A Geographically Weighted Random Forest Analysis of Spatial Non-Stationarity Association of Street View Environmental, Socioeconomic, and Lifestyle Factors with Type-2 Diabetes Prevalence in Toronto

A Geographically Weighted Random Forest Analysis of Spatial Non-Stationarity Association of Street View Environmental, Socioeconomic, and Lifestyle Factors with Type-2 Diabetes Prevalence in Toronto | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Geographically Weighted Random Forest Analysis of Spatial Non-Stationarity Association of Street View Environmental, Socioeconomic, and Lifestyle Factors with Type-2 Diabetes Prevalence in Toronto Haoxuan Ge, Jue Wang, Devin Yongzhao Wu, Hanlin Zhou This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5984185/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Numerous studies have shown that the environment can affect the prevalence of Type-2 Diabetes Mellitus (T2DM) by encouraging healthy lifestyle behaviours. Alongside traditional medical prevention strategies, social determinants have also been prioritized as a top consideration in T2DM clinical treatment and care. However, limited research has explored the association between neighbourhood perceptions of aesthetics and safety and T2DM prevalence, potentially through indirect pathways influencing behavioural response. Combining the effects, this research has two main objectives: (1) to identify the relationships between street view environmental, socioeconomic, and lifestyle factors with T2DM prevalence rates and (2) to determine how these associations vary spatially across different regions of Toronto. Methods This study applied a Geographically Weighted Random Forest regression to analyze the spatial non-stationarity associations and account for potential confounding factors in the relationship between 27 variables with T2DM prevalence across Toronto’s 3,800 Dissemination Areas. After modelling, the study examined local variations in feature effects using partial dependency plots and permutation-based feature importance maps to assess how variable associations on T2DM prevalence various around Toronto. Results The model achieved an R 2 of 77%. Having regular healthcare, age, smoking rate, and obesity prevalence have the strongest positive correlation with T2DM prevalence. Beauty and safety perception, NDVI, and mental issues have a weak positive association with T2DM prevalence. In the downtown financial districts, immigration rates and drinking rates were identified as negatively associated with T2DM prevalence. Meanwhile, marital status, obesity, life dissatisfaction, and commuting by walking or cycling were found to have positive or negative spatial non-stationary associations across different geographical regions in Toronto. Conclusions Street view-derived environmental perceptions show spatially non-stationary associations with the prevalence of T2DM in Toronto. Higher T2DM rates are observed in dissemination areas with better street-view environments and access to healthcare providers. This may reflect underdiagnosis in areas with poorer perceived environments and less frequent access to healthcare providers. Residents living in better-perceived environments may not necessarily engage in more physical activity or active transportation. The findings offer valuable insights to assist government and public health authorities design targeted prevention and intervention programs in Toronto. Type-2 Diabetes Street View Image Environmental Health Geographically Weighted Random Forest Spatial Non-Stationarity Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 1. Background Type-2 Diabetes Mellitus (T2DM) is becoming one of the most extensive global healthcare emergencies, and its prevalence is increasing worldwide. Taking Toronto, Canada, as an example, more than one in eight adults have T2DM, and this rate has almost doubled over the past decade [ 1 ]. Public health and medical researchers have studied how environmental factors and exposures significantly shape people’s daily lifestyles and behaviours. They are hypothesized to correlate with the risk of T2DM by influencing psychosocial stressors and physical activities [ 2 , 3 ]. The availability or proximity to recreational resources, green spaces, walkways, open spaces, and well-designed public areas encourages physical activity. Residents would prefer exercising in walkable neighbourhoods with pleasing environments, which reduces the risk of T2DM [ 4 – 11 ]. Neighbourhoods with strong social networks and safer, greener environments will also improve mental health and encourage physical activity [ 12 ]. However, an unpleasant environment can discourage healthy lifestyles. Neighbourhoods with limited access to supermarkets, increasing residential noise, air pollution, traffic, and proximity to roads may lead to greater reliance on fast food outlets, elevated chronic stress, and inflammation of blood vessels due to oxidative stress [ 7 , 11 , 13 – 20 ]. Crime, social disorders, and unsafe neighbourhoods will promote social isolation, anxiety, fear, and a reduction in outdoor activity time. These factors and other combined biological, behavioural, and socioeconomic status contribute to insulin resistance and increase the risk of T2DM. Aside from environmental effects, existing research has recognized the critical role of socioeconomic status and lifestyle factors in influencing the T2DM prevalence rate. Social and healthy lifestyle determinants have also been prioritized as a top consideration in T2DM clinical treatment and care, alongside traditional medical prevention strategies [ 21 ]. Various studies show that older adults and the unemployed have higher risks of diagnosing T2DM [ 22 – 24 ]. Lower education and income are linked to higher rates of T2DM [ 23 , 25 , 26 ]. Regarding ethnicity, recent studies found that the non-Hispanic white population has a lower rate of T2DM prevalence rate compared with other ethnic groups [ 27 , 28 ]. For example, South Asian and black immigrants in Canada have been reported to face a higher likelihood of developing T2DM earlier in their life course compared to immigrants from the UK [ 23 , 29 ]. Lifestyle factors such as smoking, heavy alcohol consumption, and obesity are well-known contributors to T2DM risk [ 30 – 32 ]. Psychological and emotional stress also play a significant role. Individuals experiencing depression and anxiety may find it challenging to maintain self-care routines, including monitoring blood glucose levels, maintaining a healthy diet, and engaging in regular physical activity [ 33 , 34 ]. Marital relationships may further affect T2DM management by worsening glycemic control through increasing stress [ 35 ]. Additionally, access to regular healthcare is vital for the effectiveness of T2DM management. Consistent interaction with healthcare providers will improve health outcomes, effectively preventing and managing T2DM complications [ 21 , 36 ]. Ultimately, the interplay between socioeconomic status and lifestyle behaviours creates a synergistic effect significantly influencing T2DM prevalence and management. Various measurement methods on how the socioeconomic, environmental, and lifestyle context impacts T2DM health outcomes were utilized in past studies, including generalized linear models, spatial estimations, and Bayesian approaches [ 37 ]. Nevertheless, research often assumes that the effects are spatially stationary and do not differ at different geographic locations. Environmental and socio-determinants usually have inconsistent associations with T2DM or other health outcomes [ 38 ]. Traditional, non-spatial approaches likely cause biased model performance and lead to erroneous. For instance, green space density and income were positively correlated with the T2DM prevalence rate in one part of the city but negatively correlated with the T2DM prevalence rate in other neighbourhoods [ 39 , 40 ]. Recognizing spatial non-stationarity effects further helps to investigate the ignored aspects and improve our understanding of the spatial phenomena on determinates of health outcomes. To address these limitations, recent advancements in spatial analytical methods have introduced advanced models that account for spatial heterogeneity. Once such model is Geographically Weighted Random Forest (GWRF), a machine learning-based extension of the Random Forest algorithm designed to capture spatial non-stationarity. GWRF can adaptively capture and interpret intricate patterns of spatially diverse dataset, accept non-parametric predictors, alleviate overfitting issues with its bootstrapping nature, and retain explainable quantification for the spatial variation of localized relationships [ 41 , 42 ]. The model has recently gained attention in health geography for its effectiveness in modelling spatially varying associations with improved performance compared to non-machine learning based spatial models. For example, Quiñones et al. (2021) applied GWRF to explain spatial non-stationary effects and predict T2DM prevalence through obesity, physical activity, food environment, and socioeconomic data across the United States [ 43 ]. Lotfata et al. (2023) explored how environmental and socioeconomic factors influence asthma prevalence differently across the United States using GWRF [ 44 ]. Grekousis et al. (2023) used GWRF to estimate the COVID-19 death rate through socioeconomic and health-related factors in the United States. These studies highlight the growing role of GWRF in health research, where it often outperforms ordinary least squares, conventional random forest, and geographically weighted regression models in terms of predictive accuracy and spatial interpretability [ 45 , 46 ]. Despite researchers striving to unravel the intricate connections between the environment and human health outcomes for decades, a notable gap in our knowledge currently surrounds the influence of human perceptions of the environment on health outcomes. Human perceptions and feelings about the environment represent how people view and interpret their contextual surroundings. Few studies have explored the relationship between human perceptions of the environment and T2DM prevalence across large-scale geographical areas, as challenges arise with large-scale collection of human perceptions. The techniques, such as environmental audits, are usually labour-intensive and require significant time [ 47 ]. This can be driving around communities to measure the greenness in each neighbourhood or inquiring about residents’ perceptions through surveys and interviews. With the rapid advancement in machine learning algorithms, there is a growing trend toward harnessing computer vision and big data images, such as Street View Imagery (SVI), to quantify environmental characteristics and understand their impacts on health outcomes [ 47 – 49 ]. SVI captures the environment at eye level with 360-degree panoramic pictures, revealing real-world scenery from a pedestrian perspective. Furthermore, SVI can be used to extract both objective and subjective measurements of environmental features. Objective measurements include roads, humans, sidewalks, vehicles, constructions, objects (traffic poles, lights, and signs), proportion of nature (vegetation and terrain), and sky. In contrast, perception variables can be human subjectivity of their feelings about the environment, including safety, beauty, liveliness, dullness, depression, and a wealthy environment [ 47 , 50 , 51 ]. The urban environment is a complex system in which various environmental features are interconnected and jointly influence resident’s health outcomes and behaviours. Studies have shown that the characteristics of street environments, such as greenery, building types, pedestrian infrastructure, road design, and street connectivity, were closely associated with residents’ health behaviours and the prevalence of chronic diseases, including high blood pressure, high cholesterol, diabetes, and heart disease [ 48 , 52 – 54 ]. In particular, studies using SVI identified single-lane roads and visible utility wires, which are markers of older or lower-income neighbourhoods, as being associated with an increased risk of diabetes and obesity [ 54 ]. Street lighting poles improve visibility at night, reducing crime and injury and increasing perceptions of safety. A well-lit street may encourage evening walks or other outdoor activities, whereas poorly lit areas can deter people from being active after dark [ 55 ]. Meanwhile, non-single-family home buildings were associated with decreased obesity, diabetes, and physical inactivity [ 54 ]. Physical barriers such as fences may limit access to public green spaces, thereby affecting their willingness to conduct physical activity and overall health. In contrast, perceptions of a beautiful, lively environment have been shown to benefit health outcomes, especially mental health [ 56 – 58 ]. Greener, safer, wealthier, and more beautiful environments can reduce depression and stress [ 57 , 59 , 60 ]. For Example, environments perceived as beautiful, safe, and lively are associated with lower psychological stress, better mental health, and increased physical activity through active transportation, contributing to reduced chronic disease risk [ 61 – 65 ]. Better-perceived neighbourhood environments also significantly correlate with reduced sedentary behaviour, smoking, and drinking, promoting healthier lifestyles and greater life satisfaction [ 66 ]. Conversely, unsafe neighbourhood perceptions may foster social isolation and fear, leading to reduced physical activities and hindering the prevention of chronic diseases like T2DM, which requires sustained engagement in healthy behaviours [ 67 ]. Therefore, the environment's objective and subjective characteristics can directly or indirectly influence people’s health outcomes. Although much research has identified that street view-derived environmental characteristics, socioeconomic factors, and lifestyle habits affect health outcomes by influencing behaviours, few studies have investigated their associations with T2DM prevalence in a comprehensive, city-level context from a geospatial perspective. Further, to the best of our knowledge, no research assesses the impact of neighbourhood environment perceptions on T2DM prevalence. The use of GWRF to explain spatial-non-stationarity effects of T2DM prevalence has also not been extensively explored. To address these gaps, this research uses SVI-derived environmental, socioeconomic, and lifestyle factors to investigate how they affect T2DM at the Dissemination Areas level in Toronto through Geographically Weighted Random Forest modelling. The goals are twofold: 1) To explore the relationships between SVI-derived environmental, socioeconomic, and lifestyle factors and T2DM prevalence rates, and 2) To determine how these associations vary across the city of Toronto. 2. Methodology This study examines the spatial effect of SVI-derived environmental, socioeconomic, and lifestyle determinants on T2DM prevalence rates across 3800 Dissemination Areas (DA) in Toronto. A Geographically Weighted Random Forest (GWRF) regression model is applied to address spatial autocorrelation and investigate global and spatial associations between the factors with T2DM prevalence. This is achieved through graphing partial dependency plots and mapping the importance of local permutation features. A graphical summary of this study’s workflow is provided in Fig. 1 . 2.1 Study Area Located in the southern region of Ontario, Toronto serves as Ontario’s capital city. Covering a total land area of 630.2 km 2 . It is home to three million culturally diverse citizens residing in 3800 officially registered Dissemination Areas [ 68 ]. Toronto’s environmental characteristics vary across neighbourhoods, ranging from dense urban centers in the central and southern areas to extensive green lands in the east. Its socioeconomic factors, such as income levels, education attainment, and access to healthcare, also exhibit significant spatial variation across the city. This study, therefore, conducts an observational study covering Toronto’s DA to investigate the social and environmental determinants associated with the prevalence of T2DM. 2.2 Data 2.2.1 T2DM Prevalence and Health-Related Factors The 2017/2018 T2DM prevalence data for the Toronto DA used in this study is collected and modelled by Environics Analytics as part of their Community Health Collection [ 69 ]. The Community Health data was derived from the 2017/2018 Canadian Community Health Survey and modelled using the 2021 Census demographic data to ensure it is representative of all census units across Canada by Environics Analytics [ 68 , 70 ]. This dataset includes survey responses from participants aged 12 and above who reported being diagnosed with T2DM. This study also uses additional health factors from the Community Health dataset to explore their associations with T2DM and control for potential confounding variables. The factors include obesity prevalence (for body weight index higher than 30), smoking rate, drinking rate, percentage of population with satisfaction of life reported as very dissatisfied, perceived poor mental health, perceived anxiety disorders (including phobia, obsessive-compulsive disorder, and panic disorder), weekly physical activity minutes, and access to regular health care providers. Each data record was matched and spatially joined with the Toronto DA boundary shapefile for spatial analysis. 