ArticlePDF Available

Geospatial association between adverse birth outcomes and arsenic in groundwater in New Hampshire, USA

Authors:

Abstract and Figures

There is increasing evidence of the role of arsenic in the etiology of adverse human reproductive outcomes. Because drinking water can be a major source of arsenic to pregnant women, the effect of arsenic exposure through drinking water on human birth may be revealed by a geospatial association between arsenic concentration in groundwater and birth problems, particularly in a region where private wells substantially account for water supply, like New Hampshire, USA. We calculated town-level rates of preterm birth and term low birth weight (term LBW) for New Hampshire, by using data for 1997-2009 stratified by maternal age. We smoothed the rates by using a locally weighted averaging method to increase the statistical stability. The town-level groundwater arsenic probability values are from three GIS data layers generated by the US Geological Survey: probability of local groundwater arsenic concentration >1 µg/L, probability >5 µg/L, and probability >10 µg/L. We calculated Pearson's correlation coefficients (r) between the reproductive outcomes (preterm birth and term LBW) and the arsenic probability values, at both state and county levels. For preterm birth, younger mothers (maternal age <20) have a statewide r = 0.70 between the rates smoothed with a threshold = 2,000 births and the town mean arsenic level based on the data of probability >10 µg/L; for older mothers, r = 0.19 when the smoothing threshold = 3,500; a majority of county level r values are positive based on the arsenic data of probability >10 µg/L. For term LBW, younger mothers (maternal age <25) have a statewide r = 0.44 between the rates smoothed with a threshold = 3,500 and town minimum arsenic concentration based on the data of probability >1 µg/L; for older mothers, r = 0.14 when the rates are smoothed with a threshold = 1,000 births and also adjusted by town median household income in 1999, and the arsenic values are the town minimum based on probability >10 µg/L. At the county level for younger mothers, positive r values prevail, but for older mothers, it is a mix. For both birth problems, the several most populous counties-with 60-80 % of the state's population and clustering at the southwest corner of the state-are largely consistent in having a positive r across different smoothing thresholds. We found evident spatial associations between the two adverse human reproductive outcomes and groundwater arsenic in New Hampshire, USA. However, the degree of associations and their sensitivity to different representations of arsenic level are variable. Generally, preterm birth has a stronger spatial association with groundwater arsenic than term LBW, suggesting an inconsistency in the impact of arsenic on the two reproductive outcomes. For both outcomes, younger maternal age has stronger spatial associations with groundwater arsenic.
Content may be subject to copyright.
ORIGINAL PAPER
Geospatial association between adverse birth outcomes
and arsenic in groundwater in New Hampshire, USA
Xun Shi Joseph D. Ayotte Akikazu Onda Stephanie Miller
Judy Rees Diane Gilbert-Diamond Tracy Onega Jiang Gui
Margaret Karagas John Moeschler
Received: 11 July 2013 / Accepted: 23 September 2014
Springer Science+Business Media Dordrecht 2014
Abstract There is increasing evidence of the role of
arsenic in the etiology of adverse human reproductive
outcomes.Because drinking water can be a majorsource
of arsenic to pregnant women, the effect of arsenic
exposure through drinking water on human birth may be
revealed by a geospatial association between arsenic
concentration in groundwater and birth problems,
particularly in a region where private wells substantially
account for water supply, like New Hampshire, USA.
We calculated town-levelrates of preterm birth and term
low birth weight (term LBW) for New Hampshire,
by using data for 1997–2009 stratified by maternal age.
We smoothed the rates by using a locally weighted
averaging method to increase the statistical stability.
The town-level groundwater arsenic probability values
are from three GIS data layers generated by the US
Geological Survey: probability of local groundwater
arsenic concentration [1lg/L, probability [5lg/L,
and probability [10 lg/L. We calculated Pearson’s
correlation coefficients (r) between the reproductive
outcomes (preterm birth and term LBW) and the arsenic
probability values, at both state and county levels. For
preterm birth,younger mothers (maternal age\20) have
astatewider=0.70 between the rates smoothed with a
threshold =2,000 births and the town mean arsenic
level based on the data of probability [10 lg/L; for
older mothers, r=0.19 when the smoothing thresh-
old =3,500; a majority of county level rvalues are
X. Shi (&)A. Onda
Dartmouth College, Hanover, NH, USA
e-mail: xun.shi@dartmouth.edu
A. Onda
e-mail: akikazu.onda@gmail.com
J. D. Ayotte
NH - VT Office, New England Water Science Center,
U.S. Geological Survey, Concord, NH 03301, USA
e-mail: jayotte@usgs.gov
S. Miller J. Rees D. Gilbert-Diamond
T. Onega J. Gui M. Karagas J. Moeschler
The Geisel School of Medicine at Dartmouth, Hanover,
NH, USA
e-mail: Stephanie.D.Miller2@dartmouth.edu
J. Rees
e-mail: Judith.R.Rees@dartmouth.edu
D. Gilbert-Diamond
e-mail: Diane.Gilbert-Diamond@dartmouth.edu
T. Onega
e-mail: Tracy.L.Onega@dartmouth.edu
J. Gui
e-mail: Jiang.Gui@dartmouth.edu
M. Karagas
e-mail: Margaret.R.Karagas@dartmouth.edu
J. Moeschler
e-mail: John.Moeschler@dartmouth.edu
123
Environ Geochem Health
DOI 10.1007/s10653-014-9651-2
positive based on the arsenic data of probability[10 lg/L.
For term LBW, younger mothers (maternal age \25)
have a statewide r=0.44 between the rates smoothed
with a threshold =3,500 and town minimum arsenic
concentration based on the data of probability[1lg/L;
for older mothers, r=0.14 when the rates are
smoothed with a threshold =1,000 births and also
adjusted by town median household income in 1999,
and the arsenic values are the town minimum based on
probability[10 lg/L. At the county level for younger
mothers, positive rvalues prevail, but for older
mothers, it is a mix. For both birth problems, the
several most populous counties—with 60–80 % of the
state’s population and clustering at the southwest
corner of the state—are largely consistent in having a
positive racross different smoothing thresholds. We
found evident spatial associations between the two
adverse human reproductive outcomes and groundwa-
ter arsenic in New Hampshire, USA. However, the
degree of associations and their sensitivity to different
representations of arsenic level are variable. Generally,
preterm birth has a stronger spatial association with
groundwater arsenic than term LBW, suggesting an
inconsistency in the impact of arsenic on the two
reproductive outcomes. For both outcomes, younger
maternal age has stronger spatial associations with
groundwater arsenic.
Keywords Preterm birth Low birth weight
Arsenic Groundwater Locally weighted averaging
smoothing New Hampshire
Introduction
Infant mortality can be impacted by adverse repro-
ductive outcomes, and these outcomes can be sensitive
to many environmental influences (Wilcox 2001;
WHO 2002; Stallones et al. 1992; Braud et al.
2011). Geospatial analysis has been used to study
variation in the occurrence of the adverse outcomes
(Stallones et al. 1992; Braud et al. 2011; Talbot et al.
2000; Reader 2001; Ozdenerol et al. 2005; Tu et al.
2011) and its possible association with environmental
factors, including point-source pollution, such as
hazardous waste sites and toxic release sites (Stallones
et al. 1992; Braud et al. 2011), pesticide exposure
(Xiang et al. 2000), and air pollution (Lin et al. 2004;
Slama et al. 2007; Kashima et al. 2011).
In the past 15 years, epidemiologists have been
particularly interested in the role of arsenic in the
etiology of adverse human reproductive outcomes,
especially arsenic exposure from drinking water
sourced from groundwater. A geographic concentra-
tion of the research is Bangladesh (Ahmad et al. 2001;
Milton et al. 2005; Vahter et al. 2006; Kwok et al.
2006; Huyck et al. 2007; Rahman et al. 2009; Tofail
et al. 2009; Sohel et al. 2010; Kippler et al. 2012). At
the individual level, some studies in Bangladesh
provide evidence that arsenic exposure is associated
with adverse reproductive outcomes. For example, one
study in the MATLAB region identified a significant
association in 1,578 mother–infant pairs over the
range of urinary arsenic concentrations of \100 lg/L
[but not over the entire range measured (6–978 lg/L)].
Over the lower range, a 1.68-g reduction in birth
weight was seen for every 1 lg/L increase in urinary
arsenic concentration (Rahman et al. 2009). Another
study in the Sirajdikhan region analyzed hair, toenail,
and drinking water samples collected from 52
pregnant women at multiple time points during
pregnancy and from their newborns after birth and
suggests that maternal arsenic exposure early in
pregnancy negatively affects newborn birth weight
(Huyck et al. 2007). However, one study that has
examined 2,006 pregnant women from the Faridpur
Sadar, MATLAB, and Shahrasti regions who had
been chronically exposed to a range of naturally
occurring concentrations of arsenic in drinking water
only finds small but statistically significant association
between arsenic exposure and birth defects and did not
see such an association in some other outcomes (Kwok
et al. 2006). At the ecological level, a study that
examined fetal loss and infant death in the MATLAB
region performed geospatial clustering analysis on
both reproductive outcomes and arsenic concentra-
tion. It finds that the spatial patterns of arsenic
concentrations in tube-well water are linked with the
adverse pregnancy outcome clusters (Sohel et al.
2010). Besides Bangladesh, an individual-level study
in India finds that exposure to high concentrations of
arsenic (200 lg/L) during pregnancy was associated
with a sixfold increased risk of stillbirth after adjust-
ment for potential confounders, but finds no associa-
tion between arsenic exposure and spontaneous
abortion or overall infant mortality (von Ehrenstein
et al. 2006). Ecologic studies in Taiwan and Chile
indicate that arsenic endemic areas with drinking
Environ Geochem Health
123
water contamination have significantly lower average
birth weights compared to non-endemic regions (Yang
et al. 2003; Hopenhayn et al. 2003); however, an
ecological study in Mongolia did not support an
association (Myers et al. 2010). The mechanism
through which arsenic influences birth weight is not
clear; one of many possible explanations is arsenic-
induced impaired glucose tolerance during pregnancy
(Ettinger et al. 2009; Andra et al. 2013).
