Impact of Commercial Food Environments on Local Type 2 Diabetes Burden: Cross-Sectional and Ecological Multimodeling Study (Preprint)

<sec> <title>BACKGROUND</title> Neighborhoods resulting from rapid urbanization processes are often saturated with eateries for local communities, potentially increasing exposure to unhealthy foods and creating diabetogenic residential habitats. </sec> <sec> <title>OBJECTIVE</title> We examined the association between proximity of commercial food outlets to local neighborhood residences and type 2 diabetes (T2D) cases to explore how local T2D rates vary by location and provide policy-driven metrics to monitor food outlet density as a potential control for high local T2D rates. </sec> <sec> <title>METHODS</title> This cross-sectional ecological study included 11,354 patients with active T2D aged ≥20 years geocoded using approximate neighborhood residence aggregated to area-level rates and counts by subdistricts (&lt;i&gt;mukims&lt;/i&gt;) in Penang, northern Malaysia. We used the National Diabetes Registry complemented with data from medical records across 29 primary care clinics throughout the state. Food establishment data were retrieved from the Open Data Portal sourced through the Penang GeoHub, and urbanization indicators were retrieved from MyCensus 2020. We executed point-level proximity- and density-based area-level analysis through multimodel aspatial and spatial regression methods. </sec> <sec> <title>RESULTS</title> Our final hierarchical linear regression revealed that the distance to food complexes, hawker markets, &lt;i&gt;kopitiams&lt;/i&gt; (a type of coffee shop), 24-7 convenience stores, fast food outlets, and public markets showed statistically significant associations (&lt;i&gt;P&lt;/i&gt;&amp;lt;.05) with the age and BMI of patients with T2D. In the multiscale geographically weighted regression model, the adjusted &lt;i&gt;R&lt;/i&gt;&lt;sup&gt;2&lt;/sup&gt; values ranged from 0.15 to 0.62, with lower values observed across the mainland. The multiscale geographically weighted regression model yielded average β coefficients for densities of &lt;i&gt;kopitiams&lt;/i&gt; (β=0.256), fast food outlets (β=−0.061), 24-7 convenience stores (β=0.028), supermarkets (β=0.122), public markets (β=0.067), and &lt;i&gt;nasi kandar&lt;/i&gt; (a type of rice dish) restaurants (β=−0.064), urban growth rate (β=0.189), and population density (β=−0.080; t&lt;sub&gt;65.835&lt;/sub&gt;≥1.96 in all cases). We established population-attributable fractions suggesting that, if local neighborhoods underwent township restructuring to remove food complexes, hawker markets, or &lt;i&gt;kopitiams&lt;/i&gt;, an estimated reduction of 0.21%, 0.27%, and 0.09%, respectively, in the risk of T2D cases in Penang would be anticipated. However, if local neighborhoods underwent township restructuring to add hawker complexes, &lt;i&gt;nasi kandar&lt;/i&gt; restaurants, fast food outlets, 24-7 convenience stores, public markets, or supermarkets, an estimated reduction of between 0.07% and 0.64% in the number of residents with risk of T2D was estimated. </sec> <sec> <title>CONCLUSIONS</title> The reported variations provide insights into the associations between high neighborhood T2D rates and the density of a range of food outlets. We observed that these associations varied by place, providing insight into potential monitoring and policy considerations. This work provides evidence for interpretation at the individual and aggregate levels, shifting public health interventions from a generic to a targeted approach and prompting township planners to restructure food outlet accessibility or availability in local neighborhoods and to develop health behavior interventions for local communities for healthy food purchase and consumption. </sec>