Content uploaded by Kelly Austin
Author content
All content in this area was uploaded by Kelly Austin on Jun 06, 2017
Content may be subject to copyright.
AIMS Environmental Science, 4(2): 217-231.
DOI: 10.3934/environsci.2017.2.217
Received: 31 October 2016
Accepted: 27 February 2017
Published: 01 March 2017
http://www.aimspress.com/journal/environmental
Research article
Anthropogenic forest loss and malaria prevalence: a comparative
examination of the causes and disease consequences of deforestation in
developing nations
Kelly F. Austin1,*, Megan O. Bellinger2 and Priyokti Rana2
1 Department of Sociology and Anthropology, Lehigh University
2 Integrated Degree in Engineering, Arts, and Sciences, Lehigh University
* Correspondence: Email: kellyaustin@lehigh.edu; Tel: +610 758-2103.
Abstract: Malaria represents an infectious disease keenly tied to environmental conditions, as
mosquitoes represent the disease vector. Many studies are beginning to document that changes in
environmental conditions, such as deforestation, can greatly alter the density and activity of mosquito
populations and therefore malaria rates. While numerous epidemiological studies examine the links
between forest loss and mosquito proliferation in distinct locales, comparative assessments across
multiple sites are lacking. We attempt to address this gap by imparting a cross-national analysis of
less-developed, non-desert, malaria endemic nations. Using a structural equation model of 67 nations,
we find positive associations between deforestation rates and malaria prevalence across nations. Our
results also suggest that rural population growth and specialization in agriculture are two key
influences on forest loss in developing nations. Thus, anthropogenic drivers of environmental
degradation are important to consider in explaining cross-national variation in malaria rates.
Keywords: malaria; deforestation; infectious disease
1. Introduction
The World Health Organization boasts that over the last 15 years, malaria incidence has decreased
by 37% and deaths from malaria declined by 60% globally [1]. While this is undoubtedly marked
progress, it is important to acknowledge that headway is not uniform across regions or countries, with
some nations, particularly in Sub-Saharan Africa and some areas of Southeast Asia, lagging behind in
218
AIMS Environmental Science Volume 4, Issue 2, 217-231.
malaria improvements [1]. Furthermore, malaria represents an infectious disease keenly tied to
environmental conditions with mosquitoes as the disease vector [2,3]; thus, changes in environmental
conditions can greatly alter the density and activity of mosquito populations and therefore malaria
rates (e.g., [2]). As human alterations to the natural environment are only increasing in scale and
intensity over time [3,4], we cannot assume that progress in addressing this disease will continue with
manifest success. Instead, research needs to continue to investigate how human activities may be
altering ecosystem conditions in ways that enhance mosquito habitats and transmission of malaria.
Although a number of forms of environmental degradation or ecosystem alteration could affect
mosquito populations, deforestation is an increasingly studied factor in several locales across Asia,
Sub-Saharan Africa, and, most prominently, Latin America (e.g., [5-8]). While a number of case
studies identify links between forest loss and malaria or mosquito prevalence [5-9], whether
relationship exists across areas or regions remains insufficiently examined. We attempt to contribute to
this line of inquiry by imparting a cross-national analysis of less-developed, non-desert nations to
examine the potential association deforestation and malaria prevalence across countries.
Prior examinations of deforestation and malaria tend to focus on the epidemiological aspects of
ecosystem change and mosquito habitat proliferation [5,6,8-10]. While these studies provide key
evidence on the direct mechanisms that cause deforestation to lead to increased levels of larvae,
parasite concentrations, mosquito populations, or actual malaria rates [5,6,8-10], they fail also to
consider fully the wider socio-economic context in which forest loss occurs. Our study contributes to
this endeavor by simultaneously investigating the causes and disease consequences of deforestation in
developing nations. By illuminating some of the key anthropogenic forces behind deforestation, our
study contributes to broader understanding on how environmental degradation affects global trends in
infectious disease.
2. Background
2.1. Malaria: characteristics and global trends
In 2015, around 450,000 people died from malaria worldwide [1]. The burden of this disease
spreads across around 90 countries, with 88% of malaria cases concentrated in the Sub-Saharan
African region [1,11]. Over the last several decades, partially due to major initiatives and programs,
such as the WHO‘s ―Roll Back Malaria, Roll in Development‖, there have been major reductions in
malaria deaths and incidence [1,11]. However, this disease is still a leading threat to health and
premature death in several countries [1]. Poverty represents one of the greatest risks for malaria
cross-nationally [12]; undoubtedly, socioeconomic development fundamentally relates to the global
distribution of morbidity and mortality for many health issues (e.g., [13-18]).
Caused by a parasite in the genus Plasmodium, four of the species, P. falciparum, P. vivax, P.
malariae and P. ovale, lead to malaria in humans [19]. These parasites rely on two hosts throughout
their lifecycle: a mosquito and a human (or another animal) [19]. Only around 30–40 of the 400 species
of Anopheles mosquito can act as vectors for the parasite, and only the bite of a female Anopheles
mosquito spreads the parasite from one person to another [19,20]. The different types of malaria cause
different variations of the disease in humans, with some forms being milder than other strains [19,20].
P. falciparum is the most prevalent species in Africa and causes the most deaths worldwide [1,11].
219
AIMS Environmental Science Volume 4, Issue 2, 217-231.
Early diagnosis of malaria is crucial and essential for management of the disease [1,13]. The
WHO recognizes microscopy and rapid diagnostic tests as the current most common and effective
ways of diagnosing the disease [1,11]. However, the quality of this method can be inadequate due to
lack of sensitivity and insufficient expertise of personnel required to analyze results properly [1,11,21].
