Article type
Abstract
Background
Global greenhouse gas emissions continue to rise. The resulting warming poses increasingly unacceptable risks to human and planetary health and therefore sustainable development. Already today, climate change severely negatively affects people’s lives around the globe. There are many studies documenting health impacts in relation to climate drivers, but only few formally attribute them to anthropogenic climate changes. As a result, science assessments like those by the Intergovernmental Panel on Climate Change (IPCC) or the Lancet Countdown fail to provide a comprehensive picture of attributable health impacts.
Objectives
•Develop a novel, reproducible, 2-step methodology that combines results from climate impact models with machine-learning-assisted evidence synthesis to perform a post hoc attribution of health impacts attributable to anthropogenic global warming at scale.
•Provide a comprehensive, updatable, living map of attributable human health impacts documented in the scientific literature, identify potential evidence gaps, and evaluate how reliably this map can be extended into the future without major human updates.
Methods
We train and fine-tune the transformer model climateBERT to identify scientific health impact studies sourced from OpenAlex. Relevant studies are classified with regard to the climatic driver (warming, precipitation), the specific impact, and the location. This geo-referenced set of documented climate-related health impacts is then combined with results from a climate model for the attribution of climate drivers to human influence at the grid cell level.
Results
In total, we found about 17,000 studies documenting health impacts that can be attributed to anthropogenic climate changes post hoc – orders of magnitudes more than in any study before. Most of this research on health impacts is related to infectious diseases (41%), cardiorespiratory diseases (15%), and mortality/morbidity (18%). We identify critical evidence gaps in parts of the Global South, where the disease burden is high, but there are comparatively few studies.
Conclusions
Our post hoc attribution methodology provides the first comprehensive picture of climate-related health impacts documented in the scientific literature and is now used as a new indicator in the Lancet Countdown. Our modeling pipeline is fully automated and allows for frequent updating in support of living evidence and annual Countdown updates.
Global greenhouse gas emissions continue to rise. The resulting warming poses increasingly unacceptable risks to human and planetary health and therefore sustainable development. Already today, climate change severely negatively affects people’s lives around the globe. There are many studies documenting health impacts in relation to climate drivers, but only few formally attribute them to anthropogenic climate changes. As a result, science assessments like those by the Intergovernmental Panel on Climate Change (IPCC) or the Lancet Countdown fail to provide a comprehensive picture of attributable health impacts.
Objectives
•Develop a novel, reproducible, 2-step methodology that combines results from climate impact models with machine-learning-assisted evidence synthesis to perform a post hoc attribution of health impacts attributable to anthropogenic global warming at scale.
•Provide a comprehensive, updatable, living map of attributable human health impacts documented in the scientific literature, identify potential evidence gaps, and evaluate how reliably this map can be extended into the future without major human updates.
Methods
We train and fine-tune the transformer model climateBERT to identify scientific health impact studies sourced from OpenAlex. Relevant studies are classified with regard to the climatic driver (warming, precipitation), the specific impact, and the location. This geo-referenced set of documented climate-related health impacts is then combined with results from a climate model for the attribution of climate drivers to human influence at the grid cell level.
Results
In total, we found about 17,000 studies documenting health impacts that can be attributed to anthropogenic climate changes post hoc – orders of magnitudes more than in any study before. Most of this research on health impacts is related to infectious diseases (41%), cardiorespiratory diseases (15%), and mortality/morbidity (18%). We identify critical evidence gaps in parts of the Global South, where the disease burden is high, but there are comparatively few studies.
Conclusions
Our post hoc attribution methodology provides the first comprehensive picture of climate-related health impacts documented in the scientific literature and is now used as a new indicator in the Lancet Countdown. Our modeling pipeline is fully automated and allows for frequent updating in support of living evidence and annual Countdown updates.