Climatic suitability for leishmaniasis at global and European scales

This study utilizes an XGBoost machine learning framework to demonstrate that global and European leishmaniasis risk is driven by the interplay of land use, climate change, and socio-economic inequity, revealing a significant increase in disease prevalence since the 1990s and a northward expansion of visceral leishmaniasis in Europe that underscores the need for enhanced surveillance and targeted support for vulnerable populations.

The Gist

The Big Picture: A Game of "Where's the Sand Fly?"

Imagine Leishmaniasis as a game of hide-and-seek played by a tiny, invisible enemy. The enemy is a parasite, but it needs a "messenger" to deliver it to humans. That messenger is the sand fly (a tiny insect). If the sand fly can't survive in a certain place, the game can't start there.

This paper is like a super-smart detective (using a computer program called XGBoost) that tries to predict where this game is most likely to happen. The detective looks at three main clues to solve the mystery:

1. The Weather: Is it too hot, too cold, too dry, or just right?
2. The Landscape: Is there a forest, a shrub, or a city?
3. The People: Is the area poor or wealthy? Does everyone have access to good healthcare?

The Detective's Toolkit: How They Did It

The researchers built two different "maps" to solve the puzzle:

• The Global Map: Looking at the whole world (like a globe).
• The European Map: Looking closely at Europe (like a detailed street map).

They fed the computer thousands of years of data about temperature, rain, soil, and land use. They also looked at social data, like how many people live in poverty or lack basic services. The computer learned to spot patterns: "Ah, whenever there are shrubs, it's cold in winter but humid in summer, and people are struggling financially, that's where the sand flies and the disease show up."

The Big Discoveries

1. The "Shrub" Factor is King

The most important clue wasn't the temperature or the poverty; it was land use. Specifically, shrubs and forests.

• The Analogy: Think of shrubs and forests as the sand fly's "five-star hotel." These places provide shade (to hide from the sun), moisture (to keep their eggs from drying out), and plenty of snacks (animals like rabbits and rodents that the flies bite). If you have shrubs, you likely have sand flies.

2. The Weather Matters (But Not Just "Hot")

The computer found that sand flies are picky about the weather.

• The Cold: They need a winter that isn't too freezing (measured by the "mean temperature of the coldest quarter").
• The Dry: They need humidity during the hottest or driest times of the year. If it gets too hot and dry, the flies can't survive.
• The Shift: As the climate changes, the "hotel" is moving. Some flies that like warmth are moving north, while flies that like cool, damp places are shrinking back.

3. Poverty is a Powerful Predictor

The study confirmed that this disease is a "disease of inequity."

• The Analogy: Even if the weather is perfect for sand flies, the disease often only takes hold where people are vulnerable. Think of poverty, poor housing, and lack of healthcare as "open doors" that let the disease in. In Europe, this is linked to refugees and migrants who might be living in crowded or temporary conditions.

What Changed Over Time?

The detective looked at how the "game" has changed from the 1990s to today (2024).

• Globally: The risk of the disease has gone up by about 17%. The biggest jumps are happening in Africa, Asia, and the Americas.
• In Europe: The overall risk went up only a tiny bit (about 1-2%), but the location of the risk changed dramatically.
  • Visceral Leishmaniasis (the severe kind): It is slowly moving north. Countries like Finland and Sweden are seeing higher risks, though the absolute numbers are still low.
  • Cutaneous Leishmaniasis (the skin kind): It is getting worse in Southern and Eastern Europe (like Spain, France, Italy, and the Balkans). Some specific areas saw their risk double or triple.

The Sand Fly Migration

The study also tracked the specific types of sand flies in Europe:

• The Warm-Adapters: Species that like heat (like P. ariasi and P. perniciosus) are expanding their territory, moving into new areas.
• The Cool-Lovers: Species that like cooler, damper weather (like P. perfiliewi) are shrinking back because Southern Europe is getting too hot and dry for them.

The Limitations (What the Detective Couldn't See)

The paper admits a few things the detective couldn't figure out perfectly:

• Imported Cases: Sometimes people get sick after traveling to another country and bringing the disease back. The model can't always tell if a case started locally or was "imported" like a souvenir.
• Data Gaps: In some parts of the world (especially Africa), we don't have enough data to be 100% sure of the numbers.
• Missing Pieces: The model didn't have perfect data on how many dogs or rabbits live in an area, even though these animals are the main "food" for the flies.

The Bottom Line

The paper concludes that climate change, land use (shrubs), and social inequality are working together to change where this disease lives.

• The Warning: The "safe zones" are shrinking. The disease is moving north in Europe and spreading in other parts of the world.
• The Solution: We need better surveillance (watching the sand flies), better management of the environment, and special attention to vulnerable groups like refugees and the poor, because they are the ones most likely to get caught in the game.