Especially in poor nations and rural areas, access to quality healthcare is limited [13,12,21]. Currently,
there are many drugs available to combat the infectious parasite, but some are not very effective [1,22,23].
The most effective and popular drug, artemisinin combination treatments (ACTs), contain an active
ingredient that is a derivative of artemisinin from the sweet wormwood tree [11,23]. Other drug
treatments, including quinine and chloroquine were also popular in past decades; however, the parasite
has developed resistance to these drug treatments [1,23]. In addition, recent studies already document
increased tolerance to ACTs in many regions, leading to concerns over the potentials of widespread
resistance [22].
As previously emphasized, patterns in malaria prevalence and international development are
closely linked, as rates are highest in the poorest nations and, conversely, eradication of malaria took
place decades ago in affluent nations [13,24,12,23]. Many studies in the areas of global health and
development emphasize that impoverished people lack knowledge of disease prevention techniques,
have little access to modern ‗Western‘ medicine, and face limited access to appropriate
preventative strategies, such as screened windows or insecticide-treated bed nets [13,24,21]. In
addition, many poor households lack adequate access to water or sanitation, increasing potential
mosquito habitats (e.g., [12,21,24-26]).
Education in particular signifies a key determinant of malaria incidence. For example, Dike et al. [27]
find that formal education teaches basic skills that remove cultural ideologies leading to misconceptions
that affect proper malaria management. In particular, people of higher education were more likely to
identify mosquitoes as the malaria disease vector and to use insecticide-treated bed nets [27]. Numerous
other studies of malaria report similar findings (e.g., [21,26,28]), suggesting that education also enhances
people‘s awareness of Western medicine techniques and propensity to seek proper treatment. The potential
importance of education in predicting cross-national malaria rates fits with other comparative global health
assessments, including examinations of HIV, life expectancy, and infant and child mortality, which find
that participation in schooling greatly improves health outcomes across nations (e.g., [14-18]).
While social and economic factors are important in shaping cross-national disease trends,
environmental factors are also especially relevant with mosquitoes as the vector [2]. Thus, prevalence
patterns are not only shaped by education, healthcare access, and poverty, but also the degree to which
environmental conditions exist or are created that can sustain or flourish mosquito habitats [2,3].
Location in tropical or sub-tropical zones alone is argued to account for a significant amount of
vulnerability to malaria, as the parasite quickly becomes weak and eliminated below 16 degrees
Celsius [29]. Additionally, the role of other environmental factors, such as deforestation, represents an
emerging area of inquiry (e.g., [7,8]). As deforestation results from human activities, this represents a
potential underlying anthropogenic cause of the malaria burden in many developing countries.
2.2. Links between forest loss and malaria
Today over 3 billion people worldwide remain at risk of acquiring malaria [13], and many of these
vulnerable people tend to live in areas where environmental factors facilitate Plasmodium growth and
prevalence. Mosquitoes fundamentally depend upon stagnant water and warm temperatures for their
220
AIMS Environmental Science Volume 4, Issue 2, 217-231.
breeding and life cycle, explaining why malaria largely remains a tropical disease. Deforestation represents
an additional environmental factor that can influence malaria [2,6,8,9,30,31]. The goal of this section is to
draw on case study and epidemiological evidence from distinct studies to establish a rationale for
examining the relationship between deforestation and malaria prevalence at the cross-national level.
Deforestation can impact malaria prevalence by several mechanisms, including increasing the
amount of sunlight and standing water in some areas [2,6,9,32]. Although it is important to emphasize
that different sub-types of Anopheles mosquitoes prefer shadier versus more sunlit habitats, or may be
more or less sensitive to land use changes in general, typically, increasing standing water and sunlight
is favorable for most species [2,6,9,30,32]. Deforestation potentially contributes to these processes in
multiple ways. For example, primary growth forests tend to be heavily shaded with thick debris on the
ground, which absorbs water and often leaves any standing water acidic [32]. After clearing non-steep
terrain, the land usually become flatter and more likely to pool water, which is typically less acidic and
more conducive to Anopheles larvae development [32]. Not only does ponding more readily increase
the availability of breeding grounds for the malaria vector, but increases in sunlight resulting from
deforestation also promotes more ideal breeding grounds by warming temperatures [2,6-9]. When
agriculture replaces forested areas, the plants can still provide the bushy cover needed for some species
of Anopheles mosquito or stages of larvae development. Increases in mosquito reproduction
potentially impact rates of malaria transmission to nearby populations [2,6,9].
Another possible link between forest loss and malaria concerns biodiversity (e.g., [33,34]).
Disease ecologists find that higher levels of biodiversity generally take on a protective role for the
human population, in a so-called ―dilution effect.‖ If there is a wider variety of species available, the
proportion that are able to transmit the vector are reduced and thus transmission of diseases occurs less
frequently [33,34]. One of the primary effects of deforestation is a loss in biodiversity, as a single,
typically nonnative, crop often replaces a huge variety of vegetation and animal life [33]. Establishing
a connection between forest loss and malaria is not straightforward; there are many confounding
factors or factors specific to certain regions and locales. Thus, comparative assessments that examine
trends in forest loss and malaria across regions can help to establish whether we can generalize the
findings from epidemiological case studies to other areas endemic to the Plasmodium parasite.
Several studies conducted in sites in Asia indeed establish a link between malaria prevalence and
forest loss. For example, studies conducted in China and Myanmar determine that the pupation rate of
Anopheles minimus, a malaria vector, was highest in samples collected from deforested areas ([7,8]).