This paper presents a geospatial analysis of asso-
ciations between groundwater arsenic concentration
and two adverse reproductive outcomes, preterm birth
and term LBW, in New Hampshire, USA. The
novelties of this study include: (1) To our knowledge,
geographically this is one of the earliest studies of its
kind particularly about a US cohort; (2) this might be
one of few studies exploring effect of low concentra-
tion of groundwater arsenic on reproductive outcomes;
and (3) methodologically, different from most eco-
logical studies of its kind that compare two or a few
selected regions, we compare the continuous geo-
graphic distributions of arsenic and adverse outcomes.
Our conceptual model holds that, if indeed the arsenic
in daily drinking water has an effect on human birth,
this effect may be revealed by a correlation between
the spatial variability of arsenic concentration in
groundwater and the spatial variability of adverse
reproductive outcomes and that this may be particu-
larly detectable in a region where private wells
substantially account for the water supply. In New
Hampshire, about 40 % of the population uses private
wells as a primary source for drinking water, which is
a reason for us to choose it as our study area.
Data
Our choice of preterm birth (gestational period
\37 weeks) and LBW (birth weight \2,500 g, e.g.,
Wilcox 2001) for this study is first based on the
availability of data, and is also following suggestions
by Wilcox (2001), who provides a powerful discussion
of the importance of separating preterm birth and term
birth in epidemiological studies. He states that ‘‘(a)n
exposure that affects fetal growth does not necessarily
affect the risk of preterm delivery,’’ and ‘‘(c)onverse-
ly, a factor that increases the risk of preterm delivery
would not necessarily change the average weight of
babies delivered at term.’’ He then recommends that
when the data of gestational age are available, the
preterm birth rate be selected for analysis. For term
births, Wilcox recommends to use mean birth weight
and simultaneously consider standard deviation (SD),
which has been adopted by most studies that are
comparing two or a few regions (e.g., Kwok et al.
2006; Yang et al. 2003; Hopenhayn et al. 2003; Myers
et al. 2010). However, simultaneously considering
mean and SD are difficult when working with many
areal units. In such a situation, the LBW rate is
convenient and actually to some extent characterizes
both mean and SD. Therefore, we choose to use the
rate of term LBW (i.e., gestational period C37 weeks
AND birth weight\2,500 g) as another measurement
of adverse reproductive outcome in this study.
Birth data and rate calculation
We obtained birth data from New Hampshire birth
certificates for 1997–2009 (N=187,851) provided by
New Hampshire Department of Health and Human
Services (NH DHHS). Each record in the dataset is for
an infant and contains information about (1) the infant,
including birth date, gestational age, sex, birth weight,
plurality, and birth order; and (2) the mother, including
age, residential town, and zip code at delivery.
Prior to the analysis, we removed those records of
mothers who were not residents of New Hampshire
towns, which account for about 1 % of the original
records. We then removed those records with a
plurality value [1 (i.e., twins and triplets), which
account for about 3 % of all records. We also removed
those records with apparent invalid or missing values
on gestational period, maternal age, and birth weight,
which account for less than 1 % of all records. After
these processes, a total of 177,995 records remained
and were used in the following analyses.
Rate of preterm birth
Among the 177,995 usable records, 12,501 have a
gestational period \37 weeks and were identified as
preterm births. We stratified the data into detailed
categories of maternal age and calculated the preterm
birth rate for each category (the upper part of Table 1).
The calculation reveals a step at maternal age =20, so
we grouped the detailed categories into two larger
categories: maternal age \20 and maternal age C20
(the lower part of Table 1). Although the categories of
Environ Geochem Health
123
maternal age C40 have greater rates, their relatively
small counts for both preterm births and all births may
lead to statistical instability and therefore were
grouped into the category of maternal age C20. We
performed the following analyses separately for the
two strata to address the influence of maternal age.
Rate of term LBW
The number of usable records of full-term births is
164,335. From these records we identified 2,651 LBW
cases (i.e., gestational period C37 weeks AND birth
weight \2,500 g). Similar to the process with the
preterm birth data, we stratified the data into detailed
categories of maternal age and calculated the term-
LBW rate for each (the upper part of Table 2). The
calculation reveals a clear step at maternal age =24,
so we grouped the detailed categories into two larger
categories: maternal age \25 and maternal age C25
and performed the following analyses separately for
these two strata. Similar to the preterm birth data, we
grouped the categories of maternal age C40 into the
category of maternal age C25, due to their relatively
small counts.
Population and socioeconomic data
The population data used in the rate calculations
described above are from the US Census 2010 data
(http://www.census.gov/). In this study, however, we
did not include the factor of race/ethnicity, because the
birth certificates do not contain such information.
According to the Census 2010 data, nonwhite females
in NH account for 7 % of the female population within
the age range 15–49. Thus, we assumed that any race/
ethnicity effect is negligible. The risk of birth prob-
lems may also be affected by socioeconomic status.
We collected town-level income data from the New
Hampshire Office of Energy and Planning (www.nh.
gov/oep).
Arsenic data
The exposure data used in this study are the modeled
probabilities of arsenic occurrence (at thresholds of 1,
5, and 10 lg/L) in private wells that tap groundwater
from bedrock aquifers in New Hampshire. This
probability of finding arsenic at a location above a
given threshold was estimated by using multivariate
logistic regression models (‘‘probability models’’)
developed for New Hampshire (Ayotte et al. 2012).
The probability models were developed by using
measurements of arsenic from public and private
wells as the dependent (or predicted) variable and
by using a variety of geologic, geochemical, hydro-
logic, and anthropogenic data as the independent
variable (predictor; Ayotte et al. 2006,2011,2012;
Flanagan and Ayotte 2011).
Probability models for predicting arsenic concen-
trations that were greater than or equal to 1, 5, and
10 lg/L in groundwater from bedrock wells were
developed in order to produce individual threshold-
level probability maps. These three thresholds were
Table 1 Preterm birth ratio by maternal age category, New
Hampshire, 1997–2009
Preterm births All births Ratio
Maternal age (years)
\20 1,034 11,792 0.0877
20–24 2,487 34,272 0.0726
25–29 3,317 49,665 0.0668
30–34 3,400 51,691 0.0658
35–39 1,783 25,368 0.0703
40–44 456 4,975 0.0917
C45 24 232 0.1034
Two-category stratification
\20 1,034 11,792 0.0877
C20 11,467 166,203 0.0690
Table 2 Term low birth weight (LBW) ratio by maternal age
category, New Hampshire, 1997–2009
Term LBW infants All full-term births Ratio
Age
\20 274 10,569 0.0259
20–24 685 31,207 0.0220
25–29 665 47,762 0.0139
30–34 615 47,225 0.0130
35–39 325 22,990 0.0141
40–44 80 4,404 0.0182
C45 7 198 0.0354
Two-category stratification
\25 959 41,776 0.0230
C25 1,692 122,579 0.0138
Environ Geochem Health
123
chosen because they represent common arsenic
reporting levels in groundwater in the State and
because they are considered to be possibly relevant
concentrations for exposure estimation in terms of
potential human health outcomes. Also, the current
USEPA maximum contaminant level, the standard for
safe drinking water with which public water supplies
in the USA must comply, is 10 lg/L. The multivariate
logistic regression techniques used to generate the
probability estimates are well suited for modeling
censored dependent-variable data—data reported as
‘less than’’ some laboratory reporting limit—because
data that are below reporting limits can be used
directly without having to modify or substitute values
(Helsel and Hirsch 1992; Helsel 2005; Hosmer and
Lemeshow 2000). The well-water arsenic concentra-
tion data (dependent data) include censored data that
were reported as below laboratory reporting levels
(LRLs). The model takes the form:
P½y¼1jx¼ eðb0þb1x1þb2x2þþbkxkÞ
1þeðb0þb1x1þb2x2þþbkxkÞð1Þ
where Pis the probability of observing the event, y is
an indicator (threshold) variable (‘y=1’’ denoting an
event or measurement greater than or equal to a
specific value (such as 1, 5, and 10 lg/L), and
‘‘ y=0’’ denoting a nonevent or measurement less
than a specific threshold), where x
1
,x
2
,,x
k
are
explanatory or independent variables, and where b
0
,
b
1
,,b
k
are unknown parameters (coefficients) to be
estimated. The exponential of a parameter, exp (b
i
),
specifies the proportional increase in the odds of an
arsenic concentration being above the modeled
threshold per unit increase in the explanatory variable.
Threshold values of 1, 5, and 10 lg/L were modeled to
identify areas of the State where the probabilities are
high for finding low-level (greater than or equal to
1lg/L) and high-level (greater than or equal to 10 lg/L)
arsenic contamination in groundwater. Standard
model testing and performance metrics were evaluated
and are described in detail elsewhere (Ayotte et al.
2012).
We tested data representing concentrations of
arsenic in 1,715 wells (dependent variable) to develop
the models, along with more than 250 independent
variables, all developed in a geographic information
system (GIS), and representing geologic, hydrologic,
demographic, and land-use and land-cover features.
The final models were dominated by geologic and
geochemistry variables but also included variables
such as population density, precipitation, groundwater
recharge, land use, and proximity to waste sites
(Ayotte et al. 2012).
The probability of having arsenic concentrations
exceeding 1, 5, and 10 lg/L in groundwater was
variable across the state. Generally, high probabilities
of arsenic greater than 5 or 10 lg/L were limited to
southeastern New Hampshire. However, high proba-
bilities that groundwater from bedrock aquifers would
exceed 1 lg/L were widespread across New Hamp-
shire. In fact, nearly half of the State was classified as
having at least a 50 % chance of having arsenic greater
than or equal to 1 lg/L. High probabilities of arsenic
greater than or equal to 5 and 10 lg/L were predicted
primarily in the southeastern counties of Merrimack,
Strafford, Hillsborough, and Rockingham—the coun-
ties that are home to about 75 % of the State’s
population.