2.2.2 Environmental-Related Factors To investigate the street view environmental factors associated with T2DM, this study incorporates variables derived from street view, GIS vector data, and remote sensing images. The analyzed factors include perceived environmental beauty and safety, fence density, pole density, traffic light density, wall density, road density, cycling network density, water body density, and Normalized Difference Vegetation Index (NDVI) within each DA. This study used a pre-trained machine-learning model to predict perceived safety and beauty across Toronto DA [ 58 , 64 , 71 ]. The model is trained by crowdsourced perception data named Place Pulse V2, including 1,170,000 pairwise companions of 110,988 SVI from 81,630 volunteers [ 72 ]. Street View Images (SVI) around Toronto taken in 2016 were included in the analysis [ 73 ]. Every sample location is spaced 50 meters apart, with four images assessed from view angles: North, East, South, and West, to obtain a panoramic view of the surrounding environments at each location. The predicted perception score ranged from 0 to 1, where the higher value means a higher likelihood of being perceived as beautiful or safe. After perception calculations, this study aggregated the scores by taking the average perception scores from four view angles for each location. As the beauty and safety perception scores were highly correlated, with a Pearson Correlation Coefficient of 0.98, including both variables in a model could create reduced model performance and biased interpretations. To address this issue and preserve the joint contribution of both perception measures, we combined the two scores into a single variable by multiplying them. Areas perceived as both beautiful and safe are more likely to support confortable and active engagement with the environment. This approach emphasizes locations where both variables are high while down-weighting locations where either perception is low. To further obtain built environmental characteristics in Toronto, the proportion of fences, poles, traffic lights, and walls viewed in each image along road networks were derived from SVI using the Pyramid Scene Parsing Network (PSPNet) [ 71 , 74 ]. Scene parsing employs computer image recognition to classify each pixel with a category label. PSPNet is a deep convolutional neural network-based model that captures details of raster images at different scales and combines extracted components to segment the micro-built environment [ 74 ]. A pre-trained PSPNet with a Cityscape dataset including 25,000 labelled images focusing on urban street scenes was obtained from GluonCV and used in this study [ 75 , 76 ]. Although SVI captures most of the built environment along road networks, some areas remain out of reach from the road network, such as parks and community gardens, without SVI coverage. Therefore, this study also calculates the Normalized Difference Vegetation Index (NDVI) to include in the regression model using Landsat 8 Collection 2, Level 2, Tier 1 imagery captured on June 11, 2018 [ 77 ]. Besides NDVI, the study also includes additional traditional environmental variables, including road density, cycling network density, and water body density, calculated by dividing their respective lengths or areas (in meters or square meters) by the DA (in square meters). Road network data was sourced from the 2022 Ontario Road Network Composite, cycling network data was obtained from the 2019 City of Toronto Physical Area of Sidewalks, and water body density was derived from the 2019 Government of Canada’s Lakes, Rivers, and Glaciers Hydrographic CanVec Series [ 78 – 80 ]. Access to environmental facilities and built environment exposures could happen beyond existing administrative boundaries, characterized as part of the Uncertain Geographic Context Problem (UGCoP) in spatial analysis [ 81 ]. People’s activities do not occur entirely within their residential units; they may spend time outside their home neighbourhoods [ 82 ]. Incorporating the 15-minute city concept from urban planning, this study thus expanded the boundary of each DA from its centroid using a 15-minute walking service area delineated with ArcGIS Pro’s network analysis tool (version 3.4.0) [ 83 ]. All features within the 15-minute walking range, including those extending beyond Toronto’s administrative boundaries, were summarized and standardized by the total area after the boundary expansion, accounting for the “edge effect” when facilities can be accessed beyond the given boundary [ 84 ]. 2.2.3 Socioeconomic Status To explore the associations of socioeconomic status with T2DM, this study included age, income, non-citizens population, unemployment rate, population without an education diploma or certificate, marriage and common law, immigration population, commute duration, population commuting by transit, and population commuting by walking and cycling in the regression model. The variables were all obtained from the 2021 Canadian Census at the DA level [ 68 ]. About 2% of the DA units do not have reported census and health status due to their small geographical and population size, which is intended to protect residents’ privacy. As a result, the study calculates the average values of their immediate neighbours to fill in the missing data, ensuring these neighbourhoods are not excluded during the regression modelling process. 2.3 Geographically Weighted Random Forest Regression Environmental factors and their health effects might not be stable over space [ 38 ]. Furthermore, T2DM prevalence exhibits a spatial autocorrelated pattern with similar prevalence rates clustered together in neighbourhoods [ 23 , 39 , 85 ]. Therefore, a spatial machine learning-based regression model, Geographically Weighted Random Forest (GWRF), is used in this study to explore and mitigate the effects of spatial autocorrelation. GWRF is a non-parametric, tree-based, ensemble model that inherits the classical Random Forest (RF) regression structure calibrated locally through a moving window approach using spatial weight matrixes [ 41 , 42 ]. The model captures spatial variation in localized relationships by generating random forest regression models for each spatial unit using nearby observations. The voting structure of the random forest regression can further mitigate the effects of non-linear relationships between the dependent and independent variables and multicollinearity among the independent variables [ 86 ]. This study used the “SpatialML” package (Version 0.1.7) in the R Statistical Computing Environment (Version 4.4.1) to create GWRF models [ 41 , 42 ]. The T2DM prevalence rate was the dependent variable, and 27 variables proposed in section 2.2 were included as independent variables in the modelling. The GWRF models were set to be based on adaptive spatial kernels, as the bandwidth size was adjusted based on the data’s local density. Following past studies, a random grid search was applied during the hyperparameter tuning [ 43 – 45 ]. The number of trees to grow for each round was set to be between 200 and 1500. The number of variables randomly sampled as candidates at each split was set to be between 5 and 11, and the number of nearest neighbourhoods to be used as the bandwidth ranged from 100 to 600 during the tuning process without growing complex trees to control for overfitting. The model with the lowest global Out-of-Bag (OOB) Mean Squared Error (MSE) and the highest global OOB R 2 as the best performance was selected as the final model. The model results were interpreted through local OOB R 2 , local Permutation Feature Importance (PFI), and Partial Dependent Plots (PDPs) following random forest evaluation approaches. PFI scores measured the importance of individual features in the predictive power of the GWRF model. It offered insight into identifying which independent variable best fits the model by measuring how much the model’s performance decreases when the values of a specific feature are randomly shuffled (permuted). It also highlighted critical features contributing to the dependent variable under the pretense of non-linearity or collinearity among variables in the model. PDPs aid the interpretation of the marginal relationships between the predictor and response variables. It illustrates the average effect of a predictor on the predicted outcome while holding all other variables constant. This allows us to visualize the strength and association of a predictor to the response variable. This study uses the R package “pdp” Version 0.8.1 to compute PDPs on the global GWRF model, equivalent to standard Random Forest models [ 87 ]. The generated plots provide an overview of each variable’s marginal influence across the study area. 3. Results 3.1 Summary Statistics Figures 2 a illustrate the spatial distribution of the T2DM prevalence rates across Toronto DA. On average, 9% of the Toronto population reported being diagnosed with T2DM, ranging from 0 to 21.66% in various DA. Overall, higher rates of T2DM are observed in the central, northern, southwestern, and along the southeastern shoreline. In comparison, lower rates are found in the central south, northeastern and northwestern regions. A Global Moran’s Index with 1000 Monte Carlo (MC) simulation tests of significance was performed for T2DM prevalence, resulting in a value of 0.4214 ( p = 0.001). This indicates a moderate positive spatial autocorrelation for T2DM prevalence in Toronto. Similar T2DM rates tend to be spatially clustered, as shown in Fig. 2 a. Other lifestyle indices in Fig. 2 b to 2 i also exhibit spatially clustered patterns, such as high regular alcohol drinking rates clustering in southern regions, where more than half of the population are regular drinkers. Obesity, smoking, poor mental health, and anxiety disorder rates are clustered in the central-west, accounting for 10 to 20% of the region’s population. The perception of environmental beauty and safety derived from SVI is assessed as an index ranging from 0 (least beautiful) to 1 (most beautiful). Neighbourhoods in the blue areas of Fig. 3 d, such as central Toronto and southeast and southwest Toronto, received high beauty perception scores, ranging from 0.401 to 0.719. Likewise, these regions also have the highest NDVI (Fig. 3 i) coverage across the city, ranging from 0.248 to 0.404. Regarding built environment characteristics, most factors, including road network, cycling network, wall, fence, pole, and light coverage, were found to cluster in the southern downtown regions, as illustrated in Fig. 3 . On the other hand, socioeconomic factors in Toronto show distinct spatial patterns, as illustrated in Fig. 4 . The average Toronto population is 42 years old, with 44.64% married or with common law. Higher-income populations were found in central Toronto (Fig. 4 c), where higher-income populations reached more than 70,000 CAD median annual income. Non-citizens (Fig. 4 d) and immigrant populations (Fig. 4 e) cluster in eastern and northern DA, with the latter reaching up to 100%. A higher percentage of walking and cycling commutes (Fig. 4 i) cluster in the downtown core, while longer weekly commuting durations (Fig. 4 j) are concentrated in suburban areas, reaching a maximum of 2640 minutes weekly. A smaller portion of the population lacks certificates or diplomas, with an average of 12.75%. Unemployment remains low, with an average of 2.45%, but peaks at nearly 30% in specific areas. The summary statistics for all the variables are also listed in Table 1 . Table 1 Summary statistics for variables included in the GWRF regression model Variable Minimum Maximum Mean Median Standard Deviation T2DM Prevalence (%) 0.0000 21.6578 9.0811 9.2553 4.3024 Obesity Rate (%) 3.9574 37.3653 18.3241 17.7219 5.5923 Smoking Rate (%) 0.9114 36.0052 9.8956 8.2169 5.7542 Drinking Rate (%) 23.4564 84.9271 48.1948 46.0313 13.2037 Weekly Recreation Physical Activity Minutes 18.2656 69.3730 33.3717 32.3303 6.5192 Have Regular Health Care Provider (%) 53.8117 99.9382 87.0798 89.8600 8.9792 Poor Mental Health (%) 0.0000 12.3337 2.2696 1.7691 1.8387 Life Dissatisfaction (%) 0.0000 4.3778 0.5229 0.3046 0.6416 Anxiety Disorder (%) 0.1651 25.3759 7.9948 7.4276 3.6283 Road Network Density (Distance/m²) 0.0000 0.0245 0.0129 0.0122 0.0038 Cycling Network Density (Distance/m²) 0.0000 0.0066 0.0015 0.0013 0.0012 Water Body Density (% of Coverage) 0.0000 0.3237 0.0052 0.0000 0.0172 Street View Beauty and Safety Perception 0.0131 0.7187 0.2911 0.2869 0.1271 Street View Wall Coverage 0.0000 0.0148 0.0032 0.0022 0.0028 Street View Fence Coverage 0.0005 0.0317 0.0119 0.0103 0.0054 Street View Pole Coverage 0.0005 0.0047 0.0026 0.0026 0.0006 Street View Traffic Light Coverage 0.0000 0.0003 0.0001 0.0001 0.0001 NDVI 0.0435 0.4044 0.2121 0.2160 0.0450 Age 0.0000 88.8000 42.2093 41.9000 5.5787 Married or Common-Law (%) 0.0000 62.3836 44.6355 45.9519 7.9763 Median Total Income (CAD) 0.0000 114000 42,899 38800 13,553 Non-Citizens (%) 0.0000 70.4225 14.2876 12.2388 9.6971 Immigrant Population (%) 0.0000 100.00 43.6088 44.1548 16.4581 Without Certificate or Diploma (%) 0.0000 43.1953 12.7512 11.4003 7.6821 Unemployment (%) 0.0000 29.7884 2.4512 2.0929 1.8098 Commute by Public Transit (%) 0.0000 29.5203 6.7932 6.0743 4.4106 Commute by Walking and Cycling (%) 0.0000 29.7619 2.1626 0.9288 3.1307 Weekly Commuting Duration (minutes) 0.0000 2640.00 212.4 155.00 197.34 3.2 Geographically Weighted Random Forest Regression The final tuned GWRF model results with the hyperparameters: number of trees, Mtry, and neighbourhood bandwidth of 1088, 11, and 435, respectively. The global, local OOB R 2 , global, and local OOB MSE for the model were 0.7763, 0.7541, 4.1400, and 4.5505, respectively. This represents that around 75% of the variation in T2DM prevalence rate could be explained by the relationships with the explanatory variables through the model. More than half of the neighbourhoods had local R 2 values of 0.65 or above (purple areas in Figure A1 a of the Appendix ). The random distribution of mapped OOB residuals in Figure A1 b of the Appendix and the Global Moran’s Index with 1000 MC simulations resulted in a value of 0.0196 ( p = 0.02), indicating no significant spatial autocorrelation presented in the model’s residuals. Furthermore, 10-fold spatial cross-validation of the model’s prediction performance resulted in consistent performance across folds (Table A2 in the Appendix ), with a mean R 2 , Root MSE, and Mean Absolute Error of 0.7224, 2.2770, and 1.7638 without showing signs of overfitting nor underfitting. The local PFI scores are visualized in Fig. 5 . Among all 27 variables in the model, lifestyle factors contributed the most overall, with access to regular health care ranking the highest, followed by obesity, smoking, drinking, and life dissatisfaction rates. Despite age being the second most crucial factor, socioeconomic factors such as income, immigrant status, and commuting patterns are ranked after lifestyle factors. Beauty and safety perceptions ranked the highest among all the environmental factors, followed by NDVI, which outperformed built environmental factors. The local PFI maps in Fig. 6 showed inconsistent spatial effects of socioeconomic, environmental, and lifestyle factors on the T2DM prevalence rate. For example, age contributed to predicting T2DM the most in Northern DAs (Fig. 6 a). In contrast, non-citizens, the immigrant population, drinking rate, and environmental perception of beauty and safety affect T2DM prevalence around downtown Toronto the most (Figs. 6 d, 6 e, 6 l, and 6 v). The variables with the highest PFI were mapped in Fig. 7 , representing the dominant feature associated with T2DM prevalence in each DA across Toronto. Regular health care dominates Toronto’s central DAs, while age and obesity were identified as the most important factors in the northeastern or southwestern and southeastern or southwestern DAs, respectively. Beauty and Safety Perceptions were found to be the most important features for some DAs in southern and central Toronto. The summary statistics of the PFI scores were recorded in Table A1 of the appendix . This study used PDPs to explore the associations between the dependent and independent variables while accounting for the average influence from other independent variables (Fig. 8 ). The relationships between T2DM prevalence and the independent variables were non-linear. The T2DM prevalence rate would sharply increase by about 3% on average when the population age increased from 35 to 60 (Fig. 8 a). Moreover, T2DM prevalence would also increase by 2%, 2.5%, 4%, and 1% when the smoking rate, obesity prevalence, and the population with regular health care provider rate, environmental beauty and safety perception, increased from 0–15%, 10–20%, and 75–95%, 0.1 to 0.5 respectively (Figs. 8 k, 8 m, 8 o, 8 v). Most associations have an s- or inversed s-shaped sigmoid curve, where the effect gradually flattens at both ends. The T2DM prevalence rate was weakly positively associated with median income, physical activity, poor mental health, life dissatisfaction, anxiety disorder, and NDVI. Yet weakly negatively related to non-citizens, immigrant population, commuting duration and drinking rate. 4. Discussion 4.1 Lifestyle Factors and Regular Health Care Access This study revealed a positive association between well-designed environmental factors - including perceptions of safety, aesthetic appeal, and greenery - and higher T2DM prevalence rates. This differs from the hypothesis that a well-designed, beautiful, and safe neighbourhood environment can attract and promote physical activities that reduce the burden of T2DM. By examining the distribution of variables around