Similarly, a study carried out over the course of a year in Vietnam found that there was a statistically
significant relationship between percentage of forest cover loss and prevalence of multiple strains of
malaria [10]. In Sub-Saharan Africa, one study set in the highlands of Kenya in a case-control format
found that living on land without trees led to increased risk for malaria contraction [5].
South America, particularly the Amazon region, represents perhaps the most popular site to
examine the links between environmental degradation and malaria. For example, studies conducted in
the Peruvian Amazon, which has experienced deforestation due to small-scale farming, discovered that
Anopheles darlingi had a statistically significant preference for living and breeding in areas that had
experienced a loss of forest cover [6,9]. The observed connection between malaria rates and
deforestation persisted even when considering changes in population density [9]. A study that sampled
from over 800 water sources throughout the Amazon found that Anopheles darlingi sites were more
likely to be located in areas that had experienced alterations in land cover, such as reductions in forest
and the establishment of new croplands [6].
221
AIMS Environmental Science Volume 4, Issue 2, 217-231.
These studies demonstrate that multiple researchers across different regions and variety of
Anopheles mosquito species identify a link between deforestation and malaria rates or mosquito
populations. In fact, Yasuoka and Levins [35] conduct a content analysis of case studies centered on
changes in ecology and malaria. They find around 60 examples from a variety of areas, including
Thailand, Nepal, India, China, Guyana, Uganda, and others, that demonstrate a link between
deforestation and land-use changes and growth in mosquito populations or malaria incidence [35].
It is important to emphasize that deforestation is not a natural phenomenon, but rather results
predominantly from human activities [4]. Expanding agricultural areas represents an underlying cause of
deforestation in the majority of these studies (e.g., [6,9,30]). Population growth within developing nations
can lead to increased pressure for food production, but also many developing nations are encouraged to
expand agricultural exports as a means for development (e.g., [36]). Aside from food demands, rural
population pressures overall are likely to spur forest loss [4]. Many developing nations where malaria is
most prevalent continue to have high rates of fertility and rural population growth [14]. Research also
identifies access to timber for building or for use as fuelwood as key causes of deforestation in developing
nations (e.g., [37,38]). Thus, rural population dynamics and pressures to increase food production are
likely to be key anthropogenic causes of deforestation. In our analysis, we engage an innovative
statistical method, structural equation modeling, in order to examine not only the potential role of
deforestation in explaining cross-national variation in malaria prevalence, but also to account for the
underlying causes of forest loss.
3. Materials and Methods
3.1. Sample
The sample is restricted to malaria-endemic nations. Malaria-endemic nations are nations that have
a constant and measurable incidence of malaria and are located in natural areas of transmission [1].
Countries that register malaria cases resulting from imported cases are not included in the analyses.
Malaria-endemic nations overwhelmingly have GDP per capita estimates in the lower three quartiles of
the income classification of countries [1], thus largely representing a sample of less-developed nations.
We exclude desert nations as deforestation is only relevant and measurable in countries with significant
forest stock. Our sample includes 67 non-desert, less-developed nations for which data are reported for
the key variables in the analysis, including the deforestation rate and malaria prevalence. For a complete
list of the countries included in the analyses, see Table 1.
3.2. Analytic strategy
To examine the association between deforestation and malaria prevalence, as well as the underlying
predictors of deforestation, we utilize structural equation models (SEMs). SEM is particularly useful in
this context based on its efficient ability to model direct and indirect effects (e.g., [39]). While most
comparative analyses utilize direct effects approaches, this strategy prevents modeling the mediating
and interrelationships outlined above, such as the potential pathway involving rural population growth,
deforestation, and malaria. Structural equation modeling also allows us to utilize latent or composite
constructs of multi-dimensional concepts, such as health resources, and circumvent issues of
multicollinearity, which often occur with cross-national data [40].
222
AIMS Environmental Science Volume 4, Issue 2, 217-231.
Table 1. Countries included in the Analyses (N = 67).
Angola
SSA
Gambia, The
SSA
Pakistan
AS
Bangladesh
AS
Ghana
SSA
Papua New Guinea
PAC
Belize
LA
Guatemala
LA
Peru
LA
Benin
SSA
Guinea
SSA
Philippines
AS
Bhutan
AS
Guinea-Bissau
SSA
Rwanda
SSA
Bolivia
LA
Guyana
LA
Sao Tome & Principe
SSA
Burkina Faso
SSA
Haiti
LA
Senegal
SSA
Burundi
SSA
Honduras
LA
Sierra Leone
SSA
Cabo Verde
SSA
India
AS
Solomon Islands
PAC
Cambodia
AS
Indonesia
AS
South Africa
SSA
Cameroon
SSA
Kenya
SSA
Sudan
SSA
Central African Rep.
SSA
Lao PDR
AS
Swaziland
SSA
Chad
SSA
Liberia
SSA
Tajikistan
AS
China
AS
Madagascar
SSA
Tanzania
SSA
Colombia
LA
Malawi
SSA
Timor-Leste
AS
Comoros
SSA
Mali
SSA
Togo
SSA
Congo, Dem. Rep.
SSA
Mauritania
SSA
Uganda
SSA
Congo, Rep.
SSA
Mozambique
SSA
Vanuatu
PAC
Cote d'Ivoire
SSA
Namibia
SSA
Vietnam
AS
Dominican Rep.
LA
Nepal
AS
Zambia
SSA
Ecuador
LA
Nicaragua
LA
Zimbabwe
SSA
El Salvador
LA
Niger
SSA
Ethiopia
SSA
Nigeria
SSA
Notes: AS—Asia, LA—Latin America, PAC—Pacific Islands, SSA—Sub-Saharan Africa
Another key aspect of SEMs involves the estimation of fit statistics that enable the researcher to
judge the fit of the model as a whole to the data provided and compare equally plausible models [39].