The original USGS arsenic data are in the format of
GIS raster layers, with cell size =30 m. To match the
LBW data at the town level, the cell-level data were
aggregated to town level by using the Zonal Statistics
tool of ArcGIS*, i.e., the average of the values of all
the cells falling into a town is used as the represen-
tative value of that town (Fig. 1).
Methods
Rate smoothing
To statistically stabilize the rates, we applied a locally
weighted average smoothing to the original rates.
Locally weighted average methods smooth the rate of
an areal unit (in our case, a town) by averaging all the
rates of the units in its neighborhood, during which
each rate is weighted by its associated background
value (in our case, number of births; Kafadar 1994;
Waller and Gotway 2004). The specific method we
implemented was proposed and justified by Shi et al.
(2007), and it is different from conventional locally
weighted averaging methods. This method (1)
employs a user-specified background value rather
than a constant geographic distance to define the
neighborhood for smoothing, which makes the statis-
tical stability explicit and controllable; (2) generates
the neighborhood by creating a buffer around the
polygon, rather than about the centroid of the polygon,
Environ Geochem Health
123
which takes into account the size and shape of the
polygon; and (3) if the neighborhood encloses only
part of a polygon, the weight of that polygon will be
proportionally determined, allowing a more accurate
estimation of contribution of each polygon than an in-
or-out strategy.
To address the subjectivity in determining the
threshold for defining the neighborhood, Shi et al.
proposed a strategy that calculates a series of
smoothing results using different thresholds. For each
of these results, the overall variance of the smoothed
rates is calculated and plotted (Shi et al. 2007). It is
expected that the variance values become stable as the
threshold increases, and the turning point on the plot
where the variance value starts to level out is
considered as an indication of the optimal threshold.
We used this strategy in the current study to identify
optimal thresholds.
Correlation calculation
To detect the spatial association between birth prob-
lems and groundwater arsenic, we calculated a Pear-
son’s correlation coefficient (r) between the town-
level preterm birth rates and the groundwater arsenic
levels, and between the term-LBW rates and the
0.00 - 0.10
0.10 - 0.20
0.20 - 0.30
0.30 - 0.40
0.40 - 0.50
0.50 - 0.60
0.60 - 0.80
0.80 - 0.95
County
Town
Coos
Grafton Carroll
Merrimack
Cheshire
Hillsborough
Sullivan
Rockingham
Belknap
Strafford
0.99966
0
Fig. 1 The USGS modeled probability of groundwater arsenic
occurrence in New Hampshire represented as the probability of
arsenic concentration exceeding a certain level. The left map is
the original USGS raster data showing the probability of arsenic
concentration [1lg/L; the right map is the town-level mean
values calculated from the left map by using the zonal statistics
tool of ArcGIS*
Environ Geochem Health
123
groundwater arsenic levels. For both preterm births
and term-LBW, the calculation was performed on the
original rates and a series of smoothed rates with
different thresholds. The calculation was performed
for the different maternal age strata separately.
For each stratum, we started the smoothing with a
threshold =100 births, i.e., the buffer around each
town polygon was expanded until it enclosed 100
births. With this threshold, those towns with more than
100 births during the study period would not be
smoothed. We kept increasing the threshold, by using
100 as the increment, to get smoother results. Figure 2
shows the maps of the original rates and the smoothed
rates with a threshold =2,200 side by side for a visual
comparison. These maps are for younger mothers in
the term-LBW analysis.
With the map of smoothed rates from a threshold,
we calculated rbetween the rates and arsenic level,
as well as the variance of rates across the entire
state. As an example, Fig. 3shows the results of
these calculations for younger mothers in the term-
LBW analysis. Figure 2indicates that for younger
mothers, the correlation between the term-LBW rate
and groundwater arsenic almost monotonically
increases after the only major drop associated with
the starting threshold (100 births). With a threshold
of 3,500 births, rreaches 0.3. We stopped at 3,500
to avoid over-smoothing. In fact, the variance starts
No births
0.0 - 0.5
0.5 - 1.0
1.0 - 1.5
1.5 - 2.0
2.0 - 2.5
2.5 - 3.0
3.0 - 3.5
3.5 - 4.0
4.0 - 6.0
6.0 - 10.0
10.0 - 16.7
Coos
Grafton Carroll
Merrimack
Cheshire Hillsborough
Sullivan
Rockingham
Belknap
Strafford
No births
1.5 - 2.0
2.0 - 2.5
2.5 - 3.0
3.0 - 3.5
Coos
Grafton Carroll
Merrimack
Cheshire Hillsborough
Sullivan
Rockingham
Belknap
Strafford
Fig. 2 New Hampshire town-level LBW rates for maternal age \25: the left map displays the original rates; the right map displays the
smoothed rates with smoothing threshold =2,200 births
Environ Geochem Health
123
to stabilize when the threshold =2,000 and
becomes very small after 2,500.
To eliminate possible impacts of population and
income on the risk of birth problems, following Ayotte
et al. (2006), we also applied linear regression with the
disease rate as the dependent variable and population
or income as the independent variable. If the rate was
related to population or income, we calculated resid-
uals of the dependent variable and then calculated
correlations between the residuals and the arsenic
values.
To explore the local variation of the spatial
association, we also calculated the correlation coeffi-
cient for each of the 10 counties of New Hampshire.
Results
Preterm birth
For the stratum of maternal age \20 (Table 3),
statewide, the unsmoothed town-level preterm birth
rates have slightly negative rvalues for all three
arsenic data layers, as well as for the town household
median income value. However, the smoothing
changes this situation. Even with a relatively small
smoothing threshold of 500 births, rvalues all become
positive with considerable magnitudes. Generally, the
higher the degree of smoothing, the higher the rvalue.
The largest rvalue, 0.70, occurs between the town
mean arsenic level based on the data of probability
[10 lg/L and the preterm birth rate smoothed with a
threshold =2,000 births (the largest threshold used in
this analysis). Among the three arsenic data layers, the
ones of probability [5 and [10 lg/L have stronger
positive associations with the preterm birth rate than
the one of probability [1lg/L. Among the town
minimum, maximum, and mean for the arsenic
probability, generally the mean has the highest
rvalue, whereas the minimum has the lowest.
Unexpectedly, the smoothed rates have fairly consid-
erable positive correlations with the town median
household income, i.e., higher rates tend to be
associated with higher income values. The adjustment
by income consistently lowers down the rvalues,
indicating that the income and arsenic may have an
association to a certain extent. However, even after the
adjustment, the positive association between preterm
birth and groundwater arsenic for this group of
mothers in New Hampshire is still considerable.
At the county level, generally the rvalues progres-
sively become more positive as the arsenic threshold
increases, from probability [1, [5, to [10 lg/L
(Fig. 4). For the data of probability [10 lg/L, a
majority of rvalues are positive. This progressive
variation is most distinct for the five most populous
counties (in terms of population density) in New
Hampshire, including Hillsborough, Rockingham,
Stafford, Merrimack, and Belknap (the five left-most
counties in Fig. 4), which account for 78 % of the
state’s total population, and geographically cluster at
the southwest corner of the state that is close to
the greater Boston area. For the data of probability
[1lg/L, four of these five counties have dominantly
Fig. 3 Variance of New
Hampshire town-level LBW
rates for maternal age \25
and correlation between the
rates and groundwater
arsenic probability values
against smoothing threshold
Environ Geochem Health
123
negative rvalues, but for the data of probability[10 lg/L,
most rvalues of these counties are positive. Another
noteworthy finding is that in most cases, the smoothing
‘helps’’ increase positiveness.
For the stratum of maternal age C20, the rvalues
are much smaller compared with their counterparts of
the younger mothers (Table 4). While it is hard to
claim any significant association based on these
rvalues, it seems, however, that the general pattern
of them is similar to that of the younger mothers. The
probability [5lg/L and probability [10 lg/L have
stronger positive associations with the preterm birth
rate than the probability [1lg/L. In fact, all rvalues
based on the 5 and 10 lg/L data are consistently
positive, although small. Again, the smoothing gener-
ally helps increase positiveness for r. The largest
rvalue, 0.19, occurs between the town-level mean for
probability [10 lg/L and the preterm birth rate
smoothed with the largest threshold (3,500 births).
However, in this stratum, the preterm birth rate does
not appear to have considerable associations with
household median income, and therefore, we did not
calculate the income-adjusted r.
At the county level, the progressive change of
rvalues along with the three arsenic data layers is still
obvious (Fig. 5). For the data of probability[10 lg/L,
a majority of rvalues are positive. This time, the role
of smoothing is controversial. For example, for
counties of Cheshire and Carroll, smoothing makes
rvalues stably and increasingly positive, while for
counties of Rockingham, Strafford, and Belknap,
smoothing reduces the positiveness.
