Ecological and Economic Modeling of Hookworm Burden under Climate Change

Background The disease burden of hookworm is disproportionately borne by tropical and sub-tropical regions because hookworm thrives in warm surroundings with moist soil. Climate change will alter the ecological strength of transmission, affecting the spatial extent and intensity of hookworm burden. Methods This paper explains the relationship between hookworm transmission strength and geographic factors, and constructs a global hookworm ecological index (HEI) at a 0·5-degree spatial resolution. We test the HEI against spatial variation in hookworm prevalence, and then use three climate models to estimate the change in the spatial extent of ecological suitability for hookworm. Finally, we estimate the impact of climate change on public health costs and human capital loss due to hookworm. Results The models predict that East Asia, Latin America and the Middle East will experience increases in hookworm prevalence while South Asia and Sub-Saharan Africa will experience decreases. Our estimates suggest that if climate change occurred now, public health costs to avert DALYs from hookworm would rise by $680 million, 72-268 thousand fewer children would enroll in school, up to 165 thousand fewer children would attain literacy, and incomes would decrease for people across much of the developing world. Conclusion This paper represents the first assessment of the global effect of climate change on the population at risk and economic burden of hookworm infections.


I. Background
In 2015, there were approximately 430 million cases of hookworm infection around the world [1]. Hookworm primarily causes malnutrition and iron-deficiency anemia, particularly in pregnant women and children given their lower iron stores. Infection has been shown to lead to lower school attendance, literacy, and worker productivity, thus decreasing incomes and economic growth [2].
Parasitic nematodes thrive in warm and humid environments. We construct a hookworm ecology index (HEI) as a function of how temperature, rainfall, relative humidity, and soil texture dictate the reproductive and survival rate of hookworm. Climate change will alter the ecological strength of transmission, affecting the spatial extent and intensity of hookworm burden. The use of GIS to construct suitability maps for prediction has been demonstrated for other diseases [3][4][5], and the applicability for climate change has been shown in the case of malaria [6]. We measure the effect of the HEI on spatial variation in hookworm prevalence rates, and then predict prevalence under different climate futures. Since hookworm prevalence rates are also dependent on other non-climate factors (such as availability of medical treatment), our exercise measures the effect of climate change assuming no change in current-day public health efforts or the spatial distribution of the human population.

a. The Ecology of Hookworm
The most common species of hookworm that infects humans are Ancylostoma duodenale and Necator americanus. These nematodes live and reproduce in the upper section of the human small intestine. Eggs are transferred in the stool and, under favorable conditions, embryonate within one to two days. The first larvae form lasts for five to ten days; this is followed by a three to four week infective stage during which the larvae must find a human host to prevent dessication [4]. However, infection rates highly depend on the rate at which hookworms eggs/larvae develop and survive.
The endemicity of hookworm infection is determined by the reproductive number (the lifetime number of surviving offspring a hookworm in a particular host produces in the absence of density dependent constraints) [7]. The reproductive number of hookworm independent of age and population density is: where λ is the fecundity rate and μ is the mortality rate. The fecundity rate is the number of eggs produced by a female hookworm in a day while the mortality rate is the number of eggs (2) ( ) = −0 · 0028 2 + 0 · 1348 − 0 · 7244 Figure 1 shows the data fitted using a quadratic function. Larvae cease to develop below 10 degree Celsius or above 40 degree Celsius.
Relative humidity and rainfall also influence hookworm development. Ova of soil-transmitted helminthes generally do not embryonate below relative humidity of 50 percent [4]. Hookworms are more susceptible to desiccation than other soil-transmitted helminthes. A rainfall annual average of more than 40 inches (or 1,016 millimeters) is almost required for ova to develop to the first larvae stage and for the larvae to develop to the infective stage [9]. Increased precipitation prevents egg and larvae from drying up, but excessive rainfall could also potentially reduce the hatch rate [10].
Although weather data is available at monthly or even daily temporal resolution, we note that long-term average rainfall is the relevant ecological driver for hookworm endemicity. Although a location might have months where the basic reproduction number is below unity, the long lifespan of adult hookworm (typically 1-10 years) maintains overall endemicity as long as the basic reproduction number is on average above one over longer periods [4]. The HEI is therefore constructed for every month using the required yearly rainfall divided by twelve months. The monthly HEI is then aggregated to the yearly level.
The final environmental variable we incorporate is soil texture. Hookworms thrive in moist, sandy soil, and transmission is facilitated when porous soil allows them to easily migrate to the surface [11]. When the soil is high in clay content, hookworm larvae are restricted in movement and tend to desiccate. The soils of coarser texture are more effective for eggs to develop towards the larvae stage (although very coarse sands produce drier conditions because of high permeability). Efficiency decreases rapidly as soil becomes less granulated (increase in fineness). We incorporate the effect of soil type into the HEI using data from an experiment on hookworm development using ova in feces under different soil types, documenting the relation of soil texture to hookworm development [12]. Table 1 reports the percentage of ova that develops under different soil types.