Note: The paper emphasizes that this is a research study to understand risk patterns, not a guide for immediate medical treatment.

Technical Summary

Problem Statement
Leishmaniasis, a climate-sensitive zoonotic neglected tropical disease transmitted by Phlebotomine sand flies, poses a significant global health burden, particularly affecting marginalized and displaced populations. While the disease is endemic in 99 countries, its spatio-temporal dynamics are increasingly influenced by environmental changes, land use shifts, and socio-economic inequities. Existing surveillance data is often sparse, inconsistent, or lacks the resolution necessary to identify emerging foci, particularly in Europe where the disease is considered an emerging threat due to climate change. There is a critical need for a robust modeling framework that integrates climatic suitability, land cover, and socio-economic determinants to map current risks and project future trends at both global and regional scales.

Methodology
The study employs a machine learning framework using the Extreme Gradient Boosting (XGBoost) algorithm to model the distribution of leishmaniasis and its vectors. The analysis was conducted at two spatial scales:

• Global Scale: Modeled as a combined Cutaneous (CL) and Visceral Leishmaniasis (VL) outcome at the national level. Due to the lack of global sand fly datasets, disease presence was used as a proxy for vector presence.
• European Scale: Modeled separately for CL and VL at the NUTS2 (sub-national) level. Additionally, five proven vector species (P. perniciosus, P. ariasi, P. perfiliewi, P. neglectus, P. tobbi) were modeled individually.

Data and Features:

• Outcome Data: Leishmaniasis case data (1970–2012) from Pigott et al. (2014) and WHO Global Health Observatory data for validation. European sand fly presence data (2015–2024) were sourced from ECDC/VectorNet.
• Predictors: The models incorporated 27 climate covariates (derived from ERA5 re-analysis and converted to WorldClim bioclimatic indicators, including temperature, precipitation, humidity, and soil water), land cover classes (from Copernicus Global Land Cover), elevation, and socio-economic indices (Human Development Index globally; At-Risk-of-Poverty and Social Exclusion [AROPE] in Europe).
• Modeling Process: Preliminary models identified key features using SHAP (Shapley Additive Explanations) values to ensure interpretability. Correlated features (r > 0.7) were clustered to prevent overfitting. The final models were trained on a 75:25 split of training/testing data, with hyperparameter tuning via grid search. Performance was assessed using AUC, sensitivity, specificity, and RMSE.

Key Results

• Predictive Performance: All models demonstrated high predictive accuracy (AUC > 0.95 for most European models; 0.903 for the global model). The inclusion of socio-economic predictors consistently improved model performance, confirming the role of poverty and inequity as determinants of disease distribution.
• Feature Importance: Land use features, particularly shrubland and forest cover, exhibited the greatest explanatory power across both scales, reflecting their role in providing microclimates and vertebrate hosts. Climatic factors, specifically the mean temperature of the coldest quarter and humidity in the warmest/driest quarters, were also highly influential.
• Global Trends: Since the 1990s, global leishmaniasis risk has increased by approximately 17%. The most significant rises were projected for Africa, Asia, and the Americas.
• European Trends: While continental-scale increases were modest (CL +1.28%; VL +2.47%), the risk is highly heterogeneous.
  • Visceral Leishmaniasis (VL): Shows a northward expansion, with significant risk increases in Nordic countries (e.g., Finland +41.7%, Sweden +58.2%), though absolute risk remains low.
  • Cutaneous Leishmaniasis (CL): Risk is increasing in southern and eastern Europe, with some NUTS2 regions in southern Spain, France, Italy, and the Balkans seeing >100% increases.
• Vector Dynamics: Projections indicate an expansion of warm-adapted sand fly species (P. ariasi, P. perniciosus, P. neglectus) and a contraction of species preferring cooler, humid niches (P. perfiliewi, P. tobbi).

Significance and Claims
The paper claims to provide a novel contribution by mapping leishmaniasis risk at both global and European scales using a unified machine learning approach that integrates climatic, land use, and socio-economic drivers. The study highlights that leishmaniasis is not solely a climatic issue but is deeply intertwined with socio-economic inequity, even in developed regions like Europe.

The authors assert that these findings underscore the need for:

1. Enhanced Surveillance: Particularly in emerging regions and for specific vector species.
2. Integrated Vector Management: Targeting the interaction between climate change, land use, and vector habitats.
3. Targeted Support: Focusing on vulnerable populations, including refugees and migrants, who are disproportionately affected by the socio-economic determinants of the disease.

The paper concludes that without coordinated efforts to address data gaps, surveillance limitations, and socio-economic vulnerabilities, the expansion of climatic suitability for sand flies will continue to place more people at risk of a preventable yet potentially disfiguring and deadly disease. The authors explicitly state that the preprint has not been peer-reviewed and should not be used to guide clinical practice.