Term low birth weight
The results of term LBW are generally weaker and less
consistent than those of the preterm birth. For the
stratum of maternal age\25 years, statewide positive
rvalues are dominant across all tests performed
(Table 5), but some patterns are different from those
of the preterm birth. First, the rank of the three arsenic
data layers is reversed and this time the data of
probability[1lg/L have stronger association with the
rates than the other two. Second, among the town
minimum, maximum, and mean, the minimum con-
sistently has the highest positive rvalues than the other
Table 3 Correlation coefficient (r) between town-level preterm birth rate for maternal age \20 and groundwater arsenic occur-
rence in New Hampshire, 1997–2009
Original Rate Smoothed_500 Smoothed_1000 Smoothed_1500 Smoothed_2000
Prob1
Town_Min -0.09 0.25 (0.14) 0.24 (0.14) 0.28 (0.17) 0.31 (0.15)
Town_Max -0.01 0.37 (0.30) 0.29 (0.22) 0.36 (0.29) 0.43 (0.32)
Town_Mean -0.08 0.26 (0.15) 0.23 (0.13) 0.30 (0.20) 0.39 (0.23)
Prob5
Town_Min -0.06 0.47 (0.37) 0.44 (0.35) 0.50 (0.42) 0.56 (0.42)
Town_Max -0.13 0.39 (0.26) 0.48 (0.37) 0.57 (0.46) 0.64 (0.47)
Town_Mean -0.07 0.46 (0.33) 0.48 (0.36) 0.57 (0.45) 0.65 (0.47)
Prob10
Town_Min -0.06 0.47 (0.40) 0.46 (0.38) 0.47 (0.40) 0.49 (0.38)
Town_Max -0.07 0.43 (0.31) 0.41 (0.30) 0.51 (0.41) 0.61 (0.45)
Town_Mean -0.04 0.57 (0.46) 0.56 (0.47) 0.63 (0.54) 0.70 (0.55)
Median household income -0.06 0.34 0.30 0.39 0.47
Prob1,Prob5, and Prob10 denote three GIS data layers of modeled groundwater arsenic occurrence, representing probability of
arsenic[1, 5, and 10 lg/L, respectively; Town_Min,Town_Max, and Town_Mean denote town-level minimum, maximum, and mean
for the modeled arsenic probability values, respectively; Smoothed_500 etc. denote the smoothed rates; e.g., smoothed_500 denotes
the town-level preterm birth rate smoothed from the Original Rate using a threshold that the neighborhood of smoothing must enclose
at least 500 births; the value inside the parentheses is radjusted by town-level income (i.e., rbetween the arsenic probability value
and the residual to the preterm birth–income linear regression); the bottom line contains the rvalues between the preterm birth rate
and the town median household income in 1999; because the original rate has very weak correlation with the income value, the
income-adjusted rfor the original rate is not calculated
Environ Geochem Health
123
two. The smoothing still helps increase the positive-
ness. The rvalues between the rates and income are
fairly small, and therefore, we did not calculate the
income-adjusted r.
Figure 6shows the county-specific rvalues
between the town minimum arsenic probability value
and the term-LBW rate. While statewide more
positive ris associated with the arsenic probability
Fig. 4 County-specific correlation coefficient (r) between
town-level preterm birth rate for maternal age \20 and
groundwater arsenic probability values in New Hampshire,
1997–2009. Notes Prob1,Prob5, and Prob10 denote three GIS
data layers of modeled groundwater arsenic occurrence,
representing probability of arsenic [1, 5, and 10 lg/L,
respectively; Mean_Ori denotes the rbetween the town-level
mean probability of arsenic occurrence and the original preterm
birth rate; Mean_500 denotes the rbetween the town-level mean
probability of arsenic occurrence and the preterm birth rate
smoothed with threshold =500 births; and so on
Environ Geochem Health
123
value of 1 lg/L, at the county level, the dominance of
positive rvalues is more obvious with the probability
value of 10 lg/L. It seems that the inverted results for
Merrimack County with the probabilities of 1 and
10 lg/L have caused this controversy. The geographic
pattern largely maintains: The three most populous
counties (in terms of population density), including
Hillsborough, Rockingham, and Strafford, accounting
for 62 % of the states’ population and clustering near
the greater Boston area, generally have positive
rvalues across the three arsenic data layers. The
smoothing, again, in most cases helps increase the
positiveness.
For the stratum of maternal age C25 years, state-
wide ris dominated by negative values, a few are
fairly considerable (e.g., r=-0.36 for the town
maximum of probability [10 lg/L and the rates
smoothed with a threshold =2,500 births), although
most are very small (Table 6). Along the line from 1,
5, to 10 lg/L, the negativeness generally increases,
especially for the town maximum and mean. The
effect of smoothing does not have an obvious pattern.
The rates, however, have non-negligible negative
associations with town median household income
(higher rates tend to be associated with lower
incomes). The adjustment by the income consistently
reduces the negativeness of the rvalues, indicating
that income may be a confounder that is overshadow-
ing the effect of groundwater arsenic.
The county-specific correlations for the stratum of
maternal age C25 years are weak and inconsistent
overall. Figure 7shows the county-specific rvalues
between the income-adjusted term-LBW rates and the
town minimum arsenic probability. The rvalues are
generally small (no matter positive or negative),
compared with the stratum of younger mothers.
Negative rvalues prevail, in terms of the number of
counties, with the data of 1 and 10 lg/L; with the data
of 5 lg/L, it is a mix. The most populous county,
Hillsborough, maintains to be generally positive
across the three arsenic data layers.
Discussion
We found evident spatial associations between two
adverse human reproductive outcomes, preterm birth
and term LBW, and groundwater arsenic in New
Table 4 Correlation coefficient (r) between town-level preterm birth rate for maternal age C20 and groundwater arsenic occur-
rence in New Hampshire, 1997–2009
Original
Rate
Smoothed
500
Smoothed
1000
Smoothed
1500
Smoothed
2000
Smoothed
2500
Smoothed
3000
Smoothed
3500
Prob1
Town_Min 0.00 -0.04 -0.06 -0.06 -0.05 -0.05 -0.05 -0.05
Town_Max -0.05 -0.02 0.01 0.04 0.07 0.09 0.11 0.13
Town_Mean -0.10 -0.10 -0.09 -0.06 -0.02 0.01 0.03 0.04
Prob5
Town_Min 0.04 0.10 0.09 0.10 0.11 0.12 0.13 0.14
Town_Max 0.01 0.05 0.08 0.10 0.13 0.14 0.15 0.18
Town_Mean 0.01 0.06 0.08 0.10 0.13 0.14 0.16 0.18
Prob10
Town_Min 0.04 0.06 0.06 0.06 0.07 0.07 0.07 0.07
Town_Max 0.03 0.00 0.01 0.02 0.04 0.05 0.07 0.09
Town_Mean 0.05 0.08 0.10 0.11 0.13 0.15 0.17 0.19
Median household
income
-0.12 -0.15 -0.13 -0.11 -0.08 -0.06 -0.04 -0.03
Prob1,Prob5, and Prob10 denote three GIS data layers of modeled groundwater arsenic occurrence, representing probability of arsenic
[1, 5, and 10 lg/L, respectively; Town_Min,Town_Max, and Town_Mean denote town-level minimum, maximum, and mean for the
modeled arsenic probability values, respectively; Smoothed_500 etc. denote the smoothed rates; e.g., smoothed_500 denotes the town-
level preterm birth rate smoothed from the Original Rate using a threshold that the neighborhood of smoothing must enclose at least 500
births; the bottom line contains the rvalues between the preterm birth rate and the town median household income in 1999; because the
preterm birth rate has very weak correlation with the income value, we did not calculate the income-adjusted rvalues
Environ Geochem Health
123
Environ Geochem Health
123
Hampshire, USA. However, the properties of these
associations vary, in terms of degree of association and
sensitivity to different representations of arsenic level.
Generally, preterm birth has a stronger spatial asso-
ciation with groundwater arsenic than with the term
LBW, suggesting an inconsistency in the impact of
arsenic on the two reproductive outcomes, and con-
firming the necessity to distinguish preterm births and
term births in this kind of analysis. For both
reproductive outcomes, younger maternal age has
stronger spatial associations with groundwater
arsenic. In particular, for term LBW, while a positive
spatial association between LBW and arsenic level is
observed for maternal \25, the association is unclear
for maternal age C25. However, an initial exploration
with town median household income suggests that for
the stratum of maternal age C25, the effect of
groundwater arsenic might have been shadowed by
socioeconomic or other factors.
