It could be said that public health professionals are often regarded as doctors for entire communities. Our "treatments" go beyond individual care, shaping the environments and systems that influence how societies live and thrive. As in clinical care, we face critical decisions: Where should we focus our efforts, and which interventions will have the greatest impact? For instance, while we know that well-designed environments—clean, safe sidewalks and accessible walking spaces—promote physical activity, we must ask: Should we prioritize creating more walkable spaces, or motivating people to move more?
Emerging data sources present exciting opportunities to enhance decision-making in public health, urban planning and policy. One powerful tool is Google Street View (GSV) imagery—a vast collection of panoramic, street-level images from over 100 countries. Captured by vehicle-mounted cameras, GSV enables precise analysis of urban environments, using deep learning to map features like sidewalks, parks and public spaces. Studies consistently link walkability and greenspace quality to better health, such as lower rates of cardiovascular disease, obesity and improved mental health.
However, relying solely on superficial correlations—like assuming more parks automatically lead to better health—can be misleading. Misclassified data can further distort reality, leading to flawed interpretations. This can result in poor decisions, such as building parks where they aren't needed or launching physical activity campaigns without the necessary infrastructure. Our recent paper in the Proceedings of the National Academy of Sciences shows how mislabelled data or improper interpretation can result in the wrong interventions. So, what’s the solution?
Public health professionals have a vital role to play. We can guide companies like Google in gathering data that captures key health-relevant factors, accurately labeling features like walkability and greenspace quality, and using advanced statistical models that account for underlying mechanisms, going beyond simple associations. We can guide the construction of models that weigh trade-offs between different interventions, ensuring decisions are grounded in solid evidence rather than solely on data trends that may not capture the full picture.
At the heart of this issue is the need to build technology grounded in deep local knowledge. For technology to truly make an impact, it requires cross-disciplinary collaboration. To accomplish this, public health experts must be equipped with AI skills to ensure that data-driven interventions are both effective and responsive to community needs. By bridging the gap we will ensure technology is not just built, but truly drives people to engage, thrive and benefit, ensuring sustainability and meaningful change.
Rumi Chunara, PhD
Associate Professor
Director, Center for Health Data Science