An additional feature of SEM is its utilization of maximum likelihood (ML) missing value routine that
calculates pathway coefficients based on all available data points; when cases are missing information
on select variables, those cases are dropped from those pathway estimations, but retained for others
when the data are available [40]. Thus, SEM allows us to maximize our sample of nations by retaining
cases that might be missing data on one or two control variables included in the models [40].
In SEM, we simultaneously estimate a system of linear equations that correspond to our hypotheses
about the correlations in our observed data [39]. For empirical identification, it is important that the
models estimate normally [39]. We utilize SEM software in AMOS and Mplus; in both statistical
packages, the path diagram displayed in Figures 1 and 2 all estimate normally. The assumptions of SEM
include multivariate normality, completely random missing data, sufficiently large sample, and the
correct model specification [40]. To protect against the negative consequences of multivariate
non-normality we also ran the analyses using the robust ML estimator (MLR) implemented in Mplus [41].
These estimates with the MLR estimator are robust to non-normality [41]. The results were consistent with
those achieved with the ML procedure. Although data should be completely missing at random, the use of
the maximum likelihood (ML) estimator also provides consistent estimates under the assumption of
missing at random, which is an easier condition to satisfy [42]. Additionally, we fail to see any pattern to
223
AIMS Environmental Science Volume 4, Issue 2, 217-231.
the missing data to suggest that the data are not missing at random. Third, the positive asymptotic
properties of the ML estimation procedure (consistency and efficiency) are known when the sample size is
relatively large. As our sample is relatively small (N = 67), we re-ran our analyses with bootstrap standard
errors as well as a robust ML (MLR) procedure and obtained estimates and model fit statistics for
Figures 1 & 2 that are consistent with the ML estimator. Model specification errors occur with the
omission of relevant variables. If this occurs, the errors and the exogenous variables in the model correlate,
leading to biased estimates [40]. To avoid this, we draw on prior research to select a wide range of
variables. Our review of the literature engages perspectives on development, health, and the environment
into our structural model. We test all theoretically informed paths as shown in Figures 1 and 2 below.
3.3. Variables included in the analysis
The key dependent variable in the analysis is the malaria prevalence rate. We created the malaria
prevalence variable using data on the estimated number of malaria cases from the World Health
Organization [1] and total population level from the World Bank [43] for the year 2013. We weighted
the number of malaria cases for each nation by its total population, and then multiplied by 100,000 to
form the prevalence rate.
As the malaria parasite requires particular climate conditions, latitude represents an important
exogenous, environmental condition to consider. We transformed the latitude scores into absolute
values to capture distance from the equator [43].
We examine deforestation as a prominent form of environmental degradation that is likely to be
directly associated with increasing malaria prevalence. The natural deforestation rate represents an annual
percent change score, calculated using FAO estimates of natural forest area, from 2012 to 2013 [43]. These
data come from the Global Forest Resource Assessment (GFRA) and represent point estimates for natural
forest stock measured in thousand square hectares for 2012 and 2013. The natural forest area measure
includes land area that is more than 0.5 hectares which contains trees higher than 5 meters and a
canopy cover of more than 10%. We multiply the change score by -1 to capture rate of forest loss.
We consider both agriculture and rural population growth as key drivers of forest loss. We therefore
include a measure of rural population growth, which accounts for the annual percent change in the rural
population for the year 2012 [43]. We measure specialization in agriculture using data on the percent of the
economy that comes from agricultural production, or agriculture as a percent of GDP [43].
To assess the influence of economic development, we include GDP per capita. This represents
the total annual output of a country‘s economy divided by its population, measured in current
international dollars for the year 2012 [43]. More specifically, GDP per capita is the total market value
of all final goods and services produced in a country in a given year, equal to total consumer,
investment, and government spending, divided by the mid-year population. It is converted into current
international dollars using Purchasing Power Parity (PPP) rates, providing a standard measure
allowing for comparisons of real price levels between countries [43].
Public health conditions are important non-economic factors that are likely to influence malaria
rates in less-developed nations. We therefore measure public health conditions using a latent construct
comprised of the following indicators: number of physicians, the fertility rate, access to clean water,
and secondary schooling for the year 2012. Physicians, secondary schooling, access to clean water and
the fertility rate represent some of the most prominent predictors of malaria rates, as well as other
health outcomes in developing nations (e.g., [14-16]). The variable medical doctors measures the
224
AIMS Environmental Science Volume 4, Issue 2, 217-231.
number of physicians in a nation per 100,000 people. The fertility rate is an estimate of the number of
children an average woman would have if current age-specific fertility rates remained constant during
her reproductive years. Percent access to clean water refers to the percentage of the population using an
improved drinking water source. Improved drinking water sources include piped water located inside
the user‘s dwelling, plot, or yard, and other improved drinking water sources, such as public taps or
standpipes, tube wells or boreholes, protected dug wells, protected springs, and rainwater collection.
Secondary school enrollment represents a gross enrollment ratio, which calculates the ratio of total
enrollment, regardless of age, to the population of the age group that officially corresponds to
secondary level education. We acquired all of these measures from the World Bank [43].
Additionally, we include a regional indicator for Sub-Saharan Africa, as many Sub-Saharan
African nations have an especially high level of malaria and the highest concentrations of P.
falciparum, which is one of the severest strains of the parasite. We measure this as a dummy variable,
where countries coded with a ‗1‘ indicate location in Sub-Saharan Africa and those with a ‗0‘ indicate
that a nation is located in a different region of the world [43].