Toronto, the spatial distribution of higher T2DM prevalences in DAs (Fig. 2 A) is associated with higher proportions of healthcare providers (Fig. 2 F), higher income levels (Fig. 4 C), greater perceptions of environmental beauty and safety (Fig. 3 D), higher NDVI measurements (Fig. 3 i), as well as lower active transportation time (Figs. 4 h and 4 i) and commuting duration (Fig. 4 j). The neighbourhoods in central Toronto are wealthier and receive better environmental beauty and safety perceptions; however, residents in these areas tend to commute by car, resulting in lower public transportation ridership across Toronto [ 88 ]. Compared to commuting and shopping through public transit, car ridership reduces the time spent on moderate and vigorous physical activity and increases sedentary time [ 89 ]. This lifestyle, characterized by more sedentary behaviours and decreased physical activity, may contribute to the observed increased prevalence of T2DM in these regions despite a favourable built environment. Furthermore, having a regular healthcare provider was ranked as the most important factor explaining T2DM prevalence in the GWRF model, which dominates most of Toronto’s DA (Fig. 7 ). The standard of diabetes care developed by the American Diabetes Association has emphasized that access to regular healthcare is crucial for effective diabetes management and prevention [ 21 ]. Despite that existing studies reported that 70% of Ontario received at least a blood glucose test between 2010 and 2017 [ 90 , 91 ], more recent statistics from 2019 to 2021 show that fewer than 50% of diabetic populations met Diabetes Canada’s recommended standards for regular Hemoglobin A1c (HBA1c) blood glucose testing [ 92 ]. These findings may suggest that wealthier areas with better access to healthcare are more likely to screen and diagnose T2DM at earlier stages. In many cases, the residents of more affluent neighbourhoods with regular access to healthcare are more likely to prioritize their health outcomes and seek diabetes screenings regularly. In contrast, neighbourhoods with limited healthcare access may have undiagnosed or hidden cases of T2DM due to low awareness or a lack of screening, resulting in low prevalence in the region. No matter how favourable the environment is, if residents do not undergo a healthy lifestyle with regular diabetes screening, the burden of T2DM will persist. Therefore, this study suggests that T2DM intervention and prevention efforts should focus on increasing screening rates in neighbourhoods with fewer healthcare providers and poorer environmental perceptions while emphasizing education about healthy lifestyles in higher-income neighbourhoods with better environments in Toronto. Similar to previous studies, other lifestyle factors, such as increasing smoking rates, obesity prevalence, poor mental health, and anxiety disorders, were found to be correlated with increasing T2DM prevalence [ 31 – 34 ]. They were identified as the most important determinants of T2DM compared to environmental and socioeconomic factors. Drinking rates exhibited spatial non-stationarity in their relationship with T2DM and were identified as a highly important factor in eastern downtown Toronto (Fig. 6 l). The observed spatial pattern could reveal that in downtown regions with a cluster of pubs, higher drinking rates might reflect social lifestyle rather than behaviours directly contributing to T2DM risk, providing further insights into the positive association between alcohol consumption and T2DM observed in previous studies [ 30 ]. Residents around the downtown area often walk to pubs for drinking, promoting active transportation paired with their drinking behaviour. In other regions, drinking may reflect stress-included or habitual consumption, which can negatively impact health and increase T2DM risk. 4.2 Street View Image Derived Environmental Measurements Even though the perception of environmental beauty and safety was positively associated with T2DM prevalence rates in Toronto, it is identified as the most important environmental factor in the model, contributing more than the widely used environmental factors, including satellite-derived NDVI, GIS-derived water body, road, and cycling network density. Since the perception of beauty and safety measurements rely on environmental greenery and streetscape design, the vegetation coverage along the streets extracted from SVI could contribute to the importance of explaining T2DM. Residents constantly see roadside vegetation coverage during daily commuting, exercising, and shopping compared to greenery spaces like parks. A previous study identified that green space quantified using parks, recreational areas, sports fields, open spaces, campgrounds, and golf courses is relatively less in central Toronto than in the suburban regions of eastern Toronto. Some neighbourhoods in suburban areas with larger green spaces were associated with higher T2DM prevalence [ 39 ]. However, the perceived environmental safety and beauty extracted from SVI in this study were the most important local features in central Toronto compared to other regions of Toronto (Figs. 6 y and 6aa). Even though Eastern Toronto (Scarborough) regions have the highest green space coverage in Toronto, including a national urban park, there might not be opportunities for outdoor activities without the concern of being interrupted by dangerous road traffic. This may also suggest that simply increasing the green space or neighbourhood greenery without considering its aesthetics, safety, and accessibility may not effectively control or reduce T2DM prevalence rates. An unsafe environment that limits physical activity opportunities and creates social isolation will also contribute to an increased prevalence of T2DM in neighbourhoods around Toronto. Moreover, the SVI-derived built environmental characteristics observed showed a significant positive association between increasing walls and traffic lights with higher T2DM rates from the GWRF model. This aligns with the judgement that more walls and traffic lights on the streets due to urban development create barriers to physical activity around the city that promote an active lifestyle, which was associated with lower T2DM prevalence [ 93 ]. Therefore, SVI-derived perception of the environment can provide a more comprehensive quantification and representation of both the built and contextual environmental factors. 4.3 Socioeconomic Status Similar to SVI-derived environmental variables, associations between socioeconomic status were also found with T2DM prevalence. For example, the association between age and T2DM prevalence, the second most important factor identified in the GWRF model, showed an increasing positive trend, starting with no relationship but encountering a sharp rise at age 37. It gradually transitioned to a weak positive relationship by age 50 and slowly stopped at 70. This aligns with the well-recognized phenomenon that individuals with increasing age are more prone to developing T2DM [ 94 , 95 ]. In addition, non-citizens and immigrant populations were negatively associated with the T2DM prevalence rate. Unlike the traditional hypothesis that immigrants experience a socio, environmental and dietary structural change, including increased consumption of ultra-processed foods, leading to a higher risk of developing T2DM [ 92 , 96 ]. The observed local permutation feature importance map indicated that most contributions are concentrated in Toronto’s downtown region (Figs. 6 d and 6 e). This could be explained by the fact that new immigrants living in the high-density downtown areas are wealthier, younger, and more likely to choose healthier diets. However, another possibility is that new immigrants and non-citizens typically have limited access to healthcare resources in Canada, leading to lower screening and prevalence rates among immigrant populations [ 97 – 99 ]. Furthermore, active commuting by transit was found to have a weak negative relationship with T2DM, which is aligned with the observation from previous studies [ 100 ]. 4.4 Significance of Contribution, Limitations, and Future Directions This study is one of the first to utilize the novel GWRF regression to explore street view image-derived environmental factors, including perceptions of beauty and safety with T2DM prevalence across a large-scale city-wide study area. It provides a detailed characterization of spatial non-stationarity in environmental, socioeconomic, and lifestyle factors, highlighting their varying impacts across different city regions. GWRF successfully handled the spatial autocorrelated T2DM prevalence rate, further providing insights into local feature importance and highlighting the spatial variation of factors (Figs. 6 and 7 ). Combining the results from partial dependency plots, local feature importance derived from GWRF offers a comprehensive understanding of how key variables contribute to T2DM prevalence within the study area. This approach identifies the most influential variables driving T2DM prevalence, including access to regular health care, age, and obesity prevalence. It further reveals their localized associations and spatial variability, providing a holistic perspective on the spatial dynamics of T2DM risk factors. The results of this study will support the design of targeted T2DM intervention and prevention programs around Toronto neighbourhoods. Understanding the spatial relationship of environmental, socioeconomic, and lifestyle characteristics with T2DM will inform policy and urban planning efforts to create healthier, more livable communities. Healthcare practitioners and policymakers must consider the neighbourhoods’ diverse spatial phenomena and social-environmental aspects during intervention programs and policymaking. For example, evidence from this study could suggest that identifying and optimizing resource allocation, particularly healthcare and targeted screening resources, is critical for neighbourhoods with lower incomes, poorer environmental factors, and lower regular healthcare providers. Meanwhile, promoting healthy lifestyle behaviour education in weather and neighbourhoods with high car ridership. This targeted approach could help bridge disparities in T2DM prevention and care across different communities. There are several areas for future research to improve our understanding of the underlying mechanisms and address existing knowledge gaps. One limitation of this study was the availability of a single-year T2DM prevalence rate. With multiple years of data, future studies could investigate the lag effect of environmental, socioeconomic, and lifestyle determinants on T2DM prevalence. Multitemporal data could also be utilized to evaluate the predictive performance of GWRF models and enhance their accuracy by balancing model complexity with reducing the risk of overfitting. Additionally, the predictive performance could be compared against other models, such as Multiscale Geographically Weighted Regression. Future studies should also explore whether lifestyle behaviours, such as reduced active commuting or sedentary time, mediate the relationship between environmental factors and T2DM prevalence. When studying regions with lower screening rates, using prevalence data may lead to erroneous conclusions in examining social and environmental determinants of T2DM. Therefore, exploring the interaction of environments and contextual factors with target achievement rates, such as HbA1c levels, could provide further insights into the role of environmental conditions in primary (preventing the development of diabetes) and secondary (management of diabetes to prevent complications) T2DM preventions. 5. Conclusion This study is among the first to investigate the associations between street view-derived environmental, socioeconomic, and lifestyle factors with T2DM prevalence through Geographically Weighted Random Forest regression. The results revealed that regular access to health care providers, age, obesity, smoking, mental illness, environmental beauty and safety perception, and NDVI were the most important factors positively and non-linearly associated with T2DM prevalence. This differs from the traditional hypothesis that a better-designed environment and frequent access to healthcare providers lead to reduced T2DM. However, the population with higher T2DM prevalence resided in wealthier and more affluent neighbourhoods with more frequent access to healthcare providers in Toronto. This suggests the importance of strengthening lifestyle interventions and education in communities with better healthcare conditions and environments to prevent T2DM actively. At the same time, screening efforts should be enhanced in communities with relatively poorer healthcare and environmental conditions. The findings from this study will provide evidence and recommendations to support the government and public health authorities in designing targeted education, prevention, and intervention programs that can control and reduce the increasing burden of T2DM. Abbreviations CAD Canadian Dollar DA Dissemination Areas GWR Geographically Weighted Regression GWRF Geographically Weighted Random Forest HbA 1c Hemoglobin A 1C MSE Mean Squared Error MC Monte Carlo NDVI Normalized Difference Vegetation Index OLS Ordinary Least Square OOB Out-of-Bag PDPs Partial Dependency Plots PFI Permutation Feature Importance R 2 Coefficient of Determination SVI Street View Imagery T2DM Type-2 Diabetes Mellitus VIF Variance Inflation Factor Declarations Ethics approval and consent to participate Not applicable Consent for publication Not applicable Availability of data and materials Diabetes and healthy lifestyle datasets analyzed during the study are available from Environics Analytics, but restrictions apply to the availability of these data, which were used under license for the current study and are not publicly available. The Census and GIS are publicly available through the Canadian national, Ontario provincial, and the City of Toronto open data portals listed in the method section supporting this article's findings and conclusions. Competing of Interest The authors declare that they have no competing interests. Funding This work was supported by the Social Sciences and Humanities Research Council of Canada [Grant Number: 430-2023-00116] and the Manchester-Melbourne-Toronto Research Fund. Authors’ Contributions H.G. conceived, designed, and implemented the experiments, analyzed the results, and wrote the manuscript. J.W. conceptualized the study, acquired funding, supervised the project, and contributed to the review and editing of the manuscript. D.Y.W. contributed to acquiring and processing street view data, improving the methods, interpreting the results, and conducting the manuscript discussion session. H.Z. contributed to acquiring and processing lifestyle and street view data. All authors have read and approved the published version of the manuscript. Acknowledgments The authors extend their sincere appreciation to the anonymous reviewers and editors for providing valuable comments that significantly contributed to improving the quality of the manuscript. The authors also wish to thank Environics Analytics for providing access to the health and lifestyle factor data used in this study. Special thanks to Dr. Pierre Desrochers and Dr. Chunjiang Li for their unwavering support and valuable input, which played a crucial role in enhancing the overall quality of this research. Furthermore, the authors would like to acknowledge and express their gratitude to Mr. Xianmiao Zhao, Mr. Ryan Siu, and Mr. Bruce Huang for their invaluable assistance in configuring and troubleshooting the spatial computational workstation that contributes to the modelling process in this study. Authors’ Information Department of Geography and Planning, University of Toronto, 100 St. George Street, Toronto, ON, M5S 3G3, Canada Haoxuan Ge, Jue Wang, and Devin Yongzhao Wu Department of Geography, Geomatics and Environment, University of Toronto Mississauga, 3359 Mississauga Road, Mississauga, ON, L5L 1C6, Canada Haoxuan Ge, Jue Wang, and Devin Yongzhao Wu Department of Geography, Sustainability, Community, and Urban Studies, University of Connecticut, Storrs, CT, USA Hanlin Zhou Use of Generative AI and AI-assisted technologies In the preparation of this manuscript, generative AI (ChatGPT 4o) was used solely to enhance readability and improve language clarity during the writing process. The AI assistance was limited to refining grammar, sentence structure, logistics of paragraph flow, and overall coherence without influencing the originality, accuracy, or integrity of the content. All AI-generated outputs were carefully reviewed, examined, and edited under human oversight and control to ensure that the final text was accurate, complete, and free from biases. The substantive content, analysis, and conclusions presented in this manuscript remain entirely the responsibility of the authors. References Toronto Public Health. Health Surveillance Indicators: Diabetes [Internet]. 2017 [cited 2024 Jan 7]. Available from: https://www.toronto.ca/wp-content/uploads/2017/12/8c72-tph-hsi-diabetes-july18f.pdf Christine PJ, Auchincloss AH, Bertoni AG, Carnethon MR, Sánchez BN, Moore K, et al. Longitudinal Associations Between Neighborhood Physical and Social Environments and Incident Type 2 Diabetes Mellitus: The Multi-Ethnic Study of Atherosclerosis (MESA). JAMA Internal Medicine. 2015;175:1311–20. Kahn SE, Hull RL, Utzschneider KM. Mechanisms linking obesity to insulin resistance and type 2 diabetes. Nature. 