In this study, we treated town median household
income as a confounding factor. However, the asso-
ciations between the reproductive outcome and house-
hold income are fairly variable across outcomes and
maternal ages. The preterm birth with maternal age
\20 has a stronger and unexpected positive associa-
tion with the household income, and the adjustment by
household income consistently reduces the positive-
ness in the association between the outcome and the
arsenic, suggesting that the spatial distribution of
household income may co-vary with groundwater
arsenic to some extent in NH. The preterm birth with
maternal age C20 has a slight but negative association
with the household income, as well as a much weaker
positive association with arsenic, which can be
interpreted as that the negative effect of income and
the positive effect of arsenic have cancelled each other
bFig. 5 County-specific correlation coefficient (r) between
town-level preterm birth rate for maternal age C20 and
groundwater arsenic probability values in New Hampshire,
1997–2009. Notes Prob1,Prob5, and Prob10 denote three GIS
data layers of modeled groundwater arsenic occurrence,
representing probability of arsenic [1, 5, and 10 lg/L,
respectively; Mean_Ori denotes the rbetween the town-level
mean probability of arsenic occurrence and the original preterm
birth rate; Mean_500 denotes the rbetween the town-level mean
probability of arsenic occurrence and the preterm birth rate
smoothed with threshold =500 births; and so on
Table 5 Correlation coefficient (r) between town-level term-LBW rate for maternal age\25 and groundwater arsenic occurrence in
New Hampshire, 1997–2009
Original
Rate
Smoothed
500
Smoothed
1000
Smoothed
1500
Smoothed
2000
Smoothed
2500
Smoothed
3000
Smoothed
3500
Prob1
Town_Min 0.17 0.19 0.25 0.30 0.35 0.37 0.40 0.44
Town_Max 0.03 0.00 -0.02 -0.01 0.00 0.01 0.01 0.03
Town_Mean 0.16 0.16 0.19 0.21 0.25 0.26 0.26 0.30
Prob5
Town_Min 0.21 0.15 0.18 0.23 0.27 0.28 0.29 0.34
Town_Max 0.10 0.02 0.02 0.04 0.08 0.08 0.09 0.14
Town_Mean 0.18 0.10 0.12 0.16 0.21 0.21 0.22 0.27
Prob10
Town_Min 0.16 0.23 0.25 0.30 0.34 0.36 0.38 0.43
Town_Max 0.04 -0.12 -0.18 -0.18 -0.16 -0.16 -0.15 -0.09
Town_Mean 0.13 0.02 -0.02 0.00 0.04 0.04 0.05 0.11
Median household
income
-0.02 -0.03 0.02 0.05 0.10 0.12 0.12 0.18
Prob1,Prob5, and Prob10 denote three GIS data layers of modeled groundwater arsenic occurrence, representing probability of
arsenic[1, 5, and 10 lg/L, respectively; Town_Min,Town_Max, and Town_Mean denote town-level minimum, maximum, and mean
for the modeled arsenic probability values, respectively; Smoothed_500 etc. denote the smoothed rates; e.g., smoothed_500 denotes
the town-level preterm birth rate smoothed from the Original Rate using a threshold that the neighborhood of smoothing must enclose
at least 500 births; the bottom line contains the rvalues between the preterm birth rate and the town median household income in
1999; because the term-LBW rate has very weak correlation with the income value, we did not calculate the income-adjusted rvalues
Environ Geochem Health
123
Fig. 6 County-specific correlation coefficient (r) between
town-level term-LBW rate for maternal age \25 and ground-
water arsenic probability values in New Hampshire. Notes
Prob1,Prob5, and Prob10 denote three GIS data layers of
modeled groundwater arsenic occurrence, representing proba-
bility of arsenic[1, 5, and 10, respectively; Min_Ori denotes the
rbetween the town-level minimum probability of arsenic
occurrence and the original term-LBW rate; Minimum_500
denotes the rbetween the town-level mean probability of arsenic
occurrence and the term-LBW rate smoothed with thresh-
old =500 births; and so on
Environ Geochem Health
123
Table 6 Correlation coefficient (r) between town-level term-LBW rate for maternal age C25 and groundwater arsenic occurrence in New Hampshire, 1997–2009
Original Rate Smoothed 500 Smoothed 1000 Smoothed 1500 Smoothed 2000 Smoothed 2500 Smoothed 3000 Smoothed 3500
Prob1
Town_Min -0.03 (0.04) -0.02 (0.08) -0.01 (0.12) -0.03 (0.12) -0.03 (0.11) -0.02 (0.11) -0.01 (0.11) -0.01 (0.11)
Town_Max -0.15 (-0.12) -0.05 (-0.01) -0.05 (0.00) -0.06 (0.00) -0.05 (0.01) -0.03 (0.02) -0.03 (0.02) -0.03 (0.02)
Town_Mean -0.06 (0.02) -0.07 (0.02) -0.10 (0.03) -0.12 (0.01) -0.11 (0.02) -0.09 (0.03) -0.08 (0.03) -0.08 (0.03)
Prob5
Town_Min -0.03 (0.05) 0.02 (0.13) 0.02 (0.16) -0.02 (0.13) -0.05 (0.08) -0.05 (0.08) -0.05 (0.08) -0.05 (0.08)
Town_Max -0.06 (0.02) -0.11 (0.00) -0.12 (0.03) -0.16 (-0.01) -0.20 (-0.06) -0.18 (-0.04) -0.14 (-0.01) -0.13 (0.00)
Town_Mean -0.06 (0.04) -0.08 (0.04) -0.08 (0.08) -0.11 (0.06) -0.14 (0.03) -0.12 (0.04) -0.10 (0.05) -0.09 (0.06)
Prob10
Town_Min -0.01 (0.05) 0.03 (0.12) 0.03 (0.14) 0.00 (0.12) -0.03 (0.08) -0.03 (0.08) -0.03 (0.07) -0.03 (0.07)
Town_Max -0.10 (-0.02) -0.19 (-0.09) -0.23 (-0.10) -0.31 (-0.17) -0.36 (-0.24) -0.35 (-0.22) -0.32 (-0.20) -0.32 (-0.19)
Town_Mean -0.05 (0.04) -0.11 (-0.01) -0.14 (0.00) -0.20 (-0.05) -0.24 (-0.10) -0.24 (-0.10) -0.22 (-0.09) -0.22 (-0.09)
Median household income -0.18 -0.35 -0.37 -0.39 -0.40 -0.37 -0.33 -0.33
Prob1,Prob5, and Prob10 denote three GIS data layers of modeled groundwater arsenic occurrence, representing probability of arsenic [1, 5, and 10 lg/L, respectively;
Town_Min,Town_Max, and Town_Mean denote town-level minimum, maximum, and mean for the modeled arsenic probability values, respectively; Smoothed_500 etc. denote
the smoothed rates; e.g., smoothed_500 denotes the town-level preterm birth rate smoothed from the Original Rate using a threshold that the neighborhood of smoothing must
enclose at least 500 births; the value inside the parentheses is radjusted by town-level income (i.e., rbetween the arsenic probability value and the residual to the LBW-income
linear regression); the bottom line contains the rvalues between the term-LBW rate and the town median household income in 1999
Environ Geochem Health
123
Environ Geochem Health
123
to some extent. For term LBW, the correlation
between the outcome and household income is
noticeable for maternal age C25. The adjustment by
income reduces the negativeness in the rvalues
between LBW rates and arsenic levels. For the stratum
of maternal age\25, the correlation between the LBW
rate and the income is very weak. While these findings
are seemingly variable, they can have a fairly consis-
tent interpretation: Younger mothers, who may have
not established a career and/or a stable income, may be
less sensitive to the expected negative effect of an
economical variable, especially measured at a highly
aggregated level, and as a result, the impact of
environmental hazards on them might be easier to
detect. For older mothers, income might be a stronger
independent variable in the equation. This may also
indicate that town-level income data may not well
represent the economic status of younger mothers.
This study reveals that a smoothing process may
have a considerable effect on detection of spatial
association. In the analysis of preterm birth and in the
analysis of younger mothers with term LBW, it
appears that more statistically stable rates (i.e., more
smoothing) help reveal potential associations between
the reproductive outcomes and the groundwater
arsenic. We are aware that for less populous areas
such as northern New Hampshire, there is a greater
risk of over-smoothing, i.e., the smoothed rates may
not correctly reflect the local variability of the disease
risk, and in turn, may affect the reliability of the
detected association between the disease and environ-
mental exposures.
We find that the spatial associations between the
reproductive outcomes and the groundwater arsenic
have spatial variation in New Hampshire, and this
regional pattern is consistent, as expected, for the two
reproductive outcomes we examined. Among the 10
counties of NH, the several most populous counties
usually have stronger spatial associations between
birth problems and groundwater arsenic than the
others. Geographically, these counties are clustered
in the southeastern corner of New Hampshire and are
more proximate to the greater metropolitan Boston
area. There are at least three factors that may have
contributed to the stronger correlations they possess:
(1) These counties are the most populated, having a
majority of the state’s population, and, as such,
provided a larger sample size with more stable
disease rates than in other parts of the state, which
may have better reflected the actual influence of
groundwater arsenic on reproductive outcomes; (2)
some parts of these counties have the highest arsenic
probability values in the USGS modeled data, and
the high arsenic exposure levels may have increased
the detectability of an association between the
reproductive outcomes and groundwater arsenic;
and (3) compared with the rest of the state, the
arsenic probability values vary the most in this
region, which also may have facilitated the detection
of its spatial association with LBW.
The New Hampshire population uses either public
or private water supplies. There is no arsenic regula-
tory requirement for private water supplies, which
results in some having fairly high arsenic concentra-
tions. In contrast, public supplies, by law, must have
arsenic monitored and controlled. A limitation of this
study is that we did not have precise information to
distinguish different water sources in different places.
Partially because of this, similar to most geospatial
analyses in health studies, the goal of this study has
been set to be ‘‘exploration.’’ We use geospatial
analysis and the best available data to explore if there
is a possibility that arsenic in groundwater has an
association with adverse reproductive outcomes. To
have accurate and precise information about people’s
source of drinking water, it requires much more
extensive and expensive investigations, which is
outside the scope of the study presented by this paper.
It should also be noted that, as a fairly rural state, New
Hampshire’s 40 % population using private wells
disproportionally occupy a much larger geographic
area than the other 60 % of the population. Neverthe-
less, it will be of interest to make the distinction
between private and public water supplies in future
data collection and analyses.
bFig. 7 County-specific correlation coefficient (r) between
town-level term-LBW rate for maternal age C25 and ground-
water arsenic probability values in New Hampshire,
1997–2009. Notes Prob1,Prob5, and Prob10 denote three
GIS data layers of modeled groundwater arsenic occurrence,
representing probability of arsenic [1, 5, and 10 lg/L,
respectively; the term-LBW rates used in this figure have been
adjusted by the town median household income of 1999;
Min_Ori denotes the rbetween the town-level minimum
probability of arsenic occurrence and the income-adjusted
original term-LBW rate; Min_500 denotes the rbetween the
town-level minimum probability of arsenic occurrence and the
income-adjusted term-LBW rate smoothed with thresh-
old =500 births; and so on
Environ Geochem Health
123
In this study, we used three GIS data layers of
modeled probability of arsenic exceeding certain
concentration, including 1, 5, and 10 lg/L. While
the results based on the three data layers are generally
consistent, the distinction is noticeable. For the
preterm birth analysis, the one with the highest bar,
probability [10 lg/L, tend to bring about strongest
positiveness. For the term-LBW analysis, the one with
the lowest bar, probability [1lg/L, sometimes is
more ‘‘positive’’ than the others. What is also
noteworthy is the different ‘‘performances’’ of the
three representations of town-level arsenic: minimum,
maximum, and mean. For the preterm birth analysis,
the town mean provides the most positive rvalues, and
for the term-LBW analysis, the town minimum
slightly outperforms the other two. A general lesson
learned from these findings is that in geospatial
analysis for environmental health studies, a thorough
consideration and exploration of different data, repre-
sentations, and parameter settings is necessary.
In principle, the analysis should take into account all
confounding factors, subject to availability of data.
Maternal age is the only confounding factor we have
data at the individual level. Race/ethnicity, a known
confounding factor, was not taken into account of this
study, due to lackof data of this variable at the individual
level. The New Hampshire population is over 95 %
Caucasian; hence the impact of this data limitation is
minimal. We analyzed the town-level income data as
our best-possible effort so far to address socioeconomic
factors. Nutritional factors influence biomarker concen-
trations of arsenic and could potentially be confounders.
This is a limitation of our analysis, but we are limited by
data in this study, and will certainly take them into
account whenever data allow.
