Improving the precision of public health interventions
In Malawi, Google.org recipient Cooper/Smith Earth AI combined PDFM and AlphaEarth satellite embeddings with local data to predict the use of health services at local clinics. This can help decision makers detect early warning signs of disease outbreaks and allocate limited resources more efficiently.
To combat the rise in measles, researchers at Mount Sinai and Boston Children’s Hospital/Harvard Earth used AI’s PDFM to fill in gaps and produce “super-resolution” estimates of vaccination coverage. Based on aggregated data that protects privacy, researchers can map vaccination rates down to the ZIP code level without revealing sensitive personal information and identify localized clusters of undervaccination consistent with recent outbreaks.
Forecasts for diseases where weather and geography matter
The weather affects the pace of many diseases, and specific weather patterns can signal health crises. For example, summer rains can cause dengue fever to rise, while floods can significantly increase cholera outbreaks. Combining population dynamics with predictive weather models helps improve forecasting of health emergencies weeks or months in advance.
In collaboration with the WHO Regional Office for Africa, we evaluated a subnational cholera case forecasting model using WHO centralized integrated disease surveillance data. We found that by combining Google’s TimesFM time series model with PDFM and weather data, we were able to improve the forecasting accuracy of cholera cases by over 35% compared to standard models. Better forecasting could enable public health officials to plan proactively rather than react after a crisis—for example, moving life-saving rehydration supplies to where they will be needed.
Furthermore, researchers at the University of Oxford have successfully used Earth AI models and datasets to improve dengue forecasting in Brazil. Including PDFM embeddings significantly increased the predictive accuracy of six-month forecasts, giving local authorities more time to implement preventive measures.
Understanding the needs of chronic diseases
Earth AI also unlocks critical insights into non-communicable diseases. In a recent initiative in Australia, we partnered with the Victor Chang Cardiac Research Institute, Wesfarmers Health and Latrobe Health Services to implement Population Health AI (PHAI). Currently available as a proof-of-concept to select partners, PHAI uses Earth AI’s PDFM embeddings along with other key datasets such as air quality, pollen and site insights to uncover the health needs of communities in rural Australia, with the aim of supporting their chronic disease needs and prevention efforts.
A proactive, healthier future
Technology is most powerful when it leads to action in the real world. By fusing the planetary intelligence of Google Earth AI with the deep health expertise of our partners, we are moving toward a future goal where health systems everywhere possess the data-driven insights needed to protect and improve public health.