The study, led by Emmanuella A. Asiedu of the University of Miami’s Miller School of Medicine and UMass Chan Medical School, together with Jennifer Tjia, Jonggyu Baek, and Sarah N. Forrester of UMass Chan, focused on adults with electronically confirmed type 2 diabetes mellitus. The team used version 7 of the All of Us Registered Tier Dataset, covering the years 2017 through 2023, and restricted their analysis to participants who self-identified as either non-Hispanic Black or non-Hispanic White. That restriction was deliberate. Racial disparities in diabetes complications are well documented, and the researchers wanted to test whether the neighborhood environment modifies those disparities rather than simply averaging over them.

To quantify the built environment, the team turned to the Physical Activity Neighborhood Environment Scale, known as PANES, an instrument that scores neighborhoods from zero to eight based on features that support active living. A high-scoring neighborhood, in PANES terms, is one where homes are close to shops and services, where public transit is accessible, where sidewalks and bicycle infrastructure exist, and where parks and recreation facilities are within easy reach. The researchers grouped participants into three categories: those living in the least favorable environments with scores of zero to two, those in moderately favorable environments scoring three to five, and those in the most favorable environments scoring six to eight. This categorization allowed the team to compare clinical outcomes across a gradient of neighborhood walkability rather than treating the environment as a simple yes-or-no variable.

The clinical outcomes were drawn from linked electronic health records and represent the standard panel of measures that clinicians use to judge whether diabetes is under control. The primary outcome was high hemoglobin A1c, the blood test that reflects average glucose levels over the preceding two to three months and serves as the single most important indicator of diabetes management. The secondary outcomes included reduced estimated glomerular filtration rate, a marker of kidney function that declines as diabetic nephropathy advances; elevated low-density lipoprotein cholesterol, the lipid fraction most tightly linked to cardiovascular risk; and a set of EHR-documented conditions known to be related to type 2 diabetes. Psychological distress, defined as a score of five or higher on the Patient Health Questionnaire-9 or an electronic health record diagnosis of depression, was included as a potential modifier, since depression is known to worsen diabetes self-management and outcomes.

The headline number from the study is sobering: 89 percent of the study population had at least one adverse type 2 diabetes-related clinical outcome. That figure alone underscores how difficult it is to keep diabetes fully controlled even within a healthcare system, and it frames the central question of the research—if nearly everyone is struggling, what role does the environment around them play in determining who struggles most?

The answer, at first glance, was surprisingly muted. When the researchers ran modified Poisson regression models with generalized estimating equations, adjusting for sociodemographic factors, the crude and adjusted associations between PANES scores and the diabetes-related outcomes were not statistically significant across the population as a whole. In other words, living in a walkable neighborhood with parks and transit did not, on average, predict better A1c control, better kidney function, or better cholesterol levels compared with living in a less favorable environment. For advocates of walkable urban design, this null result might seem discouraging, but the story does not end there.

The interaction analyses revealed something the main models had obscured. When the researchers tested whether the relationship between the built environment and clinical outcomes differed by racial group, they found a statistically significant interaction for LDL cholesterol. Black individuals living in neighborhoods with the least favorable built attributes had a 43 percent greater prevalence of high LDL cholesterol compared with White individuals living in neighborhoods with the most favorable built attributes, with an adjusted prevalence ratio of 1.43 and a 95 percent confidence interval of 1.03 to 1.98. The interaction term p-value of 0.01 indicates that this difference is unlikely to be a statistical fluke. The effect modification by psychological distress, by contrast, did not yield significant findings in the reported results.

What does this mean mechanistically? The authors’ interpretation, grounded in their data rather than speculation, is that neighborhood contextual factors play a role in shaping cardiometabolic risk and that built environment inequities may exacerbate racial health disparities. The biological plausibility is straightforward. Neighborhoods lacking sidewalks, safe public spaces, and recreational facilities reduce opportunities for routine physical activity, which is one of the most effective non-pharmacological tools for improving lipid profiles and glucose control. Residents of such neighborhoods may also face longer exposure to stressors, fewer options for purchasing fresh food, and greater difficulty reaching healthcare appointments, all of which can compound the metabolic burden of diabetes. The fact that the disparity appeared specifically for LDL cholesterol—a marker strongly influenced by both diet and physical activity—fits this framework, though the cross-sectional design of the study means the data cannot establish that the environment caused the worse outcomes, only that the two are associated in a pattern consistent with that mechanism.

The methodological strengths of the study deserve attention. By using EHR-confirmed diabetes rather than self-report, the researchers avoided misclassification of the disease status. By linking standardized Observational Medical Outcomes Partnership-format records from the All of Us program, they accessed clinical measurements across a geographically diverse and demographically broad population, a sample far more representative than the single-health-system cohorts that dominate much of this literature. The use of modified Poisson regression with robust variance estimation is appropriate for binary outcomes with high prevalence, since standard logistic regression in such settings can produce prevalence ratios that overstate effect sizes. The formal interaction testing, rather than stratified analysis alone, gives the racial disparity finding a firmer statistical footing than a simple subgroup comparison would.

The limitations are equally worth stating plainly. The cross-sectional design captures a single snapshot, so reverse causation and time-varying confounding cannot be excluded. PANES scores, while validated, are a coarse summary of neighborhood quality and may miss features such as traffic safety, food access, or social cohesion that also shape health. The restriction to non-Hispanic Black and non-Hispanic White participants means the findings cannot be generalized to other racial and ethnic groups, including Hispanic and Asian populations who carry substantial diabetes burden. And the null overall association raises the possibility that neighborhood effects on diabetes outcomes operate through pathways or at spatial scales that PANES does not capture, or that individual-level factors such as medication access and healthcare quality dominate within this cohort.

Even with those caveats, the implications are concrete. The authors conclude that urban planning efforts prioritizing walkability, access to recreation, and safe public spaces—especially in underserved communities—are needed to support more equitable chronic disease outcomes. In practical terms, that means the tools for reducing diabetes complications may lie partly outside the clinic, in zoning decisions, transit investments, and park development that determine what a resident’s daily environment makes easy or hard. The 43 percent excess prevalence of high LDL cholesterol among Black residents of the least favorable neighborhoods is a quantified signal that the geography of diabetes is not evenly distributed, and that closing racial health gaps may require treating the neighborhood itself as a site of clinical intervention. As datasets like All of Us continue to grow and link environmental, genomic, and clinical data at scale, studies of this kind will increasingly be able to test whether reshaping the built environment actually reshapes the trajectory of chronic disease—a question with profound consequences for the millions of Americans living with diabetes in neighborhoods that were never designed with their health in mind.

Subject of Research: Association between neighborhood built-environment attributes and type 2 diabetes-related clinical outcomes, with effect modification by race

Article Title: Characterizing the association of neighborhood built-environment and type 2 diabetes-related clinical health outcomes: insights from the All of Us Research Program

Article References: Asiedu, E. A., Tjia, J., Baek, J., & Forrester, S. N. (2026). Characterizing the association of neighborhood built-environment and type 2 diabetes-related clinical health outcomes: insights from the All of Us Research Program. BMC Public Health. https://doi.org/10.1186/s12889-026-29830-8

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29830-8

Keywords: type 2 diabetes, built environment, neighborhood walkability, All of Us Research Program, health disparities, LDL cholesterol, HbA1c, PANES, electronic health records, public health, urban planning, racial disparities