2006;444:840–6. Astell-Burt T, Feng X, Kolt GS. Is Neighborhood Green Space Associated With a Lower Risk of Type 2 Diabetes? Evidence From 267,072 Australians. Diabetes Care. 2013;37:197–201. Bodicoat DH, O’Donovan G, Dalton AM, Gray LJ, Yates T, Edwardson C, et al. The association between neighbourhood greenspace and type 2 diabetes in a large cross-sectional study. BMJ Open. 2014;4:e006076. Creatore MI, Glazier RH, Moineddin R, Fazli GS, Johns A, Gozdyra P, et al. Association of Neighborhood Walkability With Change in Overweight, Obesity, and Diabetes. JAMA. 2016;315:2211–20. Dendup T, Feng X, Clingan S, Astell-Burt T. Environmental Risk Factors for Developing Type 2 Diabetes Mellitus: A Systematic Review. International Journal of Environmental Research and Public Health. 2018;15:78. Glazier RH, Creatore MI, Weyman JT, Fazli G, Matheson FI, Gozdyra P, et al. Density, Destinations or Both? A Comparison of Measures of Walkability in Relation to Transportation Behaviors, Obesity and Diabetes in Toronto, Canada. PLOS ONE. 2014;9:e85295. Müller-Riemenschneider F, Pereira G, Villanueva K, Christian H, Knuiman M, Giles-Corti B, et al. Neighborhood walkability and cardiometabolic risk factors in australian adults: an observational study. BMC Public Health. 2013;13:755. Poortinga W. Perceptions of the environment, physical activity, and obesity. Social Science & Medicine (1982). 2006;63:2835–46. Sundquist K, Eriksson U, Mezuk B, Ohlsson H. Neighborhood walkability, deprivation and incidence of type 2 diabetes: A population-based study on 512,061 Swedish adults. Health & Place. 2015;31:24–30. Stockdale SE, Wells KB, Tang L, Belin TR, Zhang L, Sherbourne CD. The importance of social context: Neighborhood stressors, stress-buffering mechanisms, and alcohol, drug, and mental health disorders. Social Science & Medicine. 2007;65:1867–81. Ahern M, Brown C, Dukas S. A National Study of the Association Between Food Environments and County-Level Health Outcomes. The Journal of Rural Health. 2011;27:367–79. Andersen ZJ, Raaschou-Nielsen O, Ketzel M, Jensen SS, Hvidberg M, Loft S, et al. Diabetes Incidence and Long-Term Exposure to Air Pollution: A cohort study. Diabetes Care. 2011;35:92–8. Dzhambov AM, Dimitrova DD. Exposures to road traffic, noise, and air pollution as risk factors for type 2 diabetes: A feasibility study in Bulgaria. Noise and Health. 2016;18:133. Gebreab SY, Hickson DA, Sims M, Wyatt SB, Davis SK, Correa A, et al. Neighborhood social and physical environments and type 2 diabetes mellitus in African Americans: The Jackson Heart Study. Health & Place. 2017;43:128–37. Hill-Briggs F, Adler NE, Berkowitz SA, Chin MH, Gary-Webb TL, Navas-Acien A, et al. Social Determinants of Health and Diabetes: A Scientific Review. Diabetes Care. 2020;44:258–79. Mezuk B, Li X, Cederin K, Rice K, Sundquist J, Sundquist K. Beyond Access: Characteristics of the Food Environment and Risk of Diabetes. American Journal of Epidemiology. 2016;183:1129–37. Polsky JY, Moineddin R, Glazier RH, Dunn JR, Booth GL. Relative and absolute availability of fast-food restaurants in relation to the development of diabetes: A population-based cohort study. Canadian Journal of Public Health. 2016;107:eS27–33. Sørensen M, Andersen ZJ, Nordsborg RB, Becker T, Tjønneland A, Overvad K, et al. Long-Term Exposure to Road Traffic Noise and Incident Diabetes: A Cohort Study. Environmental Health Perspectives. 2013;121:217–22. American Diabetes Association Professional Practice Committee. 1. Improving Care and Promoting Health in Populations: Standards of Care in Diabetes—2025. Diabetes Care. 2024;48:S14–26. Chowdhury MAB, Uddin MJ, Khan HMR, Haque MR. Type 2 diabetes and its correlates among adults in Bangladesh: a population based study. BMC Public Health. 2015;15:1070. Kolpak P, Wang L. Exploring the social and neighbourhood predictors of diabetes: a comparison between Toronto and Chicago. Primary Health Care Research & Development. 2017;18:291–9. International Diabetes Federation. IDF Diabetes Atlas 2021 [Internet]. 2021 [cited 2023 Oct 20]. Available from: https://diabetesatlas.org/atlas/tenth-edition/ Agardh E, Allebeck P, Hallqvist J, Moradi T, Sidorchuk A. Type 2 diabetes incidence and socio-economic position: a systematic review and meta-analysis. International Journal of Epidemiology. 2011;40:804–18. Wu H, Meng X, Wild SH, Gasevic D, Jackson CA. Socioeconomic status and prevalence of type 2 diabetes in mainland China, Hong Kong and Taiwan: a systematic review. Journal of Global Health. 2017;7:011103. Cheng YJ, Kanaya AM, Araneta MRG, Saydah SH, Kahn HS, Gregg EW, et al. Prevalence of Diabetes by Race and Ethnicity in the United States, 2011-2016. JAMA. 2019;322:2389–98. Ledford CJW, Seehusen DA, Crawford PF. Geographic and Race/Ethnicity Differences in Patient Perceptions of Diabetes. Journal of Primary Care & Community Health. 2019;10:2150132719845819. Tenkorang EY. Early onset of type 2 diabetes among visible minority and immigrant populations in Canada. Ethnicity & Health. 2017;22:266–84. Knott C, Bell S, Britton A. Alcohol Consumption and the Risk of Type 2 Diabetes: A Systematic Review and Dose-Response Meta-analysis of More Than 1.9 Million Individuals From 38 Observational Studies. Diabetes Care. 2015;38:1804–12. Śliwińska-Mossoń M, Milnerowicz H. The impact of smoking on the development of diabetes and its complications. Diabetes and Vascular Disease Research. 2017;14:265–76. Verma S, Hussain ME. Obesity and diabetes: An update. Diabetes & Metabolic Syndrome: Clinical Research & Reviews. 2017;11:73–9. Garrett C, Doherty A. Diabetes and mental health. Clinical Medicine. 2014;14:669–72. Robinson DJ, Coons M, Haensel H, Vallis M, Yale J-F. Diabetes and Mental Health. Canadian Journal of Diabetes. 2018;42:S130–41. Ford KJ, Robitaille A. How sweet is your love? Disentangling the role of marital status and quality on average glycemic levels among adults 50 years and older in the English Longitudinal Study of Ageing. BMJ Open Diabetes Research & Care. 2023;11. Zhang X, Bullard KM, Gregg EW, Beckles GL, Williams DE, Barker LE, et al. Access to Health Care and Control of ABCs of Diabetes. Diabetes Care. 2012;35:1566–71. Cuadros DF, Li J, Musuka G, Awad SF. Spatial epidemiology of diabetes: Methods and insights. World Journal of Diabetes. 2021;12:1042–56. Kwan M-P. The stationarity bias in research on the environmental determinants of health. Health & Place. 2021;70:102609. Ge H, Wang J. Spatial Non-Stationarity Effects of Unhealthy Food Environments and Green Spaces for Type-2 Diabetes in Toronto. Sustainability. 2023;15:1762. Kauhl B, Schweikart J, Krafft T, Keste A, Moskwyn M. Do the risk factors for type 2 diabetes mellitus vary by location? A spatial analysis of health insurance claims in Northeastern Germany using kernel density estimation and geographically weighted regression. International Journal of Health Geographics. 2016;15:38. Georganos S, Grippa T, Niang Gadiaga A, Linard C, Lennert M, Vanhuysse S, et al. Geographical random forests: a spatial extension of the random forest algorithm to address spatial heterogeneity in remote sensing and population modelling. Geocarto International. 2021;36:121–36. Georganos S, Kalogirou S. A Forest of Forests: A Spatially Weighted and Computationally Efficient Formulation of Geographical Random Forests. ISPRS International Journal of Geo-Information. 2022;11:471. Quiñones S, Goyal A, Ahmed ZU. Geographically weighted machine learning model for untangling spatial heterogeneity of type 2 diabetes mellitus (T2D) prevalence in the USA. Scientific Reports. 2021;11:6955. Lotfata A, Moosazadeh M, Helbich M, Hoseini B. Socioeconomic and environmental determinants of asthma prevalence: a cross-sectional study at the U.S. County level using geographically weighted random forests. International Journal of Health Geographics. 2023;22:18. Lotfata A, Grekousis G, Wang R. Using geographical random forest models to explore spatial patterns in the neighborhood determinants of hypertension prevalence across chicago, illinois, USA. Environment and Planning B: Urban Analytics and City Science. 2023;50:2376–93. Wu D, Zhang Y, Xiang Q. Geographically weighted random forests for macro-level crash frequency prediction. Accident Analysis & Prevention. 2024;194:107370. Kang Y, Zhang F, Gao S, Lin H, Liu Y. A review of urban physical environment sensing using street view imagery in public health studies. Annals of GIS. 2020;26:261–75. Nguyen QC, Khanna S, Dwivedi P, Huang D, Huang Y, Tasdizen T, et al. Using Google Street View to examine associations between built environment characteristics and U.S. health outcomes. Preventive Medicine Reports. 2019;14:100859. Nguyen QC, Keralis JM, Dwivedi P, Ng AE, Javanmardi M, Khanna S, et al. Leveraging 31 Million Google Street View Images to Characterize Built Environments and Examine County Health Outcomes. Public Health Reports. 2021;136:201–11. Wei J, Yue W, Li M, Gao J. Mapping human perception of urban landscape from street-view images: A deep-learning approach. International Journal of Applied Earth Observation and Geoinformation. 2022;112:102886. He J, Zhang J, Yao Y, Li X. Extracting human perceptions from street view images for better assessing urban renewal potential. Cities. 2023;134:104189. Nguyen QC, Belnap T, Dwivedi P, Deligani AHN, Kumar A, Li D, et al. Google Street View Images as Predictors of Patient Health Outcomes, 2017–2019. Big Data and Cognitive Computing. 2022;6:15. Nguyen TT, Nguyen QC, Rubinsky AD, Tasdizen T, Deligani AHN, Dwivedi P, et al. Google Street View-Derived Neighborhood Characteristics in California Associated with Coronary Heart Disease, Hypertension, Diabetes. International Journal of Environmental Research and Public Health. 2021;18:10428. Keralis JM, Javanmardi M, Khanna S, Dwivedi P, Huang D, Tasdizen T, et al. Health and the built environment in United States cities: measuring associations using Google Street View-derived indicators of the built environment. BMC Public Health. 2020;20:215. Painter K. The influence of street lighting improvements on crime, fear and pedestrian street use, after dark. Landscape and Urban Planning. 1996;35:193–201. Vlahov D, Freudenberg N, Proietti F, Ompad D, Quinn A, Nandi V, et al. Urban as a Determinant of Health. Journal of Urban Health. 2007;84:16–26. Seresinhe CI, Preis T, Moat HS. Quantifying the Impact of Scenic Environments on Health. Scientific Reports. 2015;5:16899. Zhang F, Zhou B, Liu L, Liu Y, Fung HH, Lin H, et al. Measuring human perceptions of a large-scale urban region using machine learning. Landscape and Urban Planning. 2018;180:148–60. Galea S, Ahern J, Rudenstine S, Wallace Z, Vlahov D. Urban built environment and depression: a multilevel analysis. Journal of Epidemiology & Community Health. 2005;59:822–7. Wang R, Helbich M, Yao Y, Zhang J, Liu P, Yuan Y, et al. Urban greenery and mental wellbeing in adults: Cross-sectional mediation analyses on multiple pathways across different greenery measures. Environmental Research. 2019;176:108535. Won J, Lee C, Forjuoh SN, Ory MG. Neighborhood safety factors associated with older adults’ health-related outcomes: A systematic literature review. Social Science & Medicine. 2016;165:177–86. Wu J, Li Q, Feng Y, Bhuyan SS, Tarimo CS, Zeng X, et al. Active commuting and the risk of obesity, hypertension and diabetes: a systematic review and meta-analysis of observational studies. BMJ Global Health. 2021;6. Zhou H, Wang J, Widener M, Wilson K. Examining the relationship between active transport and exposure to streetscape diversity during travel: A study using GPS data and street view imagery. Computers, Environment and Urban Systems. 2024;110:102105. Zhou H, Wang J, Wilson K. Impacts of perceived safety and beauty of park environments on time spent in parks: Examining the potential of street view imagery and phone-based GPS data. International Journal of Applied Earth Observation and Geoinformation. 2022;115:103078. Xu J, Liu Y, Liu Y, An R, Tong Z. Integrating street view images and deep learning to explore the association between human perceptions of the built environment and cardiovascular disease in older adults. Social Science & Medicine. 2023;338:116304. Liu J, Yang L, Xiao L, Tao Z. Perceived Neighborhood Environment Impacts on Health Behavior, Multi-Dimensional Health, and Life Satisfaction. Frontiers in Public Health. 2022;10. Bennett GG, McNeill LH, Wolin KY, Duncan DT, Puleo E, Emmons KM. Safe To Walk? Neighborhood Safety and Physical Activity Among Public Housing Residents. PLOS Medicine. 2007;4:e306. Statistics Canada. Census Profile, 2021 Census of Population [Internet]. 2022 [cited 2024 Jan 6]. Available from: https://www12.statcan.gc.ca/census-recensement/2021/dp-pd/prof/index.cfm?Lang=E Environics Analytics [Internet]. 2024 [cited 2024 Oct 31]. Available from: https://environicsanalytics.com/en-ca/data/health-databases/communityhealth Government of Canada SC. Canadian Community Health Survey (CCHS) - 2018 [Internet]. 2018 [cited 2025 Jan 18]. Available from: https://www23.statcan.gc.ca/imdb/p3Instr.pl?Function=assembleInstr&a=1&&lang=en&Item_Id=839130 Wu DY, Wang J. Analyzing Urban Crime Through Street View Imagery: Insights from Urban Micro Built Environment and Perceptions. Urban Science. 2024;8:247. Dubey A, Naik N, Parikh D, Raskar R, Hidalgo CA. Deep Learning the City : Quantifying Urban Perception At A Global Scale [Internet]. arXiv; 2016 [cited 2023 Oct 21]. Available from: http://arxiv.org/abs/1608.01769 Google Maps. Google Maps Street View [Internet]. Google Maps Street View. 2024 [cited 2024 Feb 21]. Available from: https://www.google.com/streetview/ Zhao H, Shi J, Qi X, Wang X, Jia J. Pyramid Scene Parsing Network [Internet]. arXiv; 2017 [cited 2023 Oct 21]. Available from: http://arxiv.org/abs/1612.01105 Cordts M, Omran M, Ramos S, Rehfeld T, Enzweiler M, Benenson R, et al. The Cityscapes Dataset for Semantic Urban Scene Understanding [Internet]. arXiv; 2016 [cited 2023 Oct 21]. Available from: http://arxiv.org/abs/1604.01685 Guo J, He H, He T, Lausen L, Li M, Lin H, et al. GluonCV and GluonNLP: Deep Learning in Computer Vision and Natural Language Processing [Internet]. arXiv; 2020 [cited 2023 Oct 21]. Available from: http://arxiv.org/abs/1907.04433 Landsat 8 | U.S. Geological Survey [Internet]. 2025 [cited 2025 Jan 18]. Available from: https://www.usgs.gov/landsat-missions/landsat-8 Canada NR. Lakes, Rivers and Glaciers in Canada - CanVec Series - Hydrographic Features - Open Government Portal [Internet]. 2019 [cited 2025 Jan 18]. Available from: https://open.canada.ca/data/en/dataset/9d96e8c9-22fe-4ad2-b5e8-94a6991b744b Ontario Road Network [Internet]. 2022 [cited 2025 Jan 18]. Available from: https://geohub.lio.gov.on.ca/datasets/mnrf::ontario-road-network-orn-road-net-element/about Toronto C of. Sidewalk Inventory [Internet]. City of Toronto. City of Toronto; 2017 [cited 2025 Jan 18]. Available from: https://www.toronto.ca/services-payments/streets-parking-transportation/sidewalk-tours-wayfinding/sidewalks-in-toronto/sidewalk-inventory/ Kwan M-P. The Uncertain Geographic Context Problem. Annals of the Association of American Geographers. 2012;102:958–68. Vallée J, Cadot E, Grillo F, Parizot I, Chauvin P. The combined effects of activity space and neighbourhood of residence on participation in preventive health-care activities: The case of cervical screening in the Paris metropolitan area (France). Health & Place. 2010;16:838–52. Moreno C, Allam Z, Chabaud D, Gall C, Pratlong F. Introducing the “15-Minute City”: Sustainability, Resilience and Place Identity in Future Post-Pandemic Cities. Smart Cities. 2021;4:93–111. Chen X, Ye X, Widener MJ, Delmelle E, Kwan M-P, Shannon J, et al. A systematic review of the modifiable areal unit problem (MAUP) in community food environmental research. Urban Informatics. 2022;1:22. Tompkins JW, Luginaah IN, Booth GL, Harris SB. The Geography of Diabetes in London, Canada: The Need for Local Level Policy for Prevention and Management. International Journal of Environmental Research and Public Health. 2010;7:2407–22. Luo Y, Yan J, McClure S. Distribution of the environmental and socioeconomic risk factors on COVID-19 death rate across continental USA: a spatial nonlinear analysis. Environmental Science and Pollution Research. 2021;28:6587–99. Greenwell BM. pdp: An R Package for Constructing Partial Dependence Plots. The R Journal. 2017;9:421–36. Yousefzadeh Barri E, Farber S, Kramer A, Jahanshahi H, Allen J, Beyazit E. Can transit investments in low-income neighbourhoods increase transit use? Exploring the nexus of income, car-ownership, and transit accessibility in Toronto. Transportation Research Part D: Transport and Environment. 2021;95:102849. Chaix B, Kestens Y, Duncan S, Merrien C, Thierry B, Pannier B, et al. Active transportation and public transportation use to achieve physical activity recommendations? A combined GPS, accelerometer, and mobility survey study. Int J Behav Nutr Phys Act. 2014;11:124. Chu A, Shah BR, Rashid M, Booth GL, Fazli GS, Tu K, et al. Trends in glucose testing among individuals without diabetes in Ontario between 2010 and 2017: a population-based cohort study. Canadian Medical Association Open Access Journal. 2022;10:E772–80. Wilson SE, Rosella LC, Lipscombe LL, Manuel DG. The effectiveness and efficiency of diabetes screening in Ontario, Canada: a population-based cohort study. BMC Public Health. 2010;10:506. Lorraine L. Lipscombe, Fatema M. Ali, Krista Lamb, Laura C. Rosella, Baiju R. Shah, Ian S. Zenlea. Current State of Type 2 Diabetes in the Peel Region [Internet]. Novo Nordisk Network for Healthy Populations; 2024. Available from: https://healthypopulationsnetwork.utoronto.ca/sites/default/files/assets/files/nhp-reportcurrent-state-type-2-diabetes-peel-region.pdf Pasala SK, Rao AA, Sridhar GR. Built environment and diabetes. Int J Diabetes Dev Ctries. 