rates: A three-wave panel study 1990–2010. Child Abuse & Neglect , 84, 170–181.

Dettlaff

A. J.

, Rivaux

S. L.

, Baumann

D. J.

, Fluke

J. D.

, Rycraft

J. R.

, & James

(2011). Disentangling substantiation: The influence of race, income, and risk on the substantiation decision in child welfare. Children and Youth Services Review , 33(9), 1630–1637.

Dewalt

F. G.

, Cox

D. C.

, O’Haver

, Salatino

, Holmes

, Ashley

P. J.

, Pinzer

E. A.

, Friedman

, Marker

, Viet

S. M.

, & Fraser

(2015). Prevalence of lead hazards and soil arsenic in U.S. housing. Journal of Environmental Health , 78(5), 22–29.

Evans

G. W.

, & Kantrowitz

(2002). Socioeconomic status and health: The potential role of environmental risk exposure. Annual Review of Public Health , 23(1), 303–331.

10.

Farrell

C. A.

, Fleegler

E. W.

, Monuteaux

M. C.

, Wilson

C. R.

, Christian

C. W.

, & Lee

L. K.

(2017). Community poverty and child abuse fatalities in the United States. Pediatrics , 139(5), e20161616.

11.

Frioux

, Wood

J. N.

, Fakeye

, Luan

, Localio

, & Rubin

D. M.

(2014). Longitudinal association of county-level economic indicators and child maltreatment incidents. Maternal and Child Health Journal , 18(9), 2202–2208.

12.

Herrenkohl

, Leeb

, & Higgins

(2016). The public health model of child maltreatment prevention. Trauma, Violence, & Abuse , 17(4), 363–365.

13.

Jaffee

S. R.

(2019). Lead exposure and child maltreatment as models for how to conceptualize early-in-life risk factors for violence. Infant Mental Health Journal , 40(1), 23–38.

14.

Maguire-Jack

, Font

S. A.

, & Dillard

(2020). Child protective services decision-making: The role of children’s race and county factors. American Journal of Orthopsychiatry , 90(1), 48–62.

15.

Manh

B. H.

, Clements

A. C.

, Thieu

N. Q.

, Hung

N. M.

, Hung

L. X.

, Hay

S. I.

, Tran

T. H.

, Wertheim

H. F.

, Snow

R. W.

, & Horby

(2011). Social and environmental determinants of malaria in space and time in Viet Nam. International Journal for Parasitology , 41(1), 109–116.

16.

Martins

, Simpson

, Lindgren

, & Rue

(2013). Bayesian computing with INLA: New features. Computational Statistics & Data Analysis , 67, 68–83.

17.

Morton

C. M.

, Simmel

, & Peterson

N. A.

(2014). Neighborhood alcohol outlet density and rates of child abuse and neglect: Moderating effects of access to substance abuse services. Child Abuse & Neglect , 38(5), 952–961.

18.

Nguyen

, & O’Dea

(2019, January 3). Interactive map: Median age of housing in St. Louis by neighborhood. St. Louis Today . https://www.stltoday.com/interactive-map-median-age-of-housing-in-st-louis-by/html_236c83e9-103e-5337-be58-49a9d2ec1d6a.html

19.

Patrick

(2006). Lead toxicity, a review of the literature. Part I: Exposure, Evaluation, and Treatment. Environmental Research , 11(1), 2–22.

20.

Rue

, & Martino

(2007). Approximate Bayesian inference for hierarchical Gaussian Markov random field models. Journal of Statistical Planning and Inference , 137(10), 3177–3192.

21.

Sampson

, Sharkey

, & Raudenbush

(2008). Durable effects of concentrated disadvantage on verbal ability among African-American children. Proceedings of the National Academy of Sciences , 105(3), 845–852.

22.

Stretesky

P. B.

, & Lynch

M. J.

(2004). The relationship between lead and crime. Journal of Health and Social Behavior , 45(2), 214–229.

23.

U.S. Department of Housing and Urban Development. (2011). American Healthy Homes Survey: Lead and arsenic findings . Retrieved November 13, 2019, from https://www.hud.gov/sites/documents/AHHS_REPORT.PDF

24.

United States Census Bureau. (2015). American Community Survey . http://www.census.gov/acs/www/

25.

Wheeler

, & Brown

M. J.

(2013). Blood lead levels in children aged 1–5 years—United States, 1999–2010. MMWR. Morbidity and Mortality Weekly Report , 62(13), 245–248.

26.

Winter

, & Sampson

(2017). From lead exposure in early childhood to adolescent health: A Chicago birth cohort. American Journal of Public Health , 107(9), 1496–1501.