4. Results
Table 2 displays the bivariate correlation matrix and univariate statistics for all of the variables
used in the analyses. The magnitude of the bivariate correlation coefficients in Table 2 demonstrates
that many of the predictor variables in the sample are highly correlated, such as the indicators for
public health conditions variables. This further warrants the use of the SEM analytical technique given
its superior handling of inter-correlated independent variables through the creation of latent constructs
and direct and indirect pathways that circumvents the tendency to bias coefficient estimates [39].
Table 2. Correlation matrix and univariate statistics
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
1. Malaria Prevalence
1.00
2. Latitude
−0.365
1.00
3. GDP per capita
−0.591
0.175
1.00
4. Sub-Saharan Africa
0.652
−0.274
−0.563
1.00
5. Secondary Schooling
−0.582
0.157
0.777
−0.650
1.00
6. Medical Doctors
−0.531
0.367
0.698
−0.658
0.664
1.00
7. Clean Water
−0.409
0.195
0.592
−0.621
0.656
0.457
1.00
8. Fertility Rate
0.754
−0.304
−0.757
0.730
−0.793
−0.650
−0.658
1.00
9. Agriculture % GDP
0.608
−0.220
−0.753
0.445
−0.672
−0.506
−0.606
0.605
1.00
10. Rural Pop Growth
0.465
−0.196
−0.726
0.533
−0.633
−0.563
−0.487
0.738
0.530
1.00
11. Deforest-ation Rate
0.391
−0.158
0.490
0.174
−0.267
−0.140
−0.177
0.354
0.346
0.316
1.00
Mean
12433
13.49
4148.2
0.582
55.86
0.401
76.38
4.08
23.94
1.23
0.526
Standard Deviation
13996
8.40
3243.8
0.497
22.26
0.533
15.04
1.41
12.92
1.28
1.64
A preliminary step in the empirical assessment of our complete SEM was to validate empirically
whether public health conditions represent latent factors that can be appropriately estimated secondary
school enrollments, trained medical doctors, access to clean water, and the average fertility rate. To test
this, we initially construct a confirmatory factor analysis (CFA) of the measures and analyze the
225
AIMS Environmental Science Volume 4, Issue 2, 217-231.
overall and component measures of fit. By empirical standards, we find evidence at both the
component and overall model levels to support our predictions that these indicators can be used to
measure public health conditions. This fits with our substantive interpretations of prior health and
development research discussed earlier.
Figures 1 and 2 present the SEM results of malaria prevalence. We test all theoretically and
substantively informed paths in the path diagram displayed in Figure 1, and then eliminate all
non-significant relationships to present the most parsimonious model in Figure 2. Before turning to a
discussion of results, we note the model fit statistics indicate an excellent fit of both models to the data.
Specifically, for Figure 1, in accordance with empirical standards the chi-square test statistic is
non-significant (χ2 = 33.15 with df = 31); the values of the Incremental Fit Index (0.980),
Tucker-Lewis Index (0.965), and the Confirmatory Fit Index (0.979) all exceed 0.90; and the root
mean squared error of approximation (RMSEA) value (0.046) is below the suggested threshold of .10
for smaller samples [39].
Notes: ***p < 0.001, **p < 0.01, *p < 0.05 (two-tailed tests); standardized coefficients reported.
Figure 1. SEM predicting malaria prevalence, saturated Model.
The results from the trimmed model (where non-significant paths are eliminated) presented in
Figure 2 are also well within range of the empirical standards for SEMs. The chi-square test statistic is
non-significant (χ2 = 45.08 with df = 37); the values of the Incremental Fit Index (0.968),
Tucker-Lewis Index (0.966), and the Confirmatory Fit Index (0.981) all exceed 0.90; and the root
mean squared error of approximation (RMSEA) value (0.048) is below the appropriate threshold.
It is common when using SEMs to eliminate non-significant paths for the sake of parsimony.
When looking across the saturated model in Figure 1 and the trimmed model in Figure 2, there are a
few substantial differences in the path coefficients. The only changes are a small reduction in the
226
AIMS Environmental Science Volume 4, Issue 2, 217-231.
magnitudes for effects on malaria prevalence and deforestation in the more saturated model. As the
substantive findings are essentially consistent across the two models, and as the fit statistics are
acceptable in the second model as well as the first, we prefer to focus our discussion of the results on
the findings presented in Figure 2. In the path diagrams, standardized regression coefficients are
reported and flagged for statistical significance. We also include Table 3, which displays the
standardized regression coefficients in addition to the unstandardized regression coefficients and
standard errors for the model in Figure 2. However, we keep our interpretation of the results focused on
Figure 2, as the path diagram facilitates accessible interpretations.
Notes: ***p <0.001, **p < 0.01, *p < 0.05 (two-tailed tests); standardized coefficients reported.
Figure 2. SEM predicting malaria prevalence, trimmed model.
The results presented in Figure 2 suggest that deforestation is associated with increased
prevalence of malaria in developing nations, which is consistent with our predictions (0.24). In
addition, we find that economic specialization in agriculture (0.27) and rural population growth (0.28)
are positively associated with the deforestation rate. Thus, not only do we find empirical evidence for
links between deforestation and malaria across nations, but also that population growth in rural areas
and increased emphasis on agriculture contribute to deforestation.
In addition, we find that public health conditions have a significant, negative association with
malaria rates in less-developed nations. In particular, secondary school enrollments, medical doctors,
access to clean water, and the average fertility rate together represent a set of public health conditions
that are robustly associated with declines in malaria prevalence (−0.66). In addition to public health
conditions, latitude also explains cross-national variation in malaria prevalence, where nations located
further from the equator tend to have lower rates of prevalence (−0.18).
