Google's PDFM location embeddings improve five global public health tasks

Key points

• Mount Sinai and Boston Children's Hospital raised MMR vaccination variance explained from 16% to 22% using Canadian cross-border embeddings.
• NYU Grossman researchers found PDFM matched census-based models on 2023 county cardiovascular death estimates, and it covers 17 countries versus ACS's U.S.-only data.
• With TimesFM 2.0, PDFM improved one-month dengue forecasts in up to 72% of active transmission municipalities in Mexico, per the partner researchers.
• University of Washington researchers reported PDFM improved postpartum depression risk prediction in states the model had not seen during training.

Why it matters for marketers: Teams that rely on census or lagged surveillance data can test PDFM as a fresher, drop-in location input for geographic targeting and forecasting.