These boundaries are subjective, and are adjudicated by expert opinion. Users should read studies closely, and develop their own assessment of study methodology.

Criteria for inclusion

This database includes studies that conduct end-to-end detection and attribution of climate change impacts on human health, meaning:

• a statistical analysis that quantifies health impacts or risks

• resulting from anthropogenic (human-caused) forcings on the climate

• through the comparison of factual and counterfactual scenarios.

This includes a number of different study designs, with different levels of stringency in terms of both climate science and epidemiology.

We exclude studies that estimate the impacts of observed climate change based on comparison of time points (i.e., pre- and post-human influence on the climate).

Data collection

An initial round of studies were collected as part of a systematic literature review (see Wellcome Trust report; preprint forthcoming). New eligible studies are located through Google Scholar alerts, periodic searches of preprint servers (particularly bioRxiv, medRxiv, and ResearchSquare), and word of mouth.

Data definitions

Timescale

• Event studies focus on the impacts of a specific extreme weather event (e.g., a specific hurricane, a wildfire, or a heat wave), or several events.

• Trend studies focus on long-term impacts of shifts in temperature, precipitation, or other climate variables. Some studies are both trend and event focused, looking at long-term trends in the cumulative impacts of, or shifts in the frequency of, extreme events .

Attribution method

• A full end-to-end analysis connects simulated factual and counterfactual climates, weather patterns, or event characteristics to simulated health impacts.

• The fraction of attributable risk (FAR) approach estimates impacts by multiplying observed impacts of an event by an independent estimate of the fraction of attributable risk of a given weather event due to anthropogenic influence.

Counterfactual method

• Some studies simulate factual and counterfactual climate and scenarios using climate models with different combinations of natural and anthropogenic forcings. (In some contexts, this may be more rigorous than detrending.)

• Some studies detrend observed weather or climate data, on the assumption that trends are largely or exclusively the result of human influence.

• Some studies following the FAR approach may rely on an external study’s analysis, and so never design their own counterfactual.

Health impacts

• Some studies use observed data on health impacts to develop statistical analyses, producing estimates that directly correspond to real-world observations.

• Some studies extrapolate health impacts based on independent estimates of the health outcome-climate relationship (sometimes called a “transfer function”); this approach allows for more divergence between modelled and real impacts.

• Some studies use health impact data that are partially extrapolated: for example, observational data on all-cause mortality may be converted into estimated heat-related mortality through a transfer function, and the attribution analysis is then conducted on those estimates.