Our results also suggest that GDP per capita or level of economic development and location in
Sub-Saharan Africa are important underlying factors contributing to variations in the malaria burden
227
AIMS Environmental Science Volume 4, Issue 2, 217-231.
across developing nations in indirect ways. It is important to emphasize that the results across Figures 1
and 2 demonstrate that each of these variables has an indirect association to malaria prevalence.
Specifically, more economically developed nations tend to have improved public health conditions
(0.37), which then reduce the malaria burden. In addition, more affluent nations tend to have
significantly lower levels of rural population growth (−0.51) and increased economic specialization in
agricultural production (0.70), which in turn are associated with deforestation rates.
Table 3. Regression results for SEM depicted in Figure 2 predicting malaria prevalence.
Standardized
Regression
Coefficient
Unstandardized
Regression
Coefficient
Standard
Error
Latitude Malaria Prevalence
−0.179*
−294.49
144.13
Public Health Conditions Malaria Prevalence
−0.658***
−483.02
76.51
Deforestation Rate Malaria Prevalence
0.243*
784.28
217.01
Rural Pop Growth Deforestation Rate
0.280**
0.334
0.095
Agriculture % GDP Deforestation Rate
0.273**
0.087
0.002
Rural Pop Growth Public Health Conditions
−0.257**
−3.78
1.18
GDP per capita Public Health Conditions
0.368***
0.002
0.000
GDP per capita Rural Pop Growth
−0.508***
−0.001
0.000
GDP per capita Agriculutre % GDP
−0.604***
−0.003
0.000
Sub-Saharan Africa Public Health Conditions
−0.501***
18.99
2.93
Sub-Saharan Africa Rurual Pop Growth
0.270**
0.698
0.263
Public Health Conditions Secondary Schooling
0.858
-
-
Public Health Conditions Medical Doctors
0.756***
0.021
0.003
Public Health Conditions Clean Water
0.640***
0.511
0.088
Public Health Conditions Fertility Rate
−0.928***
−0.070
0.007
Notes: ***p < 0.001, **p < 0.01, *p < 0.05 (two-tailed tests); standardized coefficients flagged for statistical
significance; unstandardized coefficients reported in in italics; standard errors reported in parentheses.
We also find in Figure 2 that Sub-Saharan African nations tend to have considerably higher rates
of rural population growth than other nations (0.27), and that Sub-Saharan African nations also have
much weaker public health conditions in comparison to other nations (−0.50). Our results also suggest
that nations experiencing high levels of rural population growth are more likely to have poorer public
health conditions (−0.26). This fits with prior research identifying that rural populations face
especially poor access to and quality of a variety of health and social resources, including health
facilities and education [21].
Overall, our results suggest that deforestation is an important factor in explaining cross-national
variation in malaria prevalence. We also find that anthropogenic forces related to specialization in
agriculture and rural population pressures are associated with heightened levels of deforestation,
thereby suggesting that human activities contribute to some level of malaria prevalence. In considering
the relative weight of impact of latitude versus deforestation in the model, we find in our analyses that
forest loss is more relevant in explaining cross-national variability in malaria rates than being located
close to a tropical zone. Thus, human-induced environmental changes may be more significant in
predicting malaria prevalence than the truly ―natural‖ or inherent environmental conditions.
228
AIMS Environmental Science Volume 4, Issue 2, 217-231.
4. Conclusions & limitations
Nearly 130 million hectares of forest—an area almost equivalent in size to South Africa - have
been lost since 1990, according to a recent FAO report [4]. Not only does deforestation contribute to
profound impacts on climate change, biodiversity loss, and changing weather patterns, but this study
builds on a body of evidence that deforestation also influences malaria transmission. While isolated
studies have begun to identify links between forest loss and mosquito proliferation or malaria
prevalence in certain locales (e.g., [5-8]) this study adds to the research on this phenomena by finding
an association between deforestation and malaria prevalence at the cross-national level.
In addition, we extend our analysis to also to consider some of the most prominent causes of forest
loss in developing nations. Our findings suggest that rural population growth and economic
specialization in agriculture are significant in increasing levels of forest loss in many developing nations.
Therefore, not only are there potential links between environmental change and malaria, but we consider
the anthropogenic underpinnings of this form of degradation. Rural population pressures not only affect
forests, but public health conditions in developing nations as well. Perhaps in some areas, growing rural
populations who tend to live closer to the natural habitats of mosquitoes may experience further
compounding risks due to their lack of health resources and proximity to deforested areas.
One limitation of our study is sample size. Sample size is inherently restricted as it is only
appropriate to include malaria endemic, non-desert nations in the analyses. Given those parameters, a
sample of nearly 70 nations is actually robust, and our model adheres to the model fit and other
parameters appropriate for SEMs. We also re-ran our analyses with bootstrap standard errors as well as a
robust ML (MLR) procedure and obtained estimates and model fit statistics for Figures 1 & 2 that were
consistent with the ML estimator. Although the use of cross-national data represents a key contribution
of this research, there are notable limitations with this approach. Cross-national analyses cannot account
for the particular sub-national ecological factors that may affect malaria prevalence, such as soil
conditions, level of rainfall, or temperature variations, as well as the particular vector susceptibilities and
preferences that vary by region or sub-species of mosquito. These represent topics more appropriate for
the epidemiological field studies we draw on, and instead we prefer to focus on determining if
comparative analysis uncovers broad patterns linking deforestation to malaria across nations. In so doing,
we acknowledge that we cannot assume that deforestation will cause increased malaria rates in every
nation, region, or area, but rather emphasize that there appears to be a pattern across countries between
deforestation and malaria prevalence, where the nations with higher rates of deforestation also tend to
have higher rates of malaria, net of other factors. Our results help to expand the generalizability of case
studies exploring forest loss and mosquito proliferation across different sites (e.g., [5-10]). Certainly,
given this complex topic, we hope to encourage further research examining how forest loss influences
trends in malaria and other mosquito-borne diseases in and across diverse settings.