Whereas the correlations reported in this paper
appear to indicate a relation between the two repro-
ductive outcomes and groundwater arsenic concen-
tration, it does not lead to the conclusion that arsenic
from bedrock wells is the true cause of the relation. It
is possible that other correlates or combinations of
factors that follow a similar pattern to that of arsenic is
responsible for the relation we observed. Future
investigations of these and other relations may provide
additional insight into understanding the effect of
exposure to arsenic from private wells in the region.
Finally, we consider it is worth emphasizing the fact
that in this study, we found a relation to probabilities of
exceeding low-to-moderate arsenic concentration. The
finding is novel also because very little is known about
adverse reproductive outcomes and arsenic exposure in
a US population. It has international value because
there are so many parts of the world with low-to-
moderate concentrations of arsenic in drinking water
that are not generally considered as a health concern.
Acknowledgments Acknowledgements: This study is
supported by USNIH grants P20RO18787, P20ES018175, and
USEPA grant RD83459901.The authors thank the anonymous
reviewers for their insightful comments and suggestions.
*ArcGIS is a product name of the Environmental System
Research Institute, Redland, California, US. Any use of trade,
firm, or product names in this paper is for descriptive purposes
only and does not imply endorsement by the US Government.
References
Ahmad, S. A., Sayed, M. H., Barua, S., Khan, M. H., Faruquee,
M. H., & Jalil, A. (2001). Arsenic in drinking water and
pregnancy outcomes. Environ Mental Health Perspectives,
109, 629–631.
Andra, S. S., Makris, K. C., Christophi, C. A., & Ettinger, A. S.
(2013). Delineating the degree of association between
biomarkers of arsenic exposure and type-2 diabetes mel-
litus. International Journal of Hygiene and Environmental
Health, 216, 35–49.
Ayotte, J. D., Cahillane, M., Hayes, L., & Robinson, K. W.
(2012). Estimated probability of arsenic in groundwater
from bedrock aquifers in New Hampshire, 2011. U.S.
Geological Survey Scientific Investigations Report,
2012–5156. http://pubs.usgs.gov/sir/2012/5156/.
Ayotte, J. D., Nolan, B. T., Nuckols, J. R., Cantor, K. P., Rob-
inson, G. R., Baris, D., et al. (2006). Modeling the proba-
bility of arsenic in groundwater in New England as a tool
for exposure assessment. Environmental Science and
Technology, 40(11), 3578–3585.
Ayotte, J. D., Szabo, Z., Focazio, M. J., & Eberts, S. M. (2011).
Effects of human-induced alteration of groundwater flow on
concentrations of naturally-occurring trace elements at
water-supply wells. Applied Geochemistry, 26(5), 747–762.
Braud, T., Nouer, S., & Lamar, K. (2011). Residential proximity
to toxic release sites and the implications for low birth
weight and premature delivery. Journal of Environmental
Health, 73(6), 8–13.
Ettinger, A. S., Zota, A. R., Amarasiriwardena, C. J., Hopkins,
M. R., Schwartz, J., Hu, H., et al. (2009). Maternal arsenic
exposure and impaired glucose tolerance during preg-
nancy. Environmental Health Perspectives, 117,
1059–1064.
Flanagan, S. F., Ayotte, J. D., & Robinson, Jr. G. R. (2011).
Quality of water from crystalline rock aquifers in New
England, New Jersey, and New York, 1995–2007. U.S.
Geological Survey Scientific Investigations Report,
2011–5220. http://pubs.usgs.gov/sir/2011/5220/.
Environ Geochem Health
123
Helsel, D. R. (2005). Nondetects and data analysis: Statistics for
censored environmental data (1st ed.). New York, NY:
Wiley.
Helsel, D. R., & Hirsch, R. M. (1992). Statistical methods in
water resources. New York, NY: Elsevier Science Com-
pany Inc.
Hopenhayn, C., Ferreccio, C., Browning, S. R., Huang, B.,
Peralta, C., Gibb, H., et al. (2003). Arsenic exposure from
drinking water and birth weight. Epidemiology, 14,
593–602.
Hosmer, D. W., & Lemeshow, S. (2000). Applied logistic
regression (2nd ed.). New York, NY: Wiley.
Huyck, K. L., Kile, M. L., Mahiuddin, G., Quamruzzaman, Q.,
Rahman, M., Breton, C. V., et al. (2007). Maternal arsenic
exposure associated with low birth weight in Bangladesh.
Journal of Occupational and Environmental Medicine/
American College of Occupational and Environmental
Medicine, 49, 1097–1104.
Kafadar, K. (1994). Choosing among two-dimensional smoo-
thers in practice. Computational Statistics & Data Ana-
lysis, 18, 419–439.
Kashima, S., Naruse, H., Yorifuji, T., Ohki, S., Murakoshi, T.,
Takao, S., et al. (2011). Residential proximity to heavy
traffic and birth weight in Shizuoka, Japan. Environmental
Research, 111, 377–387.
Kippler, M., Wagatsuma, Y., Rahman, A., Nermell, B., Persson,
L. A., Raqib, R., et al. (2012). Environmental exposure to
arsenic and cadmium during pregnancy and fetal size: A
longitudinal study in rural Bangladesh. Reproductive
Toxicology, 34, 504–511.
Kwok, R. K., Kaufmann, R. B., & Jakariya, M. (2006). Arsenic
in drinking-water and reproductive health outcomes: A
study of participants in the Bangladesh integrated nutrition
programme. Journal of Health Population Nutrition, 24(2),
190–205.
Lin, C.-M., Li, C.-Y., Yang, G.-Y., & Ma, I.-F. (2004). Asso-
ciation between maternal exposure to elevated ambient
sulfur dioxide during pregnancy and term low birth weight.
Environmental Research, 96, 41–50.
Milton, A. H., Smith, W., Rahman, B., Hasan, Z., Kulsum, U.,
Dear, K., et al. (2005). Chronic arsenic exposure and
adverse pregnancy outcomes in Bangladesh. Epidemiol-
ogy, 16(1), 82–86.
Myers, S. L., Lobdell, D. T., Liu, Z., Xia, Y., Ren, H., Li, Y.,
et al. (2010). Maternal drinking water arsenic exposure and
perinatal outcomes in inner Mongolia, China. Journal of
Epidemiology and Community Health, 64, 325–329.
Ozdenerol, E., Williams, B. L., Kang, S. Y., & Magsumbol, M.
S. (2005). Comparison of spatial scan statistic and spatial
filtering in estimating low birth weight clusters. Interna-
tional Journal of Health Geographics, 4, 19.
Rahman, A., Vahter, M., Smith, A. H., Nermell, B., Yunus, M.,
& El Arifeen, S. (2009). Arsenic exposure during preg-
nancy and size at birth: A prospective cohort study in
Bangladesh. American Journal of Epidemiology, 169,
304–312.
Reader, S. (2001). Detecting and analyzing clusters of low-birth
weight incidence using exploratory spatial data analysis.
GeoJournal, 53(2), 149–159.
Shi, X., Duell, E., Demidenko, E., Onega, T., Wilson, B., &
Hoftiezer, D. (2007). A polygon-based locally weighted
average method for smoothing disease rates of small units.
Epidemiology, 18(5), 523–528.
Slama, R., Morgenstern, V., Cyrys, J., Zutavern, A., Herbarth,
O., Wichmann, H.-E., et al. (2007). Traffic-related atmo-
spheric pollutants levels during pregnancy and offspring’s
term birth weight: A study relying on a land-use regression
exposure model. Environmental Health Perspectives,
115(9), 1283–1292.
Sohel, N., Vahter, M., Ali, M., Rahman, M., Rahman, A.,
Streatfield, P. K., et al. (2010). Spatial patterns of fetal loss
and infant death in an arsenic-affected area in Bangladesh.
International Journal of Health Geographics, 9, 53.
Stallones, L., Nuckols, J. R., & Berry, J. K. (1992). Surveillance
around hazardous waste sites: Geographic information
systems and reproductive outcomes. Environmental
Research, 59(1), 81–92.
Talbot, T. O., Kulldorff, M., Forand, S. P., & Haley, V. B.
(2000). Evaluation of spatial filters to create smoothed
maps of health data, statistics in medicine. Statistics in
Medicine, 19, 2399–2408.
Tofail, F., Vahter, M., Hamadani, J. D., Nermell, B., Huda, S.
N., Yunus, M., et al. (2009). Effect of arsenic exposure
during pregnancy on infant development at 7 months in
rural matlab, Bangladesh. Environmental Health Per-
spectives, 117(2), 288–293.
Tu, W., Tedders, S., & Tian, J. (2011). An exploratory spatial
data analysis of low birth weight prevalence in Georgia.
Applied Geography, 32, 195–207.
Vahter, M. E., Li, L., Nermell, B., Rahman, A., El Arifeen, S.,
Rahman, M., et al. (2006). Arsenic exposure in pregnancy:
A population-based study in Matlab, Bangladesh. Journal
of Health, Population, and Nutrition, 24, 236–245.
von Ehrenstein, O. S., Mazumder, D. N. G., Hira-Smith, M.,
Ghosh, N., Yuan, Y., Windham, G., et al. (2006). Preg-
nancy outcomes, infant mortality, and arsenic in drinking
water in West Bengal, India. American Journal of Epide-
miology, 163, 662–669.
Waller, L. A., & Gotway, C. A. (2004). Applied spatial statistics
for public health data. Hoboken, NJ: Wiley.
WHO. (2002). The World Health Report 2002: Reducing Risks,
Promoting Healthy Life. Geneva: World Health
Organization.
Wilcox, A. J. (2001). On the importance—and the unimpor-
tance—of birthweight. International Journal of Epidemi-
ology, 30, 1233–1241.
Xiang, H., Nuckols, J. R., & Stallones, L. (2000). A geographic
information assessment of birth weight and crop produc-
tion patterns around mother’s residence. Environmental
Research Section A, 82, 160–167.
Yang, C.-Y., Chang, C.-C., Tsai, S.-S., Chuang, H.-Y., Ho, C.-
K., & Wu, T.-N. (2003). Arsenic in drinking water and
adverse pregnancy outcome in an arseniasis-endemic area
in northeastern Taiwan. Environmental Research, 91,
29–34.