II. Methods
The Hookworm Ecology Index takes the following value for every grid cell i in month m: The HEI is further modified by the following constraints: 1. HEI takes on a value of zero if relative humidity falls below 50 percent.
2. HEI takes on a value of zero if temperature is less than 10 degree Celsius or greater than 40 degree Celsius.
3. An annual precipitation of 40 inches (1,016 millimeters) is required for hookworm to survive. Since the HEI is constructed at the monthly level, 1,016 is divided by 12. The HEI for a given month is equal to zero if precipitation is less than 85 millimeters.
The HEI is calculated using gridded data on temperature, precipitation, relative humidity, and soil content. Historical simulations and future projections are generated by three global climate models from the Coupled Model Intercomparison Project Phase 5 (CMIP5) [13]. CCSM4, CESM1-CAM5, and HadGEM2-AO each have outputs for four representative concentration pathways (RCP) scenarios and historical simulations. We use RCP 4·5 and RCP 8·5 to give a plausible range of the change in hookworm burden. Three variables are obtained for each model: (1) near-surface air temperature which is the temperature reported at the 2-meter height, (2) relative humidity expressed as a percentage (ratio of current absolute humidity to the highest absolute humidity), and (3)  Clay, silt, and sand content of topsoil (expressed as a percent of total soil content) are available at a 0·5-degree spatial resolution [15]. Since relative humidity was available at a coarser resolution, the centroids of the downscaled temperature and precipitation were spatially joined with relative humidity centroids using ArcGIS. The HEI is then constructed at the monthly level at a 0·5-degree spatial resolution.
The HEI was validated using available hookworm prevalence data both at the country and village level. Country-level validation was conducted by regressing country prevalence rates (provinces in the case of China, and states in India) [

III. Results
The average Hookworm Ecology Index during 1996-2005 is mapped in Figure 2. The index is highest in central Africa, the northwestern Amazon, and the southeastern United States. The map also outlines countries with nonzero hookworm prevalence as of 2003 [16], as well as the location of subnational data on prevalence rates used for validation of the index. Table 2 shows the results from regressing the country/province prevalence rates on the 10-year HEI average for each climate model. The magnitude of the association is consistent across the three climate models and statistically significant at 1 percent levels. A one-standard deviation higher level in HEI of 0·02 translates into a 10% higher level of prevalence. Table 3 runs the analysis at the village level, and shows that the association between the HEI and village prevalence rates is statistically significant and consistent across climate models, whether or not a country fixed effect is included to absorb all unobserved country-level variables that might confound the estimated effect of HEI on prevalence. Figure 3 graphically shows the within-country relationship between hookworm prevalence and the HEI (in this case constructed using climate data from the Hadley model), clearly showing the strength of the association.
Using the mean of the three coefficients in the model with country fixed effects, a within-country one standard deviation higher HEI of 0·014 translates into a 4 percentage point higher hookworm prevalence rate (0·2 standard deviations). Note that the HEI index explains around 25% of the variation in hookworm prevalence within a country, underlining the important role of ecology in disease prevalence while leaving a large part of variation explained by socioeconomic and public health system factors not in our model.

a. Estimating disease burden
We use the results in Tables 2 and 3 combined with gridded population data for the year 2000 in urban and rural areas [18] and national age distribution data [19] in

b. Health and economic burden due to climate change
Hookworm infection rarely leads to death, however heavy infection leads to lethargy, anemia, growth stunting, and weaker immunological response to other infections. The documented social consequences of hookworm infection include school absenteeism and reduced educational attainment in children, leading to adverse outcomes as adults in terms of labor force productivity [20].
One way to measure the direct cost of a disease is by modeling cost of treatment. In terms of hookworm, mass drug treatments are conducted by distributing deworming pills, usually in school-based treatment programs. Soil-transmitted helminth infections are treated with ivermectin, albendazole, azithromycin, or praziquantel. These drugs can cost $0·46 to $0·90 per patient if economies of scale are achieved by commitment of pharmaceuticals to provide free drugs, synergizing delivery modes, and commitments of communities and schools to distribute the drugs [21]. Since most infections occur in childhood age, the economic burden of hookworm occurs through adversely affecting educational outcomes. Estimates from the successful eradication of hookworm in the American South in 1910 suggest that enrollment, attendance, and literacy increased by 8%, 16%, and 5% respectively in areas where hookworm prevalence was 100 percent [22]. Economic benefits from deworming are also realized later in life with increased income. One estimate suggests that deworming increases the net present value of wages by over $40 per treated person [21]. An experiment in Kenya measured the positive effects of schoolbased deworming on health, education, and labor market outcomes and found that a decade after the deworming, moderate to heavy infections decreased by 17%, total years enrolled in school increased by around 0.3 years, and time spent working among men increased by around 17% or 3·49 hours a week [20]. We use these estimates together with the changes in HEI to estimate the change in hookworm's economic burden due to climate change. Table 7

c. Sensitivity Analyses
We analyzed the sensitivity of the results by presenting both RCP 4·5 and 8·5 scenarios from three climate models, and present ranges on our estimates of children at risk, public health costs and human capital losses for both rural and urban areas. These ranges result from the two regression estimates of the association between HEI and hookworm prevalence (the first using country-level data and the second using village-level data) as reported in Tables 2 and 3.

IV. Discussion
The

V. Conclusion
This paper represents the first assessment of the global effect of climate change on the population at risk and economic burden of hookworm infections. Results suggest that climate change will lead to an intensification of hookworm burden in some regions, and the opposite in other regions where the optimal temperature threshold is surpassed or changes in precipitation patterns reduce the ecological suitability for hookworm. Understanding the prevalence and spatial distribution of hookworm infections under climate change scenarios contributes to measuring the overall social impact of climate change as well as focusing cost-effective health interventions to areas of greatest need.

Declarations
Authors' contributions: GCM conceived the study and guided analysis, KR collected data and conducted analysis. Both authors wrote the manuscript.
Ethics approval and consent to participate: N/A

Consent for publication: N/A
Competing Interests: The authors declare that no conflicts of interest exist.

Funding:
No funding was received for this work.
Availability of data: The datasets used and analyzed during the current study are available from the corresponding author upon request.