Our study makes an interdisciplinary contribution by bringing together ideas and insights from
across fields of global development and sociology, global health, epidemiology, environmental science,
and human ecology. We consider many of the social, demographic, economic, and environmental
predictors found to be important in predicting both malaria prevalence and deforestation from across
these disciplines. The use of structural equation models (SEMs) allows us to test the interconnections
and complex pathways among these different types of indicators, and serves to illuminate how
international inequalities frame the proximate predictors of malaria rates in developing nations, such
as forest loss and public health conditions.
229
AIMS Environmental Science Volume 4, Issue 2, 217-231.
Importantly, we find there is no direct association between GDP per capita and Sub-Saharan Africa
and malaria prevalence. Although many studies make direct links between rising affluence and malaria
management, our results suggest that economic growth will be most effective if channeled to improving
public health conditions or addressing rural population growth. Similarly, the indirect associations
involving Sub-Saharan Africa imply that there is nothing inherent about this region that explain
disproportionately high malaria rates; rather our results suggest that Sub-Saharan African nations suffer
from a lack of public health services and extreme levels of rural population growth. Our results also find
links between GDP per capita and economic specialization in agriculture, where poorer nations in our
sample are much more likely to generate a larger share of their economic revenues from food production,
which in turn are associated with increased deforestation rates and malaria prevalence. Although we do
not focus on exports specifically, it is important to keep in mind that a notable amount of food produced
in poor nations is destined for markets in affluent nations [44]. Developing nations are often encouraged
by supra national organizations and core governments to specialize in agriculture as a means to
development [36]; perhaps these policy recommendations should be interpreted with caution given the
potentially contaminant influences on forests and malaria rates.
Although there have been major improvements in malaria prevention, diagnosis, and treatment in
many nations over the last several decades [1], malaria remains a leading cause of death and threat to
health in many regions and countries across the Global South [1,12]. Some patterns in climate change,
deforestation, and other human-induced changes to the natural environment could alter and amplify
malaria transmission (e.g., [31]). In addition to the influence of environmental changes, resistance to
both insecticides and antimalarial medications is increasing [22,23]. As human alterations to the
natural environment are only intensifying, understanding and mitigating the underlying anthropogenic
causes of malaria transmission deserves vigilant attention.
Conflict of interest
The authors declare there is no conflict of interest.
References
1. World Health Organization (WHO), World Malaria Report 2016, 2016. Available from:
http://apps.who.int/iris/bitstream/10665/252038/1/9789241511711-eng.pdf?ua=1.
2. Norris D (2004) Mosquito-borne Diseases as a Consequence of Land Use Change. EcoHealth 1: 19-24.
3. Lewis S, Maslin M (2015) Defining the Anthropocene. Nature 519: 171-180.
4. Food and Agriculture Organization (FAO) of the United Nations, Global Forest Resources
Assessment 2015, 2015. Available from: http://www.fao.org/3/a-i4808e.pdf.
5. Ernst KC, Lindblade KA, Koech D, et al. (2009) Environmental, socio-demographic and
behavioural determinants of malaria risk in the western Kenyan highlands: a case-control study.
Trop Med Int Health 14: 1258-1265.
6. Vittor AY, Pan W, Gilman RH, et al. (2009) Linking deforestation to malaria in the Amazon:
characterization of the breeding habitat of the principal malaria vector, Anopheles darlingi. Am J
Trop Med Hyg 81: 5-12.
7. Hanandita W, Tampubolon G (2016) Geography and social distribution of malaria in Indonesian
Papua: a cross-sectional study. Int J Health Geogr 15: 13.
230
AIMS Environmental Science Volume 4, Issue 2, 217-231.
8. Wang X, Zhou G, Zhong D, et al. (2016) Life-table studies revealed significant effects of deforestation
on the development and survivorship of Anopheles minimus larvae. Parasite vector 9: 323.
9. Vittor AY, Gilman RH, Tielsch J, et al. (2006) The effect of deforestation on the human-biting
rate of Anopheles darlingi, the primary vector of falciparum malaria in the Peruvian Amazon.
Am J Trop Med Hyg 74: 3-11.
10. Manh BH, Clements ACA, Thieu NQ, et al. (2011) Social and environmental determinants of
malaria in space and time in Viet Nam. Int J Parasitol 41: 109-116.
11. World Health Organization (WHO), World Malaria Report 2015, 2015. Available from:
http://www.who.int/malaria/publications/world-malaria-report-2015/report/en/.
12. Packard RM, The making of a tropical disease: a short history of malaria. JHU Press, 2007.
13. Stratton L, O'Neill MS, Kruk ME, et al. (2008) The persistent problem of malaria: Addressing
the fundamental causes of a global killer. Soc Sci Med 67: 854-862.
14. Austin KF, Noble MD, Mejia MT (2014) Gendered vulnerabilities to a neglected disease: A
comparative investigation of the effect of women‘s legal economic rights and social status on
malaria rates. Int J Comp Sociol 55: 204-228.
15. Brady D, Kaya Y, Beckfield J (2007) Reassessing the effect of economic growth on well-being
in less-developed countries, 1980–2003. Stud Comp Int Develop 42: 1-35.