2010;30:63–8. de Miguel-Yanes JM, Shrader P, Pencina MJ, Fox CS, Manning AK, Grant RW, et al. Genetic Risk Reclassification for Type 2 Diabetes by Age Below or Above 50 Years Using 40 Type 2 Diabetes Risk Single Nucleotide Polymorphisms. Diabetes Care. 2010;34:121–5. Kirkman MS, Briscoe VJ, Clark N, Florez H, Haas LB, Halter JB, et al. Diabetes in Older Adults. Diabetes Care. 2012;35:2650–64. Berggreen-Clausen A, Pha SH, Alvesson HM, Andersson A, Daivadanam M. Food environment interactions after migration: a scoping review on low- and middle-income country immigrants in high-income countries. Public Health Nutrition. 2022;25:136–58. Pandey M, Kamrul R, Michaels CR, McCarron M. Identifying Barriers to Healthcare Access for New Immigrants: A Qualitative Study in Regina, Saskatchewan, Canada. Journal of Immigrant and Minority Health. 2022;24:188–98. Sundareswaran M, Martignetti L, Purkey E. Barriers to primary care among immigrants and refugees in Peterborough, Ontario: a qualitative study of provider perspectives. BMC Primary Care. 2024;25:199. Bajgain BB, Bajgain KT, Badal S, Aghajafari F, Jackson J, Santana M-J. Patient-Reported Experiences in Accessing Primary Healthcare among Immigrant Population in Canada: A Rapid Literature Review. International Journal of Environmental Research and Public Health. 2020;17:8724. Bopp M, Gayah VV, Campbell ME. Examining the Link Between Public Transit Use and Active Commuting. International Journal of Environmental Research and Public Health. 2015;12:4256–74. Additional Declarations No competing interests reported. Supplementary Files RevisedPlainAppendix20250410.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5984185","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":441165221,"identity":"5041a977-7415-460d-b632-ecac6af49cca","order_by":0,"name":"Haoxuan Ge","email":"","orcid":"","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Haoxuan","middleName":"","lastName":"Ge","suffix":""},{"id":441165222,"identity":"edd2bac7-52a6-475a-8302-9d84cd1d00d2","order_by":1,"name":"Jue Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3klEQVRIiWNgGAWjYDACZijNT7oWyQYIS4J4rQYHiNUi787+gJmn5o7d5hv5Bx9XVNyp42dgfvgBnxbDwzwGzDzHniVvu5HMbHjmzDMJyQY2Y7xWGTbzMDDnsB1ONruRzCbZ2HZYwuAAD37XGTYDHZbz73Cy8QyEFuYfeP3CzGDAnNt22M5AAqGFDa8tQI8YHP7bdzhB4sxjY8OGM4clZzazmVngtaX/+MOHM74dtudvT3z4sKHiMD8/e/PjG3htOcDAAEQMiQ1wIWYcSuG2QJXaE1A3CkbBKBgFIxkAAPeJRZxPzbT/AAAAAElFTkSuQmCC","orcid":"","institution":"University of Toronto","correspondingAuthor":true,"prefix":"","firstName":"Jue","middleName":"","lastName":"Wang","suffix":""},{"id":441165223,"identity":"c88b7bb9-914d-4a17-b3a3-8345a3a06543","order_by":2,"name":"Devin Yongzhao Wu","email":"","orcid":"","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Devin","middleName":"Yongzhao","lastName":"Wu","suffix":""},{"id":441165224,"identity":"a8020b84-a9d4-424d-9fd0-c81deead8045","order_by":3,"name":"Hanlin Zhou","email":"","orcid":"","institution":"University of Connecticut","correspondingAuthor":false,"prefix":"","firstName":"Hanlin","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2025-02-07 23:38:06","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5984185/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5984185/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":80394700,"identity":"20c3ac5f-0f50-4279-9f69-f5e8636075a6","added_by":"auto","created_at":"2025-04-11 12:14:46","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":185094,"visible":true,"origin":"","legend":"\u003cp\u003eAnalysis workflow of the study\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/d9e9f7d5dbd7147ca581ac2f.png"},{"id":80395871,"identity":"e7fa7c65-6b38-4e74-bffd-84141b9183f5","added_by":"auto","created_at":"2025-04-11 12:30:46","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":615159,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial distribution of lifestyle factors: \u003cstrong\u003ea\u003c/strong\u003e) T2DM prevalence rate (%), \u003cstrong\u003eb\u003c/strong\u003e) obesity rate (%); \u003cstrong\u003ec\u003c/strong\u003e) smocking rate (%); \u003cstrong\u003ed\u003c/strong\u003e) drinking rate (%); \u003cstrong\u003ee\u003c/strong\u003e) weekly recreation physical activity minutes; \u003cstrong\u003ef\u003c/strong\u003e) access to regular health care provide (%); \u003cstrong\u003eg\u003c/strong\u003e) poor mental health (%); \u003cstrong\u003eh\u003c/strong\u003e) life dissatisfaction (%) \u003cstrong\u003ei\u003c/strong\u003e) anxiety disorder (%).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/554b03a84e31b9d0fef0c2b0.png"},{"id":80394699,"identity":"10fe7e23-022b-4214-b238-cbebc71e68ea","added_by":"auto","created_at":"2025-04-11 12:14:46","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":620313,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial distribution of environmental factors: \u003cstrong\u003ea\u003c/strong\u003e) road network density coverage (m/m\u003csup\u003e2\u003c/sup\u003e); \u003cstrong\u003eb\u003c/strong\u003e) cycling network density coverage (m/m\u003csup\u003e2\u003c/sup\u003e); \u003cstrong\u003ec\u003c/strong\u003e) water body density coverage (% area); \u003cstrong\u003ed\u003c/strong\u003e) street view beauty and safety perception; \u003cstrong\u003ee\u003c/strong\u003e) street view wall coverage; \u003cstrong\u003ef\u003c/strong\u003e) street view fence coverage; \u003cstrong\u003eg\u003c/strong\u003e) street view pole coverage; \u003cstrong\u003eh\u003c/strong\u003e) street view traffic light coverage; \u003cstrong\u003ei\u003c/strong\u003e) NDVI.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/c734950b7591d5a0966261ab.png"},{"id":80396614,"identity":"97905377-2435-48ad-8754-66b858ce814d","added_by":"auto","created_at":"2025-04-11 12:38:47","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":394845,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial distribution of socioeconomic factors: \u003cstrong\u003ea\u003c/strong\u003e) age; \u003cstrong\u003eb\u003c/strong\u003e) married or common law (%); \u003cstrong\u003ec\u003c/strong\u003e) median income (CAD); \u003cstrong\u003ed\u003c/strong\u003e) non-Canadian citizens (%); \u003cstrong\u003ee\u003c/strong\u003e) immigrant population (%); \u003cstrong\u003ef\u003c/strong\u003e) low education rate (%); \u003cstrong\u003eg\u003c/strong\u003e) unemployment rate; \u003cstrong\u003eh\u003c/strong\u003e) commuting by public transit (%); \u003cstrong\u003ei\u003c/strong\u003e) commuting by walking and cycling (%); \u003cstrong\u003ej\u003c/strong\u003e) weekly commuting duration (minutes)\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/c335a4ed009cb6f51fe80a9d.png"},{"id":80395121,"identity":"c3747cc7-543c-4a54-b777-e55b7c9d1eef","added_by":"auto","created_at":"2025-04-11 12:22:46","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":122014,"visible":true,"origin":"","legend":"\u003cp\u003eLocal permutation feature importance for the GWRF model\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/106996cb88314f7459d83579.png"},{"id":80394709,"identity":"85bef379-888e-4980-934a-61ebd1810b68","added_by":"auto","created_at":"2025-04-11 12:14:47","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":322989,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of local permutation feature importance for socioeconomic factors \u003cstrong\u003e(a to j)\u003c/strong\u003e; lifestyle factors \u003cstrong\u003e(k to r)\u003c/strong\u003e; and environmental factors \u003cstrong\u003e(s\u003c/strong\u003e \u003cstrong\u003eto aa) \u003c/strong\u003eacross Toronto\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/3787d195d17c536c26735704.png"},{"id":80395124,"identity":"760a4943-dfde-4d13-9897-23a855f9bdf0","added_by":"auto","created_at":"2025-04-11 12:22:47","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":339958,"visible":true,"origin":"","legend":"\u003cp\u003eDominant local permutation feature importance across Toronto\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/f1d6c4da21da9d48f570aabf.png"},{"id":80395873,"identity":"c2ab21b2-aa82-4227-9ebd-180620d9ad29","added_by":"auto","created_at":"2025-04-11 12:30:47","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":78285,"visible":true,"origin":"","legend":"\u003cp\u003ePartial dependence plots for the GWRF model on predicted averaged T2DM prevalence rate with independent variables: socioeconomic factors \u003cstrong\u003e(a to j)\u003c/strong\u003e; lifestyle factors \u003cstrong\u003e(k to r)\u003c/strong\u003e; and environmental factors \u003cstrong\u003e(s\u003c/strong\u003e \u003cstrong\u003eto aa)\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/f6aab699ebc1d15eecdb1287.png"},{"id":81660133,"identity":"0d325cc2-d503-4b7f-ad5c-15bdc841892c","added_by":"auto","created_at":"2025-04-29 21:31:18","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3610987,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/ae9975f4-379e-41b4-99d2-0eaa7beff3b4.pdf"},{"id":80394705,"identity":"450ad740-ac56-458f-b98c-ecc8d06a1d3f","added_by":"auto","created_at":"2025-04-11 12:14:46","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":612838,"visible":true,"origin":"","legend":"","description":"","filename":"RevisedPlainAppendix20250410.docx","url":"https://assets-eu.researchsquare.com/files/rs-5984185/v1/cee92f13f62eed9af7583572.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Geographically Weighted Random Forest Analysis of Spatial Non-Stationarity Association of Street View Environmental, Socioeconomic, and Lifestyle Factors with Type-2 Diabetes Prevalence in Toronto","fulltext":[{"header":"1. Background","content":"\u003cp\u003eType-2 Diabetes Mellitus (T2DM) is becoming one of the most extensive global healthcare emergencies, and its prevalence is increasing worldwide. Taking Toronto, Canada, as an example, more than one in eight adults have T2DM, and this rate has almost doubled over the past decade [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Public health and medical researchers have studied how environmental factors and exposures significantly shape people\u0026rsquo;s daily lifestyles and behaviours. They are hypothesized to correlate with the risk of T2DM by influencing psychosocial stressors and physical activities [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The availability or proximity to recreational resources, green spaces, walkways, open spaces, and well-designed public areas encourages physical activity. Residents would prefer exercising in walkable neighbourhoods with pleasing environments, which reduces the risk of T2DM [\u003cspan additionalcitationids=\"CR5 CR6 CR7 CR8 CR9 CR10\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Neighbourhoods with strong social networks and safer, greener environments will also improve mental health and encourage physical activity [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. However, an unpleasant environment can discourage healthy lifestyles. Neighbourhoods with limited access to supermarkets, increasing residential noise, air pollution, traffic, and proximity to roads may lead to greater reliance on fast food outlets, elevated chronic stress, and inflammation of blood vessels due to oxidative stress [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan additionalcitationids=\"CR14 CR15 CR16 CR17 CR18 CR19\" citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Crime, social disorders, and unsafe neighbourhoods will promote social isolation, anxiety, fear, and a reduction in outdoor activity time. These factors and other combined biological, behavioural, and socioeconomic status contribute to insulin resistance and increase the risk of T2DM.\u003c/p\u003e \u003cp\u003eAside from environmental effects, existing research has recognized the critical role of socioeconomic status and lifestyle factors in influencing the T2DM prevalence rate. Social and healthy lifestyle determinants have also been prioritized as a top consideration in T2DM clinical treatment and care, alongside traditional medical prevention strategies [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Various studies show that older adults and the unemployed have higher risks of diagnosing T2DM [\u003cspan additionalcitationids=\"CR23\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Lower education and income are linked to higher rates of T2DM [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Regarding ethnicity, recent studies found that the non-Hispanic white population has a lower rate of T2DM prevalence rate compared with other ethnic groups [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. For example, South Asian and black immigrants in Canada have been reported to face a higher likelihood of developing T2DM earlier in their life course compared to immigrants from the UK [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Lifestyle factors such as smoking, heavy alcohol consumption, and obesity are well-known contributors to T2DM risk [\u003cspan additionalcitationids=\"CR31\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Psychological and emotional stress also play a significant role. Individuals experiencing depression and anxiety may find it challenging to maintain self-care routines, including monitoring blood glucose levels, maintaining a healthy diet, and engaging in regular physical activity [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Marital relationships may further affect T2DM management by worsening glycemic control through increasing stress [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Additionally, access to regular healthcare is vital for the effectiveness of T2DM management. Consistent interaction with healthcare providers will improve health outcomes, effectively preventing and managing T2DM complications [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. Ultimately, the interplay between socioeconomic status and lifestyle behaviours creates a synergistic effect significantly influencing T2DM prevalence and management.\u003c/p\u003e \u003cp\u003eVarious measurement methods on how the socioeconomic, environmental, and lifestyle context impacts T2DM health outcomes were utilized in past studies, including generalized linear models, spatial estimations, and Bayesian approaches [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. Nevertheless, research often assumes that the effects are spatially stationary and do not differ at different geographic locations. Environmental and socio-determinants usually have inconsistent associations with T2DM or other health outcomes [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. Traditional, non-spatial approaches likely cause biased model performance and lead to erroneous. For instance, green space density and income were positively correlated with the T2DM prevalence rate in one part of the city but negatively correlated with the T2DM prevalence rate in other neighbourhoods [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. Recognizing spatial non-stationarity effects further helps to investigate the ignored aspects and improve our understanding of the spatial phenomena on determinates of health outcomes.