AI-Analyzed Neighborhood Features Linked to Cardiovascular Events in Veterans

A nationwide study of 777 990 US veterans with atherosclerotic cardiovascular disease (ASCVD) found that specific built environment characteristics extracted from Google Street View images were significantly associated with major adverse cardiovascular events (MACE), including heart attack, stroke, and cardiovascular death.

Using 164 million street-level images and fine-tuned computer vision models, researchers evaluated 11 interpretable features—such as greenery, sidewalks, crosswalks, and housing type—at the census tract level. These were linked to clinical outcomes using multivariable models accounting for demographics, comorbidities, and social deprivation.

Over a median 4.33-year follow-up, several features were independently associated with increased MACE risk: greenery, single-lane roads, sidewalks, crosswalks, non-single-family homes, and the presence of 2 or more cars. In contrast, old or dilapidated buildings and visible utility wiring were associated with decreased risk.

“Built environment features influence MACE risk in US Veterans with ASCVD, emphasizing the role of the exposome in cardiovascular health,” the authors stated.

The association between greenery and MACE was strongest in neighborhoods with higher social deprivation. Crosswalks were protective in rural areas but associated with increased risk in urban ones. These findings suggest context-specific effects of neighborhood features.

This is the first large-scale study to use explainable built environment data to predict MACE in a high-risk population. While causal relationships cannot be inferred, the authors concluded, “These preliminary findings open new areas for future research and interventions targeted towards reducing the burden of ASCVD and improving global public health.”

Reference
Moorthy S, Chen Z, Zhang T, et al. The built environment and adverse cardiovascular events in US veterans with cardiovascular disease. Sci Total Environ. 2025; 980. doi: 10.1016/j.scitotenv.2025.179596.