Environ Geochem Health
123
... Furthermore, the majority of mothers live in counties where private well use is rare. Our finding of a null association of arsenic with gestational age differs from comparable epidemiologic studies conducted within the U. S. (Almberg et al., 2017;(Claus Henn et al., 2016); Shi et al., 2015), but not all (Gilbert-Diamond et al., 2016;Howe et al., 2020). Several of the aforementioned studies directly measured individual-level total arsenic exposure using biomarkers ((Claus Henn et al., 2016) Note: CI, confidence interval. ...
... In general, studies that have used arsenic exposure biomarkers have observed null associations with gestational age (Gilbert-Diamond et al., 2016;Howe et al., 2020). In contrast, Almberg et al. (2017) estimated exposures using total arsenic concentrations measured exclusively in public water systems and averaged across counties, whereas Shi et al. (2015) estimated exposures using a logistic regression model of total arsenic in private well water that was aggregated to the town-level. Despite a very similar study design, Shi et al. (2015) observed positive associations of arsenic levels in private well water with preterm birth in New Hampshire. ...
... In contrast, Almberg et al. (2017) estimated exposures using total arsenic concentrations measured exclusively in public water systems and averaged across counties, whereas Shi et al. (2015) estimated exposures using a logistic regression model of total arsenic in private well water that was aggregated to the town-level. Despite a very similar study design, Shi et al. (2015) observed positive associations of arsenic levels in private well water with preterm birth in New Hampshire. However, some key differences in study design might explain the discrepancy in findings. ...
Article
Background Prenatal exposure to drinking water with arsenic concentrations >50 μg/L is associated with adverse birth outcomes, with inconclusive evidence for concentrations ≤50 μg/L. In a collaborative effort by public health experts, hydrologists, and geologists, we used published machine learning model estimates to characterize arsenic concentrations in private wells—federally unregulated for drinking water contaminants—and evaluated associations with birth outcomes throughout the conterminous U.S. Methods Using several machine learning models, including boosted regression trees (BRT) and random forest classification (RFC), developed from measured groundwater arsenic concentrations of ∼20,000 private wells, we characterized the probability that arsenic concentrations occurred within specific ranges in groundwater. Probabilistic model estimates and private well usage data were linked by county to all live birth certificates from 2016 (n = 3.6 million). We evaluated associations with gestational age and term birth weight using mixed-effects models, adjusted for potential confounders and incorporated random intercepts for spatial clustering. Results We generally observed inverse associations with term birth weight. For instance, when using BRT estimates, a 10-percentage point increase in the probability that private well arsenic concentrations exceeded 5 μg/L was associated with a −1.83 g (95% CI: −3.30, −0.38) lower term birth weight after adjusting for covariates. Similarly, a 10-percentage point increase in the probability that private well arsenic concentrations exceeded 10 μg/L was associated with a −2.79 g (95% CI: −4.99, −0.58) lower term birth weight. Associations with gestational age were null. Conclusion In this largest epidemiologic study of arsenic and birth outcomes to date, we did not observe associations of modeled arsenic estimates in private wells with gestational age and found modest inverse associations with term birth weight. Study limitations may have obscured true associations, including measurement error stemming from a lack of individual-level information on primary water sources, water arsenic concentrations, and water consumption patterns.
... We identified 10 peer-reviewed papers that investigated the association of arsenic exposure with PTB [24,31,[49][50][51][52][53][54][55][56] (Table 4). A majority were conducted in Asian developing countries: three papers were from Bangladesh, two from China, and one from each of Taiwan and Myanmar. ...
... Papers from western countries included two from the United States and one from Spain. A majority of the studies assessed ground water in the form of tube wells or dug wells while one determined arsenic level from safe drinking water [49][50][51][52][53][54]56]. The remaining studies measured arsenic levels in maternal serum [55], placental tissue [24], and maternal urine [31]. ...
Research
Full-text available
Preterm birth (PTB) and its complications are the leading causes of under-five year old child deaths, accounting worldwide for an estimated one million deaths annually. The etiology of PTB is complex and multifactorial. Exposures to environmental metals or metalloids are pervasive and prenatal exposures to them are considered important in the etiology of PTB. We conducted a scoping review to determine the extent of prenatal exposures to four metals/metalloids (lead, mercury, cadmium and arsenic) and their association with PTB. We reviewed original research studies published in PubMed, Embase, the Cochrane Library, Scopus, POPLINE and the WHO regional indexes from 2000 to 2019; 36 articles were retained for full text review. We documented a higher incidence of PTB with lead and cadmium exposures. The findings for mercury and arsenic exposures were inconclusive. Metal-induced oxidative stress in the placenta, epigenetic modification, inflammation, and endocrine disruptions are the most common pathways through which heavy metals and metalloids affect placental functions leading to PTB. Most of the studies were from the high-income countries, reflecting the need for additional data from low-middle-income countries, where PTB rates are higher and prenatal exposure to metals are likely to be just as high, if not higher.
... According to Huang et al. (2021), the prevalence of preterm birth in Bangladesh, known as the most polluted country with chronic As poisoning (arsenicosis), is the highest in the world, accounts for 35% of infant mortality, and is particularly prevalent in rural regions due to exposure to As 3+ -contaminated well water. The association of preterm birth and As-contaminated groundwater was also confirmed by a study performed by Shi et al. (2015). Punshon et al. (2015) reported that placental As levels were related to As levels in maternal urine, maternal toenails and infants' toenails, and household drinking water. ...
Article
Full-text available
According to recent research, even low levels of environmental chemicals, particularly heavy metals, can considerably disrupt placental homeostasis. This review aims to explore the profile of non-essential trace metals in placental tissues across the globe and to specify trace metal(s) that can be candidates for impaired placental health. Accordingly, we conducted an extensive survey on relevant databases of peer-reviewed papers published in the last two decades. Among a considerable number of non-essential trace metals, arsenic (As), lead (Pb), cadmium (Cd), and mercury (Hg) were identified as the most detrimental to placental health. Comparative analysis showed remarkable differences in placental levels of these trace metals worldwide. Based on current data reported across the globe, a median (min-max) range from 0.55 to 15 ng/g for placental As levels could be deemed safe. The placental Cd and Pb levels were markedly higher in smokers than in non-smokers. Occupationally exposed pregnant women had several orders of magnitude higher Cd, Pb, and Hg levels in placental tissues than non-occupationally exposed women. Also, we concluded that even low-level exposure to As, Cd, Pb, and Hg could be deleterious to proper fetal development. This review implies the need to reduce exposure to non-essential trace metals to preserve placental health and prevent numerous poor pregnancy outcomes. Overall, the information presented is expected to help plan future fundamental and applied investigations on the placental toxicity of As, Cd, Pb, and Hg.
... Arsenic exposure of fetuses and babies dramatically increases the risk of cancer and other diseases in adulthood (Farzan et al., 2013). Exposure has been associated with adverse birth outcomes, such as reduced birth weight, in New Hampshire (Shi et al., 2015). Exposure to even low levels of arsenic have been shown to result in lower IQ in children in Maine (Wasserman et al., 2014). ...
Article
Full-text available
Secondary schools in Maine and New Hampshire have been involved in a citizen science program called "All About Arsenic" aimed at addressing arsenic contamination of well water, one of the most pressing public health issues in both states. Nearly half of the population of Maine and New Hampshire derive their drinking water from private wells which often have arsenic levels above the EPA limit of 10 ppb. Arsenic exposure can cause cancer, adverse cardiovascular effects, and other health problems. Addressing this issue in schools provides context and motivation for students to engage in scientific inquiry and acquire data literacy skills. This project involves students collecting well water samples for arsenic analysis, entering their data into an online citizen science data portal, Anecdata, and using Tuva online software tools to visualize and interpret their data. Students present their data at public meetings to inform community members of their findings with the goal of moving "data to action". The COVID-19 pandemic presented multiple challenges for teachers engaging their students in this citizen science project. We adapted our program and implemented a series of interventions aimed at supporting teachers in their continued efforts to engage their students the "All About Arsenic" project.
... Laine et al., 2015), USA (e.g. Shi et al., 2015;Claus Henn et al., 2016;Gilbert-Diamond et al., 2016;Almberg et al., 2017), where very high levels of As are found in drinking water, effects such as a low birth weight and death of the unborn child or young infant have been observed relatively frequently, as evident in the reviews based on global data by various workers (e.g. Bloom et al., 2014 andQuansah et al., 2015). ...
Article
This review presents a holistic overview of the occurrence, mobilization, and pathways of arsenic (As) from predominantly geogenic sources into different near-surface environmental compartments, together with the respective reported or potential impacts on human health in Latin America. The main sources and pathways of As pollution in this region include: (i) volcanism and geothermalism: (a) volcanic rocks, fluids (e.g., gases) and ash, including large-scale transport of the latter through different mechanisms, (b) geothermal fluids and their exploitation; (ii) natural lixiviation and accelerated mobilization from (mostly sulfidic) metal ore deposits by mining and related activities; (iii) coal deposits and their exploitation; (iv) hydrocarbon reservoirs and co-produced water during exploitation; (v) solute and sediment transport through rivers to the sea; (vi) atmospheric As (dust and aerosol); and (vii) As exposure through geophagy and involuntary ingestion. The two most important and well-recognized sources and mechanisms for As release into the Latin American population's environments are: (i) volcanism and geothermalism, and (ii) strongly accelerated As release from geogenic sources by mining and related activities. Several new analyses from As-endemic areas of Latin America emphasize that As-related mortality and morbidity continue to rise even after decadal efforts towards lowering As exposure. Several public health regulatory institutions have classified As and its compounds as carcinogenic chemicals, as As uptake can affect several organ systems, viz. dermal, gastrointestinal, peptic, neurological, respiratory, reproductive, following exposure. Accordingly, ingesting large amounts of As can damage the stomach, kidneys, liver, heart, and nervous system; and, in severe cases, may cause death. Moreover, breathing air with high As levels can cause lung damage, shortness of breath, chest pain, and cough. Further, As compounds, being corrosive, can also cause skin lesions or damage eyes, and long-term exposure to As can lead to cancer development in several organs.
