SRF045: Neighborhood Assessment at Social Distance: Measuring Street “Exposures” to Food, Alcohol, & Tobacco Using Google Street View
Sean Lucan, MD, MPH, MS, MPH, MS; Clyde Schechter, MD, MA, FACPM; Jason Gilbert; Sara Glickman; Michael Yang; Charles Pan
CONTEXT: Unhealthful consumable products—ultraprocessed foods, alcohol, & tobacco—are linked to chronic disease, early death, & health disparities. “Exposures” to such products may vary in patient neighborhoods. Neighborhood assessments could provide insights. In a COVID era, with unhealthful consumption increasing, neighborhood assessments may require social distance. OBJECTIVE: To develop & use a tool to assess neighborhoods remotely via Google Street View (GSV), a freely available, web-based archive of omnidirectional street images. HUMAN SUBJECTS REVIEW: Exempt. DESIGN: GSV scan of streets for unhealthful-product “exposures”: advertisements (cues to consume; e.g., posters, billboards) & availability (opportunities to obtain; e.g., fast-food outlets, liquor stores, tobacco shops). “Exposures” will later be linked to neighborhood demographics. SETTING: Incremental buffers (0-100, 100-200, 200-300ft) along streets around all subway stations (N=68) across the Bronx, NY. OUTCOME MEASURES: Inter-rater reliability of assessment tool using Cohen’s Kappa (1=perfect agreement, 0=agreement no better than by chance); differences in unhealthful-product “exposures” by station neighborhood. ANTICIPATED RESULTS: Using a first-draft tool to assess streets around a pilot sample of 8 subway stations, agreement about the presence of ads & availability was initially poor-to-moderate (kappas 0.07-0.64). Reasons for inter-rater discrepancy included differences in the following: image dates in GSV; demarcation lines for buffers; inclusion of 2nd-story businesses; how to count street vendors; how to count businesses with closed riot doors; definitions of food/drink categories. After modifying the tool, a later pilot around 5 additional stations resulted in kappas 0.50-1.0 (lower values were driven by rare events). With this refined tool, anticipated future results: unhealthful-product “exposures” will cluster around subway stations (higher counts & proportions 0-100ft > 100-200ft > 200-300ft); “exposures” will be greater (higher counts & proportions) in neighborhoods home to vulnerable populations (e.g., Black & Hispanic communities, higher poverty). CONCLUSIONS: GSV may be a useful tool for neighborhood assessments at social distance. For this project, GSV is likely to show ads & availability for unhealthful products clustering around subway stations—particularly in vulnerable communities. Findings will be relevant to counseling patients and to suggesting community advocacy.