16. Burroway R (2010) Schools against AIDS: Secondary school enrollment and cross-national
disparities in AIDS death rates. Soc Probl 57: 398-420.
17. Shen C, Williamson JB (2001) Accounting for cross-national differences in infant mortality
decline (1965–1991) among less developed countries: Effects of women's status, economic
dependency, and state strength. Soc Indic Res 53: 257-288.
18. Williamson JB, Boehmer U (1997) Female life expectancy, gender stratification, health status,
and level of economic development: A cross-national study of less developed countries. Soc Sci
Med 45: 305-317.
19. Mayxay M, Pukrittayakamee S, Newton PN, et al. (2004) Mixed-species malaria infections in
humans. Trends Parasitol 20: 233-240.
20. Neafsey DE, Waterhouse RM, Abai MR, et al. (2015) Highly evolvable malaria vectors: the
genomes of 16 Anopheles mosquitoes. Science 347: 1258522.
21. Bates I, Fenton C, Gruber J, et al. (2004) Vulnerability to malaria, tuberculosis, and HIV/AIDS
infection and disease. Part II: determinants operating at environmental and institutional level.
Lancet Infect Diseases 4: 368-375.
22. Hastings IM, Ward SA (2005) Coartem (artemether-lumefantrine) in Africa: the beginning of
the end? J Infect Diseases 192: 1303-1304.
23. Shah S, The Fever, Picador, 2010.
24. Sachs J, Malaney P (2002) The economic and social burden of malaria. Nature 415: 680-685.
25. Lowassa A, Mazigo HD, Mahande AM, et al. (2012) Social economic factors and malaria
transmission in Lower Moshi, northern Tanzania. Parasite vector 5: 129.
26. Williams PCM, Martina A, Cumming RG, et al. (2009) Malaria prevention in Sub-Saharan
Africa: a field study in rural Uganda. J Commun Health 34: 288-294.
27. Dike N, Onwujekwe O, Ojukwu J, et al. (2006) Influence of education and knowledge on
perceptions and practices to control malaria in Southeast Nigeria. Soc Sci Med 63: 103-106.
28. Williams HA, Jones COH (2004) A critical review of behavioral issues related to malaria control in
sub-Saharan Africa: what contributions have social scientists made? Soc Sci Med 59: 501-523.
231
AIMS Environmental Science Volume 4, Issue 2, 217-231.
29. Blanford JI, Blanford S, Crane RG, et al. (2013) Implications of temperature variation for
malaria parasite development across Africa. Sci Rep 3: 1300.
30. Pattanayak S, Dickinson K, Corey C, et al. (2006) Deforestation, malaria, and poverty: a call for
transdisciplinary research to support the design of cross-sectoral policies. Sustainability:
Science, Practice, Policy 2(2).
31. Pimentel D, Cooperstein S, Randell H, et al. (2007) Ecology of increasing diseases: population
growth and environmental degradation. Hum Ecol 35: 653-668.
32. Patz JA, Graczyk TK, Geller N, et al. (2000) Effects of environmental change on emerging
parasitic diseases. Int J Parasitol 30: 1395-1405.
33. Myers SS, Gaffikin L, Golden CD, et al. (2013) Human health impacts of ecosystem alteration.
Proc Nat Acad Sci USA 110: 18753-18760.
34. Laporta GZ, de Prado PIKL, Kraenkel RA, et al. (2013) Biodiversity can help prevent malaria
outbreaks in tropical forests. PLoS Negl Trop Dis 7: e2139.
35. Yasuoka J, Levins R (2007) Impact of deforestation and agricultural development on anopheline
ecology and malaria epidemiology. Am J Trop Med Hyg 76: 450-460.
36. World Bank, World development report: Agriculture for development, 2008.
37. Bensel T (2008) Fuelwood, deforestation, and land degradation: 10 years of evidence from
Cebu Province, the Philippines. Land Degrad Dev 19: 587-605.
38. Jorgenson AK, Burns TJ (2007) Effects of Rural and Urban Population Dynamics and National
Development on Deforestation in Less-Developed Countries, 1990–2000. Sociol Inq 77: 460-482.
39. Bollen K, Structural Equations with Latent Variables, New York: John Wiley and Sons, Inc, 1989.
40. Kaplan D, Structural equation modeling: Foundations and extensions. Sage Publications, 2008.
41. Muthén LK, Muthén BO, Mplus User‘s Guide, 5th ed., Los Angeles: Muthén and Muthén, 2007.
42. Arbuckle JL (1996) Full information estimation in the presence of incomplete data. Advanced
structural equation modeling: Issues and technique 243: 277.
43. World Bank, World Databank, World Development Indicators, 2016. Available from:
http://databank.worldbank.org/data/views/variableSelection/selectvariables.aspx?source=world-
development-indicators#.
44. Austin KF, McKinney LA (2012) Disease, war, hunger, and deprivation: A cross-national
investigation of the determinants of life expectancy in less-developed and sub-Saharan African
nations. Sociol Perspect 55: 421-447.
45. Shandra CL, Shandra JM, London B (2011) World bank structural adjustment, water, and sanitation:
A cross-national analysis of child mortality in Sub-Saharan Africa. Organ Environ 24: 107-129.
46. Austin KF, McKinney LA, Thompson G (2012) Agricultural Trade Dependency and the Threat of
Starvation: A Cross-National Analysis of Hunger as Unequal Exchange. Int J Sociol 42: 68-89.
© 2017 Kelly F. Austin et al., licensee AIMS Press. This is an
open access article distributed under the terms of the Creative
Commons Attribution License
(http://creativecommons.org/licenses/by/4.0)