... The strongest association between arsenic exposure and PTB was in the sample of "younger" mothers exposed to the highest level of arsenic (>10 mg/L). 52 This study included any delivery occurring before 37 weeks' gestation and did not differentiate between spontaneous vs medically indicated PTBs. There is a growing body of US-based population studies examining the associations between arsenic exposure during pregnancy and the development of preeclampsia or PTB. ...
Article
Preeclampsia and preterm birth are among the most common pregnancy complications and are the leading causes of maternal and fetal morbidity and mortality in the United States (U.S.). Adverse pregnancy outcomes are multifactorial in nature and increasing evidence suggests that the pathophysiology behind preterm birth and preeclampsia may be similar – specifically, both of these disorders may involve abnormalities in placental vasculature. A growing body of literature supports that exposure to environmental contaminants in the air, water, soil, consumer and household products serves as a key factor influencing the development of adverse pregnancy outcomes. In pregnant women, toxic metals have been detected in urine, peripheral blood, nail clippings, and amniotic fluid. The placenta serves as a ‘gatekeeper’ between maternal and fetal exposures, as it can reduce or enhance fetal exposure to various toxicants. Proposed mechanisms underlying toxicant-mediated damage include disrupted placental vasculogenesis, an upregulated proinflammatory state, oxidative stressors contributing to prostaglandin production and consequent cervical ripening, uterine contractions, and ruptured membranes, and epigenetic changes that contribute to disrupted regulation of endocrine and immune system signaling. The objective of this review is to provide an overview of studies examining the relationships between environmental contaminants in the U.S. setting, specifically inorganic (e.g., cadmium, arsenic, lead, and mercury) and organic [e.g., per- and polyfluoroalkyl substances (PFAS)] toxicants, and the development of preeclampsia and preterm birth among U.S. women.
Article
To investigate prenatal exposure to arsenic and its effect on birth size, we conducted a cross-sectional study in Wujiang City, Jiangsu, China, from June 2009 to June 2010. A total of 1722 mother-infant pairs were included in the study. A questionnaire was administered to the pregnant women and umbilical cord blood(UCB) samples were collected. Arsenic concentration in UCB was detected by inductively coupled plasma emission mass spectroscopy (ICP-MS). The birth size included birth weight, birth body length and head circumference of the newborns. The effects of arsenic exposure on birth size were assessed by multiple linear regression analysis. Arsenic concentrations in UCB ranged from 0.11 to 30.36 μg/L, the median was 1.71 μg/L. In this range of exposure, arsenic concentration was significantly negatively associated with birth weight, especially among male infants. Our results showed that prenatal exposure to arsenic level was low in Wujiang City, China. However, low prenatal arsenic exposure could have negative effects on birth weight. Our research provided evidence for the adverse effects of prenatal low-level arsenic exposure on the intrauterine growth of the fetus.
Chapter
Geographical information system (GIS) is gaining its popularity beyond geography and information technology (IT) with its strong power in managing and analysing spatial data. In medical geology, GIS provides two main useful functions: (a) mapping and (b) spatial analysis. It contains specialised computer software and hardware designed to process data with locational information. Besides geology and medical information, data in medical geology usually contain locational information which is suitable for mapping and spatial analysis using GIS.
Chapter
Medical Geology provides a holistic framework to analyze the potential health effects from exposures to natural materials and to the environmental impacts of mineral resource development and chemical and nuclear waste disposal. The Environmental Pathways/Biological Impact Analysis comprises a systematic progression of analyses starting with a potential release of a contaminant into the environment, leading to population exposures and resulting health effects. This chapter describes the first part of the analyses, the Environmental Pathways Analysis, which produces an estimate of the amount of a pollutant that could reach a potentially exposed population. Typically, the amount of a pollutant that is available to affect the health of humans is less than the amount released from the source. This reduction of the amount and mobility of the contaminant in the geosphere is associated with a number of abiotic and biotic processes in the environment and is related to its geoavailability as discussed in this chapter. Different compartments within the Environmental Pathways Analysis involve many different disciplines in the physical, chemical, geological, and environmental sciences. The methods include materials analysis, laboratory studies, and theoretical predictions of the speciation and solubility of contaminants based on principles of chemical thermodynamics and kinetics, hydrogeologic modeling and field measurements, remote sensing, and spatial analysis. Analysis of the geoavailability of a contaminant provides realistic estimates of the potential exposure resulting from releases from pollutant sources. This chapter summarizes the basic concepts that underlie geoavailability and identifies key references to aid the reader in applying them to practical Medical Geology applications.
Article
This study analyzed the exposure and risk assessment of four toxic (Hg, Cd, As, Tl) and two essential (Se, Mo) elements in 119 Spanish women of reproductive age. The focus here was on the elements for which risk‐based benchmark, biomonitoring equivalents (BE), or health‐related human biomonitoring (HBM) values have already been established. All elements presented frequencies of detection of 100% (% > limit of detection), except for Cd (99%). The 95th percentile concentrations for the toxic metals were 358.37 µg/L (total As), 1.10 µg/L (Cd), 0.41 µg/L (Tl) and 3.03 µg/L (total Hg), and for the essential elements, they were 68.95 µg/L (total Se) and 154.67 (Mo). We examined socio‐demographic factors and dietary habits of women as predictors of urinary metal concentrations. Arsenic was positively associated with fish, shellfish, and canned fish consumption, while Mo was found to be associated with the consumption of cereals and pastry products. Maternal urine levels of As were negatively correlated with gestational age. This article is protected by copyright. All rights reserved.
Chapter
In this chapter, we introduce statistical methods for the analysis of spatial data that arise as point events in space (e.g., the locations of disease “cases”). We compare and contrast the ideas of “clustered”, “random”, and “regular” spatial patterns and provide introductory descriptions of the first- and second-order statistical properties of a spatial point pattern. The heterogeneous Poisson process is introduced to allow us to assess spatial patterns in disease events after accounting for spatial variations in population density. We recommend Monte Carlo simulation as a flexible tool for hypothesis testing and illustrate its utility in several “data breaks” that provide examples of how our hypotheses can be sequentially refined to answer a variety of practical questions about the nature of the spatial patterns we observe.
Conference Paper
The well water in Lanyang Basin, which is located in the northeastern portion of Taiwan island, was found to have high levels of arsenic ranging from undetectable levels (< 0.15 ppb) to 3.59 ppm. We performed a study to compare the risk of adverse pregnancy outcomes (preterm delivery and birthweight) between an area with historic high well water arsenic levels (AE) and a comparison area with no historic evidence of arsenic water contamination (NAE). The mean birth weight in the AEs and NAEs were 3132.6 gm and 3162.6 gm respectively. Babies born in AEs were on average 30 gm lighter than those born in NAEs. AEs had a higher rate of preterm delivery than NAEs (3.74% vs 3.43%). The results of this study suggest that, after adjustment for potential confounders, arsenic exposure from drinking well water, was associated, although not significantly, with the risk of preterm delivery, with an odds ration of 1.10 (0.91-1.33). The estimated reduction in birth weight was 29.05 gm (95% CI= 13.55-44.55). The findings from this investigation provides evidence for a potential role for arsenic exposure through drinking water in increasing the risk of low birthweight.
Book
From the reviews of the First Edition."An interesting, useful, and well-written book on logistic regression models . . . Hosmer and Lemeshow have used very little mathematics, have presented difficult concepts heuristically and through illustrative examples, and have included references."—Choice"Well written, clearly organized, and comprehensive . . . the authors carefully walk the reader through the estimation of interpretation of coefficients from a wide variety of logistic regression models . . . their careful explication of the quantitative re-expression of coefficients from these various models is excellent."—Contemporary Sociology"An extremely well-written book that will certainly prove an invaluable acquisition to the practicing statistician who finds other literature on analysis of discrete data hard to follow or heavily theoretical."—The StatisticianIn this revised and updated edition of their popular book, David Hosmer and Stanley Lemeshow continue to provide an amazingly accessible introduction to the logistic regression model while incorporating advances of the last decade, including a variety of software packages for the analysis of data sets. Hosmer and Lemeshow extend the discussion from biostatistics and epidemiology to cutting-edge applications in data mining and machine learning, guiding readers step-by-step through the use of modeling techniques for dichotomous data in diverse fields. Ample new topics and expanded discussions of existing material are accompanied by a wealth of real-world examples-with extensive data sets available over the Internet.
Book
Out of print. Was replaced by the 2nd edition, titled "Statistics for Censored Environmental Data Using Minitab and R". See that entry on my list of publications.
Book
Has been replaced by the 2020 edition, with all new R code and updated methods. Download the 2020 version at https://doi.org/10.3133/tm4a3 . The ebook sold by Elsevier is the outdated first edition. The full text of the newer 2020 version at https://doi.org/10.3133/tm4a3 is much more current, and free.
Article
Low birth weight (LBW), defined as a live birth weighing less than 2500 g, is a significant public health problem in the United States, but the complex nature of the etiology of this problem is not fully understood. Moreover, significant disparities in LBW prevalence are well documented among certain populations, particularly in minority and underserved communities. The identification of spatial patterns of LBW prevalence is a critical first step in a more complete understanding of the epidemiology of this public health challenge and these techniques are instrumental in designing valid observational and analytical studies to more fully study the problem. This paper examines the spatial patterns of LBW prevalence, as well as the presence of spatial clusters in the State of Georgia at both the county and census tract levels. Unadjusted and empirical Bayes smoothed LBW rates were mapped to visualize the spatial variation of LBW rates, and the Moran’s I statistic and the Local Indicator of Spatial Association (LISA) statistic were computed to assess the degree of spatial dependence in the LBW rates. Results revealed marked geographical variation in LBW prevalence in Georgia in 2000. In addition, these data validate the significant disparity (two-fold difference) between white and black racial subgroups as documented in the literature. Trends associated with positive and negative spatial autocorrelations illustrated variation with respect to race. Limitations of data and methods, as well as plans for utility of the results of this study for further investigation were discussed.