\u003c/p\u003e \u003cp\u003eTo address these limitations, recent advancements in spatial analytical methods have introduced advanced models that account for spatial heterogeneity. Once such model is Geographically Weighted Random Forest (GWRF), a machine learning-based extension of the Random Forest algorithm designed to capture spatial non-stationarity. GWRF can adaptively capture and interpret intricate patterns of spatially diverse dataset, accept non-parametric predictors, alleviate overfitting issues with its bootstrapping nature, and retain explainable quantification for the spatial variation of localized relationships [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. The model has recently gained attention in health geography for its effectiveness in modelling spatially varying associations with improved performance compared to non-machine learning based spatial models. For example, Qui\u0026ntilde;ones et al. (2021) applied GWRF to explain spatial non-stationary effects and predict T2DM prevalence through obesity, physical activity, food environment, and socioeconomic data across the United States [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. Lotfata et al. (2023) explored how environmental and socioeconomic factors influence asthma prevalence differently across the United States using GWRF [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Grekousis et al. (2023) used GWRF to estimate the COVID-19 death rate through socioeconomic and health-related factors in the United States. These studies highlight the growing role of GWRF in health research, where it often outperforms ordinary least squares, conventional random forest, and geographically weighted regression models in terms of predictive accuracy and spatial interpretability [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDespite researchers striving to unravel the intricate connections between the environment and human health outcomes for decades, a notable gap in our knowledge currently surrounds the influence of human perceptions of the environment on health outcomes. Human perceptions and feelings about the environment represent how people view and interpret their contextual surroundings. Few studies have explored the relationship between human perceptions of the environment and T2DM prevalence across large-scale geographical areas, as challenges arise with large-scale collection of human perceptions. The techniques, such as environmental audits, are usually labour-intensive and require significant time [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. This can be driving around communities to measure the greenness in each neighbourhood or inquiring about residents\u0026rsquo; perceptions through surveys and interviews. With the rapid advancement in machine learning algorithms, there is a growing trend toward harnessing computer vision and big data images, such as Street View Imagery (SVI), to quantify environmental characteristics and understand their impacts on health outcomes [\u003cspan additionalcitationids=\"CR48\" citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. SVI captures the environment at eye level with 360-degree panoramic pictures, revealing real-world scenery from a pedestrian perspective. Furthermore, SVI can be used to extract both objective and subjective measurements of environmental features. Objective measurements include roads, humans, sidewalks, vehicles, constructions, objects (traffic poles, lights, and signs), proportion of nature (vegetation and terrain), and sky. In contrast, perception variables can be human subjectivity of their feelings about the environment, including safety, beauty, liveliness, dullness, depression, and a wealthy environment [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe urban environment is a complex system in which various environmental features are interconnected and jointly influence resident\u0026rsquo;s health outcomes and behaviours. Studies have shown that the characteristics of street environments, such as greenery, building types, pedestrian infrastructure, road design, and street connectivity, were closely associated with residents\u0026rsquo; health behaviours and the prevalence of chronic diseases, including high blood pressure, high cholesterol, diabetes, and heart disease [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e, \u003cspan additionalcitationids=\"CR53\" citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. In particular, studies using SVI identified single-lane roads and visible utility wires, which are markers of older or lower-income neighbourhoods, as being associated with an increased risk of diabetes and obesity [\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. Street lighting poles improve visibility at night, reducing crime and injury and increasing perceptions of safety. A well-lit street may encourage evening walks or other outdoor activities, whereas poorly lit areas can deter people from being active after dark [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. Meanwhile, non-single-family home buildings were associated with decreased obesity, diabetes, and physical inactivity [\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. Physical barriers such as fences may limit access to public green spaces, thereby affecting their willingness to conduct physical activity and overall health. In contrast, perceptions of a beautiful, lively environment have been shown to benefit health outcomes, especially mental health [\u003cspan additionalcitationids=\"CR57\" citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e]. Greener, safer, wealthier, and more beautiful environments can reduce depression and stress [\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e]. For Example, environments perceived as beautiful, safe, and lively are associated with lower psychological stress, better mental health, and increased physical activity through active transportation, contributing to reduced chronic disease risk [\u003cspan additionalcitationids=\"CR62 CR63 CR64\" citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e]. Better-perceived neighbourhood environments also significantly correlate with reduced sedentary behaviour, smoking, and drinking, promoting healthier lifestyles and greater life satisfaction [\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e]. Conversely, unsafe neighbourhood perceptions may foster social isolation and fear, leading to reduced physical activities and hindering the prevention of chronic diseases like T2DM, which requires sustained engagement in healthy behaviours [\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e]. Therefore, the environment's objective and subjective characteristics can directly or indirectly influence people\u0026rsquo;s health outcomes.\u003c/p\u003e \u003cp\u003eAlthough much research has identified that street view-derived environmental characteristics, socioeconomic factors, and lifestyle habits affect health outcomes by influencing behaviours, few studies have investigated their associations with T2DM prevalence in a comprehensive, city-level context from a geospatial perspective. Further, to the best of our knowledge, no research assesses the impact of neighbourhood environment perceptions on T2DM prevalence. The use of GWRF to explain spatial-non-stationarity effects of T2DM prevalence has also not been extensively explored. To address these gaps, this research uses SVI-derived environmental, socioeconomic, and lifestyle factors to investigate how they affect T2DM at the Dissemination Areas level in Toronto through Geographically Weighted Random Forest modelling. The goals are twofold: 1) To explore the relationships between SVI-derived environmental, socioeconomic, and lifestyle factors and T2DM prevalence rates, and 2) To determine how these associations vary across the city of Toronto.\u003c/p\u003e"},{"header":"2. Methodology","content":"\u003cp\u003eThis study examines the spatial effect of SVI-derived environmental, socioeconomic, and lifestyle determinants on T2DM prevalence rates across 3800 Dissemination Areas (DA) in Toronto. A Geographically Weighted Random Forest (GWRF) regression model is applied to address spatial autocorrelation and investigate global and spatial associations between the factors with T2DM prevalence. This is achieved through graphing partial dependency plots and mapping the importance of local permutation features. A graphical summary of this study\u0026rsquo;s workflow is provided in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Study Area\u003c/h2\u003e \u003cp\u003eLocated in the southern region of Ontario, Toronto serves as Ontario\u0026rsquo;s capital city. Covering a total land area of 630.2 km\u003csup\u003e2\u003c/sup\u003e. It is home to three million culturally diverse citizens residing in 3800 officially registered Dissemination Areas [\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e]. Toronto\u0026rsquo;s environmental characteristics vary across neighbourhoods, ranging from dense urban centers in the central and southern areas to extensive green lands in the east. Its socioeconomic factors, such as income levels, education attainment, and access to healthcare, also exhibit significant spatial variation across the city. This study, therefore, conducts an observational study covering Toronto\u0026rsquo;s DA to investigate the social and environmental determinants associated with the prevalence of T2DM.\u003c/p\u003e\u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Data\u003c/h2\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003e2.2.1 T2DM Prevalence and Health-Related Factors\u003c/h2\u003e \u003cp\u003eThe 2017/2018 T2DM prevalence data for the Toronto DA used in this study is collected and modelled by Environics Analytics as part of their Community Health Collection [\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e]. The Community Health data was derived from the 2017/2018 Canadian Community Health Survey and modelled using the 2021 Census demographic data to ensure it is representative of all census units across Canada by Environics Analytics [\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e, \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e]. This dataset includes survey responses from participants aged 12 and above who reported being diagnosed with T2DM. This study also uses additional health factors from the Community Health dataset to explore their associations with T2DM and control for potential confounding variables. The factors include obesity prevalence (for body weight index higher than 30), smoking rate, drinking rate, percentage of population with satisfaction of life reported as very dissatisfied, perceived poor mental health, perceived anxiety disorders (including phobia, obsessive-compulsive disorder, and panic disorder), weekly physical activity minutes, and access to regular health care providers. Each data record was matched and spatially joined with the Toronto DA boundary shapefile for spatial analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003e2.2.2 Environmental-Related Factors\u003c/h2\u003e \u003cp\u003eTo investigate the street view environmental factors associated with T2DM, this study incorporates variables derived from street view, GIS vector data, and remote sensing images. The analyzed factors include perceived environmental beauty and safety, fence density, pole density, traffic light density, wall density, road density, cycling network density, water body density, and Normalized Difference Vegetation Index (NDVI) within each DA.\u003c/p\u003e \u003cp\u003eThis study used a pre-trained machine-learning model to predict perceived safety and beauty across Toronto DA [\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e, \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e]. The model is trained by crowdsourced perception data named Place Pulse V2, including 1,170,000 pairwise companions of 110,988 SVI from 81,630 volunteers [\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e]. Street View Images (SVI) around Toronto taken in 2016 were included in the analysis [\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e]. Every sample location is spaced 50 meters apart, with four images assessed from view angles: North, East, South, and West, to obtain a panoramic view of the surrounding environments at each location. The predicted perception score ranged from 0 to 1, where the higher value means a higher likelihood of being perceived as beautiful or safe. After perception calculations, this study aggregated the scores by taking the average perception scores from four view angles for each location. As the beauty and safety perception scores were highly correlated, with a Pearson Correlation Coefficient of 0.98, including both variables in a model could create reduced model performance and biased interpretations. To address this issue and preserve the joint contribution of both perception measures, we combined the two scores into a single variable by multiplying them. Areas perceived as both beautiful and safe are more likely to support confortable and active engagement with the environment. This approach emphasizes locations where both variables are high while down-weighting locations where either perception is low. To further obtain built environmental characteristics in Toronto, the proportion of fences, poles, traffic lights, and walls viewed in each image along road networks were derived from SVI using the Pyramid Scene Parsing Network (PSPNet) [\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e, \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e]. Scene parsing employs computer image recognition to classify each pixel with a category label. PSPNet is a deep convolutional neural network-based model that captures details of raster images at different scales and combines extracted components to segment the micro-built environment [\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e]. A pre-trained PSPNet with a Cityscape dataset including 25,000 labelled images focusing on urban street scenes was obtained from GluonCV and used in this study [\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e, \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAlthough SVI captures most of the built environment along road networks, some areas remain out of reach from the road network, such as parks and community gardens, without SVI coverage. Therefore, this study also calculates the Normalized Difference Vegetation Index (NDVI) to include in the regression model using Landsat 8 Collection 2, Level 2, Tier 1 imagery captured on June 11, 2018 [\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e]. Besides NDVI, the study also includes additional traditional environmental variables, including road density, cycling network density, and water body density, calculated by dividing their respective lengths or areas (in meters or square meters) by the DA (in square meters). Road network data was sourced from the 2022 Ontario Road Network Composite, cycling network data was obtained from the 2019 City of Toronto Physical Area of Sidewalks, and water body density was derived from the 2019 Government of Canada\u0026rsquo;s Lakes, Rivers, and Glaciers Hydrographic CanVec Series [\u003cspan additionalcitationids=\"CR79\" citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e]. Access to environmental facilities and built environment exposures could happen beyond existing administrative boundaries, characterized as part of the Uncertain Geographic Context Problem (UGCoP) in spatial analysis [\u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e]. People\u0026rsquo;s activities do not occur entirely within their residential units; they may spend time outside their home neighbourhoods [\u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e82\u003c/span\u003e]. Incorporating the 15-minute city concept from urban planning, this study thus expanded the boundary of each DA from its centroid using a 15-minute walking service area delineated with ArcGIS Pro\u0026rsquo;s network analysis tool (version 3.4.0) [\u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e83\u003c/span\u003e]. All features within the 15-minute walking range, including those extending beyond Toronto\u0026rsquo;s administrative boundaries, were summarized and standardized by the total area after the boundary expansion, accounting for the \u0026ldquo;edge effect\u0026rdquo; when facilities can be accessed beyond the given boundary [\u003cspan citationid=\"CR84\" class=\"CitationRef\"\u003e84\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.2.3 Socioeconomic Status\u003c/h2\u003e \u003cp\u003eTo explore the associations of socioeconomic status with T2DM, this study included age, income, non-citizens population, unemployment rate, population without an education diploma or certificate, marriage and common law, immigration population, commute duration, population commuting by transit, and population commuting by walking and cycling in the regression model. The variables were all obtained from the 2021 Canadian Census at the DA level [\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e]. About 2% of the DA units do not have reported census and health status due to their small geographical and population size, which is intended to protect residents\u0026rsquo; privacy. As a result, the study calculates the average values of their immediate neighbours to fill in the missing data, ensuring these neighbourhoods are not excluded during the regression modelling process.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Geographically Weighted Random Forest Regression\u003c/h2\u003e \u003cp\u003eEnvironmental factors and their health effects might not be stable over space [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. Furthermore, T2DM prevalence exhibits a spatial autocorrelated pattern with similar prevalence rates clustered together in neighbourhoods [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e, \u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e]. Therefore, a spatial machine learning-based regression model, Geographically Weighted Random Forest (GWRF), is used in this study to explore and mitigate the effects of spatial autocorrelation. GWRF is a non-parametric, tree-based, ensemble model that inherits the classical Random Forest (RF) regression structure calibrated locally through a moving window approach using spatial weight matrixes [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. The model captures spatial variation in localized relationships by generating random forest regression models for each spatial unit using nearby observations. The voting structure of the random forest regression can further mitigate the effects of non-linear relationships between the dependent and independent variables and multicollinearity among the independent variables [\u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e86\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis study used the \u0026ldquo;SpatialML\u0026rdquo; package (Version 0.1.7) in the R Statistical Computing Environment (Version 4.4.1) to create GWRF models [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. The T2DM prevalence rate was the dependent variable, and 27 variables proposed in section \u003cspan refid=\"Sec4\" class=\"InternalRef\"\u003e2.2\u003c/span\u003e were included as independent variables in the modelling. The GWRF models were set to be based on adaptive spatial kernels, as the bandwidth size was adjusted based on the data\u0026rsquo;s local density. Following past studies, a random grid search was applied during the hyperparameter tuning [\u003cspan additionalcitationids=\"CR44\" citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. The number of trees to grow for each round was set to be between 200 and 1500. The number of variables randomly sampled as candidates at each split was set to be between 5 and 11, and the number of nearest neighbourhoods to be used as the bandwidth ranged from 100 to 600 during the tuning process without growing complex trees to control for overfitting. The model with the lowest global Out-of-Bag (OOB) Mean Squared Error (MSE) and the highest global OOB R\u003csup\u003e2\u003c/sup\u003e as the best performance was selected as the final model. The model results were interpreted through local OOB R\u003csup\u003e2\u003c/sup\u003e, local Permutation Feature Importance (PFI), and Partial Dependent Plots (PDPs) following random forest evaluation approaches. PFI scores measured the importance of individual features in the predictive power of the GWRF model. It offered insight into identifying which independent variable best fits the model by measuring how much the model\u0026rsquo;s performance decreases when the values of a specific feature are randomly shuffled (permuted). It also highlighted critical features contributing to the dependent variable under the pretense of non-linearity or collinearity among variables in the model. PDPs aid the interpretation of the marginal relationships between the predictor and response variables. It illustrates the average effect of a predictor on the predicted outcome while holding all other variables constant. This allows us to visualize the strength and association of a predictor to the response variable. This study uses the R package \u0026ldquo;pdp\u0026rdquo; Version 0.8.1 to compute PDPs on the global GWRF model, equivalent to standard Random Forest models [\u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e]. The generated plots provide an overview of each variable\u0026rsquo;s marginal influence across the study area.