PED-PHAM: An Automated and Scalable Spatial Tool That Predicts and Monetizes Health Impacts of the Built, Natural, and Social Environment

点击翻译按钮获取中文摘要

The proposed PEDestrian Public Health Assessment Model (PED-PHAM) uses proven Artificial Intelligence (AI) detection methods to derive pedestrian environment features (PEF), such as sidewalks, crosswalks, and lighting, from digital images. It will add these PEFs to Urban Design 4 Health's existing peer reviewed National- PHAM that predicts health outcomes. PED-PHAM will better enable analysts and decision makers at public planning agencies, consulting firms, land developers, health care providers, lending institutions, and research and big data entities to account for health benefits of modifiable and cost-effective design features known to predict physical activity and BMI. Urban Design 4 Health's and Arizona State University's aims are: 1: Evaluate Degree AI Models Detect PEFs, 2: Create and Optimize Block Group Level AI-derived PEF Indices, and 3: Evaluate PEF enhanced Models. Create PED-PHAM. Physical inactivity is a primary risk factor for obesity, heart disease, stroke, and Type II diabetes. Most adults are inactive. This physical activity (PA) deficiency has not changed meaningfully for the US population in the last two decades. Lack of PA is partially due to hostile pedestrian environments, sedentary car-dependent lifestyles, and sprawling urban environments. Significant relationships have been documented between the built environment, PA, body mass index (BMI), diabetes, and overall cardiometabolic health. Few peer-reviewed evidence-based tools quantify and predict physical activity and health impacts of community-based transportation investments, land use, and community design decisions, and none capture PEFs. We will calculate PEFs for 2,173 participant home locations in two NIH- funded R01 studies, create new objective physical activity and reported BMI models, and integrate the results into the N-PHAM tool. The following steps will be taken [1] Objectively detect PEF presence using trained and validated AI computer vision models applied to Google Street View omnidirectional imagery every 15 meters along roads in census block groups containing 2173 participants in the Baltimore, Phoenix, San Diego, and Seattle regions from two NIH funded studies, [2] Construct block group level metrics for each detected PEF, [3] Optimize block group level PEF indices by translating PEF metrics into summed standardized distributions, [4] Spatially join new PEFs, macro walkability, greenspace, and demographic measures with objectively assessed and self reported PA and BMI into a combined person-level database. [5] Conduct statistical analysis to determine which PEFs (individually and in combined indices) and weights best explain PA and BMI when adjusting for demographics, walkability, and greenspace, and [6] Add new statistical models to the existing N- PHAM platform creating PED-PHAM. Phase 2 will scale PED-PHAM for national application and commercialization and further account for air pollution exposure to create optimized community design place- based prescriptions to increase PA and reduce chronic diseases for user-selected locations across the US.