Background <p>Diabetes remains a major public health concern in the United States, with a disproportionate burden among low-income populations. Despite rapid economic growth in several U.S. states, improvements in population health have not been evenly distributed. Understanding how modifiable behavioral and healthcare access factors correspond to variation in diagnosed diabetes prevalence is important for informing effective public health strategies. This study evaluates how diagnosed diabetes prevalence among low-income adults in high-growth U.S. states may vary under alternative behavioral and healthcare access scenarios.</p> Methods <p>We used Behavioral Risk Factor Surveillance System (BRFSS) data from 2020 to 2024 to analyze diagnosed diabetes prevalence among low-income adults in the ten U.S. states with the highest economic growth. Scenario-based simulations were constructed by modifying key risk factors, including body mass index (BMI), physical activity, smoking status, cost-related barriers to healthcare, and utilization of routine preventive checkups. A conditional generative modeling framework was applied to estimate model-based predicted prevalence under these alternative scenarios while accounting for nonlinear relationships, latent heterogeneity, and survey design features.</p> Results <p>Across all states, lower BMI profiles corresponded to the largest and most consistent reductions in predicted diabetes prevalence, with substantially greater magnitudes than those observed under other scenarios. Higher levels of physical activity were also associated with consistent and statistically significant reductions, although of smaller magnitude. Scenarios characterized by the absence of cost-related barriers to care corresponded to moderate but consistent decreases in predicted prevalence. Similarly, scenarios in which current smoking was eliminated produced uniformly negative but comparatively modest changes. In contrast, increased utilization of preventive checkups corresponded to higher predicted diagnosed diabetes prevalence in several states, consistent with greater detection of previously undiagnosed cases rather than changes in underlying disease risk.</p> Conclusions <p>Within the predictive simulation framework, lower BMI scenarios corresponded to the largest reductions in model-predicted diagnosed diabetes prevalence among the individual covariate modifications examined, whereas healthcare access scenarios primarily influenced predicted diagnosed prevalence through detection-related pathways. These findings highlight the importance of distinguishing between underlying disease risk and diagnostic processes when interpreting population-level patterns in diagnosed diabetes prevalence. Because each scenario modified selected predictors individually while all other observed characteristics were held constant, the estimated differences should be interpreted as model-based responses to isolated covariate modifications rather than realistic changes in the joint distribution of risk factors or causal intervention effects. Such predictive simulations may nevertheless provide useful evidence to inform public health planning and the prioritization of future interventions among low-income populations.</p>