QUANTIFYING THE DISEASE BURDEN ATTRIBUTABLE TO ARSENIC IN GROUNDWATER GLOBALLY (Invited Presentation)

Arsenic is a naturally-occurring element in groundwater that is linked to numerous adverse health outcomes. However, we do not yet know how many deaths globally are attributed to arsenic. Current estimates of mortality from unsafe drinking water, compiled by the United Nations for Sustainable Development Goal (SDG) Indicator 3.9.2, do not account for arsenic contamination. We present an effort to produce globally consistent, national-scale estimates of mortality attributable to arsenic in groundwater.

We train a geospatial machine-learning model (XGBoost) to predict the probability of elevated arsenic at six threshold concentrations from 10 to 250 µg/L, using newly-compiled arsenic measurements from 725,000 wells and springs in 133 countries. We multiply the machine-learning predictions by the percent untreated groundwater used for drinking water and by population density and then combine them with global health data and dose-response relationships to estimate the burden of disease attributable to arsenic. We validate our results against representative surveys of household drinking water quality.

We predict that about 100,000 annual deaths may be attributable to arsenic, and suggest current estimates of mortality from drinking unsafe water (Indicator 3.9.2) are under-predicted by 10% because they do not include arsenic. Most of the arsenic-attributed deaths occur in Asian countries, but many others also have high per-capita rates of arsenic-attributed mortality. The results highlight the need for improved monitoring and mitigation efforts.