Synergies between environmental degradation and climate variation on malaria re-emergence in southern Venezuela: a spatiotemporal modelling study

Abstract

Background: Environmental degradation facilitates the emergence of vector-borne diseases, such as malaria, through changes in the ecological landscape that increase human-vector contacts and that expand vector habitats. However, the modifying effects of environmental degradation on climate-disease relationships have not been well explored. Here, we investigate the rapid re-emergence of malaria in a transmission hotspot in southern Venezuela and explore the synergistic effects of environmental degradation, specifically gold-mining activity, and climate variation.

Methods: In this spatiotemporal modelling study of the 46 parishes of the state of Bolívar, southeast Venezuela, we used data from the Venezuelan Ministry of Health including population data and monthly cases of Plasmodium falciparum malaria and Plasmodium vivax malaria between 1996 and 2016. We estimated mean precipitation and temperature using the ERA5-Land dataset and used monthly anomalies in sea-surface temperature as an indicator of El Niño events between 1996 and 2016. The location of suspected mining sites in Bolívar in 2009, 2017, and 2018 were sourced from the Amazon Geo-Referenced Socio-Environmental Information Network. We estimated measures of cumulative forest loss and urban development by km² using annual land cover maps from the European Space Agency Climate Change Initiative between 1996 and 2016. We modelled monthly cases of P falciparum and P vivax malaria using a Bayesian hierarchical mixed model framework. We quantified the variation explained by mining activity before exploring the modifying effects of environmental degradation on climate-malaria relationships.

Findings: We observed a 27% reduction in the additional unexplained spatial variation in incidence of P falciparum malaria and a 23% reduction in P vivax malaria when mining was included in our models. The effect of temperature on malaria was greater in high mining areas than low mining areas, and the P falciparum malaria effect size at temperatures of 26·5°C (2·4 cases per 1000 people [95% CI 1·78-3·06]) was twice as high as the effect in low mining areas (1 case per 1000 people [0·68-1·49]).

Interpretation: We show that mining activity in southern Venezuela is associated with hotspots of malaria transmission. Increased temperatures exacerbated malaria transmission in mining areas, highlighting the need to consider how environmental degradation modulates climate effect on disease risk, which is especially important in areas subjected to rapidly rising temperatures and land-use change globally. Our findings have implications for the progress towards malaria elimination in the Latin American region. Our findings are also important for effectively targeting timely treatment programmes and vector-control activities in mining areas with high rates of malaria transmission.

Funding: Biotechnology and Biological Sciences Research Council, Royal Society, US National Institutes of Health, and Global Challenges Research Fund.

Translation: For the Spanish translation of the abstract see Supplementary Materials section.

Figure 1:. Geographical context and spatiotemporal trends of malaria incidence in Bolívar 1996–2016

A) Location of the state of Bolívar in southern Venezuela. B) API per 1000 people, log transformed, of Plasmodium falciparum and Plasmodium vivax malaria across Bolívar in 2016. C) API per 1000 people of P falciparum (dark purple) and P vivax (light purple) malaria in the 11 municipalities of Bolívar between 1996 and 2016. Inset maps show locations of each municipality in Bolívar (grey shading). API=annual parasite incidence.

Figure 2:. Environmental and socioeconomic drivers of malaria in Bolívar

(A) Effect size and 95% CIs for spatiotemporal models of Plasmodium falciparum malaria (purple bars) and Plasmodium vivax malaria (blue bars) incidence. The model included an interaction term between levels of mining (high [>2 mines] and low [≤2 mines]) and non-linear functions of temperature and rainfall. The model also included random effects to account for seasonality, interannual variability, and spatial dependency structures. (B) Locations of mining sites (red dots) in Bolívar identified through remote sensing and total number of mining sites per parish. Dark grey colours show parishes with a high number of mining sites and light grey colours represent parishes with few or no mining sites. Labels are shown for the ten parishes with the highest mining activity. (C) Variation in malaria incidence explained by mining activity. Marginal effect (mean and 95% CIs of the spatial random effect) of log API, of spatiotemporal models for P falciparum and P vivax malaria that exclude (blue) and include (red) mining activity across Bolívar as a covariate. A reduction in mean estimate towards zero indicates areas in which mining activity explains the spatial variation in malaria incidence. Estimates are shown for the ten parishes in Bolívar with the highest number of mines. The model also includes linear effects of mining, deforestation, urban development, sea-surface temperature anomalies characteristic of El Niño Southern Oscillation events, an interaction term between high and low levels of mining and non-linear functions of temperature and rainfall, as well as random effects to account for seasonality, interannual variability, and spatial dependency structures. API=annual parasite incidence.

Figure 3:. Combined effect of mining activity and climate variation on malaria risk in Bolívar

Distribution (median, upper, and lower quartiles) of API, log transformed in 2016, of P falciparum and P vivax malaria in areas of Bolívar with low and high mining activity (A). Effect size (API; solid line) and 95% CIs (shading) for the relationship between mean temperature (B) and precipitation (C) and P falciparum and P vivax incidence in areas of low and high mining activity. The model included an interaction term between high and low levels of mining and non-linear functions of temperature and rainfall. The model also included linear effects of mining, deforestation, urban development, sea-surface temperature anomalies characteristic of El Niño Southern Oscillation events, as well as random effects, to account for seasonality, interannual variability, and spatial dependency structures. API=annual parasite incidence.