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Summary Statistics\u003c/h2\u003e \u003cp\u003eFigures \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea illustrate the spatial distribution of the T2DM prevalence rates across Toronto DA. On average, 9% of the Toronto population reported being diagnosed with T2DM, ranging from 0 to 21.66% in various DA. Overall, higher rates of T2DM are observed in the central, northern, southwestern, and along the southeastern shoreline. In comparison, lower rates are found in the central south, northeastern and northwestern regions. A Global Moran\u0026rsquo;s Index with 1000 Monte Carlo (MC) simulation tests of significance was performed for T2DM prevalence, resulting in a value of 0.4214 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001). This indicates a moderate positive spatial autocorrelation for T2DM prevalence in Toronto. Similar T2DM rates tend to be spatially clustered, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea. Other lifestyle indices in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb to \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ei also exhibit spatially clustered patterns, such as high regular alcohol drinking rates clustering in southern regions, where more than half of the population are regular drinkers. Obesity, smoking, poor mental health, and anxiety disorder rates are clustered in the central-west, accounting for 10 to 20% of the region\u0026rsquo;s population.\u003c/p\u003e \u003cp\u003eThe perception of environmental beauty and safety derived from SVI is assessed as an index ranging from 0 (least beautiful) to 1 (most beautiful). Neighbourhoods in the blue areas of Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed, such as central Toronto and southeast and southwest Toronto, received high beauty perception scores, ranging from 0.401 to 0.719. Likewise, these regions also have the highest NDVI (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ei) coverage across the city, ranging from 0.248 to 0.404. Regarding built environment characteristics, most factors, including road network, cycling network, wall, fence, pole, and light coverage, were found to cluster in the southern downtown regions, as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eOn the other hand, socioeconomic factors in Toronto show distinct spatial patterns, as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. The average Toronto population is 42 years old, with 44.64% married or with common law. Higher-income populations were found in central Toronto (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec), where higher-income populations reached more than 70,000 CAD median annual income. Non-citizens (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed) and immigrant populations (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee) cluster in eastern and northern DA, with the latter reaching up to 100%. A higher percentage of walking and cycling commutes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ei) cluster in the downtown core, while longer weekly commuting durations (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ej) are concentrated in suburban areas, reaching a maximum of 2640 minutes weekly. A smaller portion of the population lacks certificates or diplomas, with an average of 12.75%. Unemployment remains low, with an average of 2.45%, but peaks at nearly 30% in specific areas. The summary statistics for all the variables are also listed in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary statistics for variables included in the GWRF regression model\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMinimum\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMaximum\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMedian\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eStandard Deviation\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT2DM Prevalence (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21.6578\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9.0811\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e9.2553\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4.3024\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObesity Rate (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.9574\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e37.3653\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e18.3241\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e17.7219\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5.5923\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSmoking Rate (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9114\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36.0052\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9.8956\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e8.2169\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5.7542\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDrinking Rate (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e23.4564\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e84.9271\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e48.1948\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e46.0313\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13.2037\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWeekly Recreation Physical Activity Minutes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e18.2656\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e69.3730\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e33.3717\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e32.3303\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6.5192\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHave Regular Health Care Provider (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e53.8117\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e99.9382\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e87.0798\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e89.8600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8.9792\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoor Mental Health (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12.3337\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.2696\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.7691\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.8387\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLife Dissatisfaction (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.3778\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.5229\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.3046\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.6416\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnxiety Disorder (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.1651\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e25.3759\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.9948\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7.4276\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.6283\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRoad Network Density (Distance/m\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0245\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0129\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0122\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0038\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCycling Network Density (Distance/m\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0066\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0012\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWater Body Density (% of Coverage)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.3237\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0052\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0172\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStreet View Beauty and Safety Perception\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0131\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.7187\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.2911\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.2869\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.1271\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStreet View Wall Coverage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0148\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0032\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0022\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0028\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStreet View Fence Coverage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0317\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0119\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0103\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0054\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStreet View Pole Coverage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0047\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0026\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0026\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0006\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStreet View Traffic Light Coverage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0435\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.4044\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.2121\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.2160\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0450\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e88.8000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e42.2093\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e41.9000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5.5787\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarried or Common-Law (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e62.3836\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e44.6355\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e45.9519\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7.9763\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedian Total Income (CAD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e114000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e42,899\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e38800\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13,553\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNon-Citizens (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e70.4225\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e14.2876\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e12.2388\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e9.6971\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eImmigrant Population (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e100.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e43.6088\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e44.1548\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e16.4581\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWithout Certificate or Diploma (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e43.1953\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12.7512\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11.4003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7.6821\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnemployment (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29.7884\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.4512\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.0929\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.8098\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCommute by Public Transit (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29.5203\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6.7932\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e6.0743\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4.4106\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCommute by Walking and Cycling (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29.7619\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.1626\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.9288\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.1307\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWeekly Commuting Duration (minutes)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2640.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e212.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e155.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e197.34\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Geographically Weighted Random Forest Regression\u003c/h2\u003e \u003cp\u003eThe final tuned GWRF model results with the hyperparameters: number of trees, Mtry, and neighbourhood bandwidth of 1088, 11, and 435, respectively. The global, local OOB R\u003csup\u003e2\u003c/sup\u003e, global, and local OOB MSE for the model were 0.7763, 0.7541, 4.1400, and 4.5505, respectively. This represents that around 75% of the variation in T2DM prevalence rate could be explained by the relationships with the explanatory variables through the model. More than half of the neighbourhoods had local R\u003csup\u003e2\u003c/sup\u003e values of 0.65 or above (purple areas in Figure \u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003eA1\u003c/span\u003ea of the \u003cspan refid=\"Sec18\" class=\"InternalRef\"\u003eAppendix\u003c/span\u003e). The random distribution of mapped OOB residuals in Figure \u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003eA1\u003c/span\u003eb of the \u003cspan refid=\"Sec18\" class=\"InternalRef\"\u003eAppendix\u003c/span\u003e and the Global Moran\u0026rsquo;s Index with 1000 MC simulations resulted in a value of 0.0196 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.02), indicating no significant spatial autocorrelation presented in the model\u0026rsquo;s residuals. Furthermore, 10-fold spatial cross-validation of the model\u0026rsquo;s prediction performance resulted in consistent performance across folds (Table \u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003eA2\u003c/span\u003e in the \u003cspan refid=\"Sec18\" class=\"InternalRef\"\u003eAppendix\u003c/span\u003e), with a mean R\u003csup\u003e2\u003c/sup\u003e, Root MSE, and Mean Absolute Error of 0.7224, 2.2770, and 1.7638 without showing signs of overfitting nor underfitting.