Climate warming is expanding dengue burden in the Americas and Asia

Abstract

Climate change is expected to pose significant threats to public health, particularly including vector-borne diseases. Despite dramatic recent increases in the burden of dengue that many anecdotally connect with climate change, the effect of past and future anthropogenic climate change on dengue remains poorly quantified. To assess the link between climate warming and dengue we assembled a dataset covering 21 countries in Asia and the Americas, and found a nonlinear relationship between temperature and dengue incidence with the largest impact of warming at lower temperatures (below about 20°C), peak incidence at 27.8°C, and subsequent declines at higher temperatures. Using this inferred temperature response, we estimate that historical climate change has increased dengue incidence by 18% (11 - 27%) on average across our study countries, and that future warming could further increase it by 49% (16 - 136%) to 76% (27 - 239%) by mid-century for low or high emissions scenarios, respectively, with some cooler regions projected to see dengue doubling due to warming and other currently hot regions seeing no impact or even small declines. Under the highest emissions scenario, we estimate that 262 million people are currently living in places in these 21 countries where dengue incidence is expected to more than double due to climate change by mid-century. These insights highlight the major impacts of anthropogenic warming on dengue burden across most of its endemic range, providing a foundation for public health planning and the development of strategies to mitigate future risks due to climate change.

Figure 1:. Subnational data on dengue and temperature from 21 countries.

Darker red indicates warmer temperatures and darker blue indicates higher incidence of dengue. White areas within study countries do not have reported dengue cases in the database. Insets show four examples of epidemic dynamics in different countries.

Figure 2:. Effect of temperature on dengue.

(a) Global nonlinear relationship between dengue cases and temperature and (b) the slope of that relationship indicating the marginal effect of temperature on dengue incidence. Main panel regression model fit in black with 95% confidence interval in gray shading. Vertical lines in (a) indicate country mean temperatures, with labels highlighting the coldest and warmest countries as well as the three highest population countries in the sample: Bolivia (BOL), Mexico (MEX), Brazil (BRA), Indonesia (IDN), and Cambodia (KHM). Thin gray lines in (b) represent variations on the main model using alternative specifications. Histogram in (b) shows the distribution of observed monthly temperatures. Model estimates are restricted to the 1st to 99th percentiles of the observed temperature distribution. We anchor the relative log(dengue) at 0 at the population-weighted average monthly temperature in the sample (24.5 °C). Central estimate for the main model is from the full sample, and confidence interval is calculated from bootstrapped regressions.

Figure 3:. Spatial heterogeneity in the dengue-temperature relationship.

Marginal effect of temperature on dengue incidence by (a) continental region, (b) health expenditure tercile, (c) population density tercile, and (d) dengue incidence tercile. Dengue incidence and population density are subnational covariates, and continent and health expenditure are country-level covariates. Lines are the mean estimates and shaded areas are 95% confidence intervals, with estimated marginal effects trimmed to the 1st to 99th percentiles of the observed temperature distribution for each tercile or continental region. Where relevant, confidence intervals were truncated for visibility. Density plots show the distribution of monthly temperatures for each tercile or continental region and inset maps depict the spatial units included in each tercile for the different covariates, with colors matching those of the estimated marginal effects in each panel.

Figure 4:. Climate change has already increased dengue incidence.

(a) Estimated percent of dengue burden attributable to anthropogenic climate warming for administrative units with observed dengue cases during the study period, estimated as the average % change in dengue between observed and counterfactual climate during 1995–2014. (b) Distributions show variation across administrative units within each country. Black tick marks indicate individual unit mean values, and colored points and bars show the mean and 95% CI of the population-weighted average estimate by country. Countries are ordered by current average temperatures with warmer countries to the right. Only administrative units with reported dengue are included in the distributions and country averages. Inset: marginal effect of temperature on dengue from the main model specification, with country-average temperatures indicated with vertical lines matching the colors in the main panel. (c) Impacts of historical warming are largest in cooler countries where current incidence is low, but impacts are also substantial in moderate temperature countries with high current dengue burdens (e.g., Brazil, Honduras, El Salvador, and Nicaragua). Point colors (indicating temperature) are the same in (a) and (b) and point sizes in (c) indicate population size. Line ranges are 95% CIs as in (b).

Figure 5:. Estimated impacts of climate change on dengue incidence for 2040–2059 are widespread and largest in more temperate regions.

(a) Most locations are expected to see an increase in dengue under climate change, with a small fraction of locations projected to experience slight declines due to temperatures exceeding the inferred optimal temperature. Many of the areas where large increases are predicted (shown in dark red), especially in the Americas, are areas with large cities and high population density. Black circles show cities over 5 million in population. Administrative units are colored by the mean percent change in dengue incidence under the high emissions scenario at mid-century (SSP3–7.0 in 2040–2059) compared to current temperatures (1995–2014). (b) Across all future climate scenarios, dengue incidence is predicted to increase for a majority of countries, with the largest increases in cold countries (left panel), and these impacts are estimated to be up to 100% larger for some countries under the highest emissions scenarios compared to the lowest (right panel). Estimated impacts are in percent change from current incidence in comparison between current and future scenarios, and the absolute difference between percentage changes in comparison between between different future scenarios. Countries retain ordering by average temperature, and country colors are consistent across all figures. Points show mean estimates and lines show 95% CIs. Points and confidence intervals are limited to 300% in comparisons between current and future and to 200% in comparisons between future scenarios for visibility. Where mean estimates are limited, points are shown as triangles.