**Background:** People with serious mental illness (SMI) die approximately 10–25 years earlier than the general population, largely due to elevated rates of obesity, diabetes, cardiovascular disease, and cancer. Blacks and Hispanics with SMI face compounded health inequities. While individual and interpersonal determinants of health behaviors are well studied, the role of place-based neighborhood factors (e.g., food environment, walkability, recreational facilities) remains largely unexplored for people with SMI. This study used baseline data from the Peer-Led Group Lifestyle Balance (PGLB) trial—a multisite effectiveness trial testing a peer-led healthy lifestyle intervention for overweight/obese adults with SMI living in supportive housing in New York City and Philadelphia—to demonstrate how place-based factors can be examined within intervention trials.
**Methods:** This cross-sectional ecological analysis integrated zip-code-level neighborhood data from multiple public and proprietary geospatial sources (e.g., US Census American Community Survey, city open data portals, National Neighborhood Data Archive) with individual-level baseline data from 314 PGLB trial participants (190 in Philadelphia, 124 in New York City). The analytic sample included 243 zip codes (197 in NYC, 46 in Philadelphia), of which 42 contained PGLB participants (17 in NYC, 25 in Philadelphia). Complete dietary and physical activity data were available for 292 participants. Neighborhood measures covered six domains: social environment (e.g., population density, poverty, racial composition, housing type), urban design (walkability index, land-use mix, road-network connectivity, transit stop density), healthcare infrastructure (hospital and pharmacy density), legal drug environment (tobacco and liquor store density), physical activity environment (park density, recreation centers, YMCAs, private fitness facilities, sidewalk and bicycle lane coverage), and food environment (supermarket, bodega/convenience store, and fast-food restaurant density). Individual-level outcomes included fruit/vegetable intake (dichotomized as meeting vs. not meeting USDA guideline of ≥5 portions/day), daily sugar-sweetened beverage consumption (any vs. none), and physical activity (meeting vs. not meeting 150 min/week of MVPA via walking and via non-walking MVPA, measured by IPAQ). Fisher's Exact tests and chi-square tests were used to compare neighborhood characteristics between cities and between PGLB vs. non-PGLB neighborhoods, and to examine associations between place-based factors and baseline health behaviors.
**Key Results:** Within cities, PGLB neighborhoods were significantly more likely to have >50% of residents living in poverty compared to other neighborhoods. In Philadelphia, PGLB neighborhoods had significantly fewer private fitness facilities and lower hospital density than non-PGLB neighborhoods; these disparities were not observed in NYC. Between cities, significant differences were found across most social environment variables (population density, age distribution, racial composition, poverty, housing type, vacancy rates, rental rates), urban design (walkability, land-use mix, transit access), healthcare (hospital and pharmacy density), and several physical activity environment variables (park access, sidewalk and bicycle lane coverage). For dietary behaviors, insufficient fruit/vegetable intake (<5 portions/day) was significantly higher among PGLB participants residing in neighborhoods with high educational attainment, high connectivity and walkability, and high bicycle lane coverage. No food environment characteristics were associated with fruit/vegetable intake. SSB intake did not differ significantly by any place-based factor. For physical activity, meeting guidelines via walking did not vary by any place-based factor. However, not meeting guidelines via non-walking MVPA was significantly associated with high population density, apartment housing, high housing vacancy, high renter-occupancy, high road-network connectivity, high walkability, high transit access, sidewalk and bicycle lane availability, high hospital access, and low tobacco retail access.
**Clinical Implications:** The findings demonstrate that neighborhood environments differ significantly both within and between cities where lifestyle interventions are conducted, and these differences may moderate intervention effectiveness. The counterintuitive finding that participants in highly walkable neighborhoods had lower fruit/vegetable intake and lower non-walking MVPA suggests that environmental assets (e.g., walkability, parks) may not translate to healthier behaviors if they are unaffordable, unsafe, or poorly maintained—a potential "house-poor" effect. The null findings for walking-based physical activity likely reflect that both NYC and Philadelphia are already highly walkable, limiting variability. The site-level differences in PGLB outcomes (NYC showed significant weight loss and CVD risk reduction; Philadelphia sites showed null findings) may be partially explained by between-city differences in neighborhood environments. The authors recommend that future behavioral interventions targeting place-dependent behaviors should be powered and designed to assess moderation by place-based factors, use spatially stratified sampling or propensity score matching, and account for city-level clustering in multisite trials. For policymakers and urban planners, the study underscores that people with SMI—an often-overlooked population—should be considered in "health in all policies" approaches and urban design initiatives aimed at creating healthy, equitable cities.