A living systematic map of human health impacts attributable to climate change

Session Type
Living syntheses and prospective meta-analyses
Minx J1, Callaghan M1, Lück S1, Repke T1, Schleussner C2
1Mercator Research Institute on Global Commons and Climate Change, Germany
2Climate Analytics, Germany

Background: Impacts of climate changes on human health continue to spread and intensify globally. However, efforts to identify and quantify health impacts that are attributable to climate change remain limited. One fundamental problem is that studies documenting health impacts of climate drivers often fail to perform a formal attribution of these to human interference.
Objectives: We develop a novel methodology rooted in artificial intelligence and impact attribution to produce a comprehensive, fully-automated living evidence map of health impacts attributable to anthropogenic global warming.
Methods:Natural language processing classifiers are trained using a large-language model (ClimateBERT) and applied to identify and classify scientific health impact studies with regard to the climatic driver, the specific impact, as well as the location. This geo-referenced set of documented climate-related health impacts is then combined with results of a modelling pipeline of human-attributable changes in climate drivers on the grid-cell-level.
Results: Attributable health impacts from climate change are found across the globe. In total, we find 25,804 studies documenting health impacts that are linked to climate drivers. Of these, 52% are at least broadly attributable to climate change. Most research on health impacts is related to infectious diseases (38%), cardiorespiratory diseases (12%) as well as mortality/morbity (12%). We identify critical evidence gaps, provide a full validation of our machine learning pipeline and reflect on implications for protocol and guideline development. We develop a new set of indicators that can be used in upcoming science assessment such as the Lancet Countdown on Health and Climate Change.
Conclusions: Our study highlights the value of our new approach combining automated systematic mapping approaches with results from climate change attribution modelling to provide a comprehensive living picture of how climate change is impacting human health today. Our system can be updated daily basis and made available publicly on an interactive website. Such an approach is ideally suited to inform scientific assessments such as the IPCC or the Lancet Countdown and a freely available living evidence map would greatly benefit dedicated systematic review efforts in this important field.
Patient, public and/or healthcare consumer involvement: None