\u003c/p\u003e \u003cp\u003eThe local PFI scores are visualized in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. Among all 27 variables in the model, lifestyle factors contributed the most overall, with access to regular health care ranking the highest, followed by obesity, smoking, drinking, and life dissatisfaction rates. Despite age being the second most crucial factor, socioeconomic factors such as income, immigrant status, and commuting patterns are ranked after lifestyle factors. Beauty and safety perceptions ranked the highest among all the environmental factors, followed by NDVI, which outperformed built environmental factors. The local PFI maps in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e showed inconsistent spatial effects of socioeconomic, environmental, and lifestyle factors on the T2DM prevalence rate. For example, age contributed to predicting T2DM the most in Northern DAs (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). In contrast, non-citizens, the immigrant population, drinking rate, and environmental perception of beauty and safety affect T2DM prevalence around downtown Toronto the most (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed, \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ee, \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003el, and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ev). The variables with the highest PFI were mapped in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e, representing the dominant feature associated with T2DM prevalence in each DA across Toronto. Regular health care dominates Toronto\u0026rsquo;s central DAs, while age and obesity were identified as the most important factors in the northeastern or southwestern and southeastern or southwestern DAs, respectively. Beauty and Safety Perceptions were found to be the most important features for some DAs in southern and central Toronto. The summary statistics of the PFI scores were recorded in Table \u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003eA1\u003c/span\u003e of the \u003cspan refid=\"Sec18\" class=\"InternalRef\"\u003eappendix\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThis study used PDPs to explore the associations between the dependent and independent variables while accounting for the average influence from other independent variables (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e). The relationships between T2DM prevalence and the independent variables were non-linear. The T2DM prevalence rate would sharply increase by about 3% on average when the population age increased from 35 to 60 (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ea). Moreover, T2DM prevalence would also increase by 2%, 2.5%, 4%, and 1% when the smoking rate, obesity prevalence, and the population with regular health care provider rate, environmental beauty and safety perception, increased from 0\u0026ndash;15%, 10\u0026ndash;20%, and 75\u0026ndash;95%, 0.1 to 0.5 respectively (Figs.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ek, \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003em, \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eo, \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ev). Most associations have an s- or inversed s-shaped sigmoid curve, where the effect gradually flattens at both ends. The T2DM prevalence rate was weakly positively associated with median income, physical activity, poor mental health, life dissatisfaction, anxiety disorder, and NDVI. Yet weakly negatively related to non-citizens, immigrant population, commuting duration and drinking rate.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Lifestyle Factors and Regular Health Care Access\u003c/h2\u003e \u003cp\u003eThis study revealed a positive association between well-designed environmental factors - including perceptions of safety, aesthetic appeal, and greenery - and higher T2DM prevalence rates. This differs from the hypothesis that a well-designed, beautiful, and safe neighbourhood environment can attract and promote physical activities that reduce the burden of T2DM. By examining the distribution of variables around Toronto, the spatial distribution of higher T2DM prevalences in DAs (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA) is associated with higher proportions of healthcare providers (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF), higher income levels (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC), greater perceptions of environmental beauty and safety (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD), higher NDVI measurements (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ei), as well as lower active transportation time (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eh and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ei) and commuting duration (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ej). The neighbourhoods in central Toronto are wealthier and receive better environmental beauty and safety perceptions; however, residents in these areas tend to commute by car, resulting in lower public transportation ridership across Toronto [\u003cspan citationid=\"CR88\" class=\"CitationRef\"\u003e88\u003c/span\u003e]. Compared to commuting and shopping through public transit, car ridership reduces the time spent on moderate and vigorous physical activity and increases sedentary time [\u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e89\u003c/span\u003e]. This lifestyle, characterized by more sedentary behaviours and decreased physical activity, may contribute to the observed increased prevalence of T2DM in these regions despite a favourable built environment. Furthermore, having a regular healthcare provider was ranked as the most important factor explaining T2DM prevalence in the GWRF model, which dominates most of Toronto\u0026rsquo;s DA (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). The standard of diabetes care developed by the American Diabetes Association has emphasized that access to regular healthcare is crucial for effective diabetes management and prevention [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Despite that existing studies reported that 70% of Ontario received at least a blood glucose test between 2010 and 2017 [\u003cspan citationid=\"CR90\" class=\"CitationRef\"\u003e90\u003c/span\u003e, \u003cspan citationid=\"CR91\" class=\"CitationRef\"\u003e91\u003c/span\u003e], more recent statistics from 2019 to 2021 show that fewer than 50% of diabetic populations met Diabetes Canada\u0026rsquo;s recommended standards for regular Hemoglobin A1c (HBA1c) blood glucose testing [\u003cspan citationid=\"CR92\" class=\"CitationRef\"\u003e92\u003c/span\u003e]. These findings may suggest that wealthier areas with better access to healthcare are more likely to screen and diagnose T2DM at earlier stages. In many cases, the residents of more affluent neighbourhoods with regular access to healthcare are more likely to prioritize their health outcomes and seek diabetes screenings regularly. In contrast, neighbourhoods with limited healthcare access may have undiagnosed or hidden cases of T2DM due to low awareness or a lack of screening, resulting in low prevalence in the region. No matter how favourable the environment is, if residents do not undergo a healthy lifestyle with regular diabetes screening, the burden of T2DM will persist. Therefore, this study suggests that T2DM intervention and prevention efforts should focus on increasing screening rates in neighbourhoods with fewer healthcare providers and poorer environmental perceptions while emphasizing education about healthy lifestyles in higher-income neighbourhoods with better environments in Toronto.\u003c/p\u003e \u003cp\u003eSimilar to previous studies, other lifestyle factors, such as increasing smoking rates, obesity prevalence, poor mental health, and anxiety disorders, were found to be correlated with increasing T2DM prevalence [\u003cspan additionalcitationids=\"CR32 CR33\" citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. They were identified as the most important determinants of T2DM compared to environmental and socioeconomic factors. Drinking rates exhibited spatial non-stationarity in their relationship with T2DM and were identified as a highly important factor in eastern downtown Toronto (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003el). The observed spatial pattern could reveal that in downtown regions with a cluster of pubs, higher drinking rates might reflect social lifestyle rather than behaviours directly contributing to T2DM risk, providing further insights into the positive association between alcohol consumption and T2DM observed in previous studies [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Residents around the downtown area often walk to pubs for drinking, promoting active transportation paired with their drinking behaviour. In other regions, drinking may reflect stress-included or habitual consumption, which can negatively impact health and increase T2DM risk.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Street View Image Derived Environmental Measurements\u003c/h2\u003e \u003cp\u003eEven though the perception of environmental beauty and safety was positively associated with T2DM prevalence rates in Toronto, it is identified as the most important environmental factor in the model, contributing more than the widely used environmental factors, including satellite-derived NDVI, GIS-derived water body, road, and cycling network density. Since the perception of beauty and safety measurements rely on environmental greenery and streetscape design, the vegetation coverage along the streets extracted from SVI could contribute to the importance of explaining T2DM. Residents constantly see roadside vegetation coverage during daily commuting, exercising, and shopping compared to greenery spaces like parks. A previous study identified that green space quantified using parks, recreational areas, sports fields, open spaces, campgrounds, and golf courses is relatively less in central Toronto than in the suburban regions of eastern Toronto. Some neighbourhoods in suburban areas with larger green spaces were associated with higher T2DM prevalence [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. However, the perceived environmental safety and beauty extracted from SVI in this study were the most important local features in central Toronto compared to other regions of Toronto (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ey and 6aa). Even though Eastern Toronto (Scarborough) regions have the highest green space coverage in Toronto, including a national urban park, there might not be opportunities for outdoor activities without the concern of being interrupted by dangerous road traffic. This may also suggest that simply increasing the green space or neighbourhood greenery without considering its aesthetics, safety, and accessibility may not effectively control or reduce T2DM prevalence rates. An unsafe environment that limits physical activity opportunities and creates social isolation will also contribute to an increased prevalence of T2DM in neighbourhoods around Toronto. Moreover, the SVI-derived built environmental characteristics observed showed a significant positive association between increasing walls and traffic lights with higher T2DM rates from the GWRF model. This aligns with the judgement that more walls and traffic lights on the streets due to urban development create barriers to physical activity around the city that promote an active lifestyle, which was associated with lower T2DM prevalence [\u003cspan citationid=\"CR93\" class=\"CitationRef\"\u003e93\u003c/span\u003e]. Therefore, SVI-derived perception of the environment can provide a more comprehensive quantification and representation of both the built and contextual environmental factors.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Socioeconomic Status\u003c/h2\u003e \u003cp\u003eSimilar to SVI-derived environmental variables, associations between socioeconomic status were also found with T2DM prevalence. For example, the association between age and T2DM prevalence, the second most important factor identified in the GWRF model, showed an increasing positive trend, starting with no relationship but encountering a sharp rise at age 37. It gradually transitioned to a weak positive relationship by age 50 and slowly stopped at 70. This aligns with the well-recognized phenomenon that individuals with increasing age are more prone to developing T2DM [\u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e94\u003c/span\u003e, \u003cspan citationid=\"CR95\" class=\"CitationRef\"\u003e95\u003c/span\u003e]. In addition, non-citizens and immigrant populations were negatively associated with the T2DM prevalence rate. Unlike the traditional hypothesis that immigrants experience a socio, environmental and dietary structural change, including increased consumption of ultra-processed foods, leading to a higher risk of developing T2DM [\u003cspan citationid=\"CR92\" class=\"CitationRef\"\u003e92\u003c/span\u003e, \u003cspan citationid=\"CR96\" class=\"CitationRef\"\u003e96\u003c/span\u003e]. The observed local permutation feature importance map indicated that most contributions are concentrated in Toronto\u0026rsquo;s downtown region (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ee). This could be explained by the fact that new immigrants living in the high-density downtown areas are wealthier, younger, and more likely to choose healthier diets. However, another possibility is that new immigrants and non-citizens typically have limited access to healthcare resources in Canada, leading to lower screening and prevalence rates among immigrant populations [\u003cspan additionalcitationids=\"CR98\" citationid=\"CR97\" class=\"CitationRef\"\u003e97\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR99\" class=\"CitationRef\"\u003e99\u003c/span\u003e]. Furthermore, active commuting by transit was found to have a weak negative relationship with T2DM, which is aligned with the observation from previous studies [\u003cspan citationid=\"CR100\" class=\"CitationRef\"\u003e100\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Significance of Contribution, Limitations, and Future Directions\u003c/h2\u003e \u003cp\u003eThis study is one of the first to utilize the novel GWRF regression to explore street view image-derived environmental factors, including perceptions of beauty and safety with T2DM prevalence across a large-scale city-wide study area. It provides a detailed characterization of spatial non-stationarity in environmental, socioeconomic, and lifestyle factors, highlighting their varying impacts across different city regions. GWRF successfully handled the spatial autocorrelated T2DM prevalence rate, further providing insights into local feature importance and highlighting the spatial variation of factors (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e and \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). Combining the results from partial dependency plots, local feature importance derived from GWRF offers a comprehensive understanding of how key variables contribute to T2DM prevalence within the study area. This approach identifies the most influential variables driving T2DM prevalence, including access to regular health care, age, and obesity prevalence. It further reveals their localized associations and spatial variability, providing a holistic perspective on the spatial dynamics of T2DM risk factors.\u003c/p\u003e \u003cp\u003eThe results of this study will support the design of targeted T2DM intervention and prevention programs around Toronto neighbourhoods. Understanding the spatial relationship of environmental, socioeconomic, and lifestyle characteristics with T2DM will inform policy and urban planning efforts to create healthier, more livable communities. Healthcare practitioners and policymakers must consider the neighbourhoods\u0026rsquo; diverse spatial phenomena and social-environmental aspects during intervention programs and policymaking. For example, evidence from this study could suggest that identifying and optimizing resource allocation, particularly healthcare and targeted screening resources, is critical for neighbourhoods with lower incomes, poorer environmental factors, and lower regular healthcare providers. Meanwhile, promoting healthy lifestyle behaviour education in weather and neighbourhoods with high car ridership. This targeted approach could help bridge disparities in T2DM prevention and care across different communities.\u003c/p\u003e \u003cp\u003eThere are several areas for future research to improve our understanding of the underlying mechanisms and address existing knowledge gaps. One limitation of this study was the availability of a single-year T2DM prevalence rate. With multiple years of data, future studies could investigate the lag effect of environmental, socioeconomic, and lifestyle determinants on T2DM prevalence. Multitemporal data could also be utilized to evaluate the predictive performance of GWRF models and enhance their accuracy by balancing model complexity with reducing the risk of overfitting. Additionally, the predictive performance could be compared against other models, such as Multiscale Geographically Weighted Regression. Future studies should also explore whether lifestyle behaviours, such as reduced active commuting or sedentary time, mediate the relationship between environmental factors and T2DM prevalence. When studying regions with lower screening rates, using prevalence data may lead to erroneous conclusions in examining social and environmental determinants of T2DM. Therefore, exploring the interaction of environments and contextual factors with target achievement rates, such as HbA1c levels, could provide further insights into the role of environmental conditions in primary (preventing the development of diabetes) and secondary (management of diabetes to prevent complications) T2DM preventions.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis study is among the first to investigate the associations between street view-derived environmental, socioeconomic, and lifestyle factors with T2DM prevalence through Geographically Weighted Random Forest regression. The results revealed that regular access to health care providers, age, obesity, smoking, mental illness, environmental beauty and safety perception, and NDVI were the most important factors positively and non-linearly associated with T2DM prevalence. This differs from the traditional hypothesis that a better-designed environment and frequent access to healthcare providers lead to reduced T2DM. However, the population with higher T2DM prevalence resided in wealthier and more affluent neighbourhoods with more frequent access to healthcare providers in Toronto. This suggests the importance of strengthening lifestyle interventions and education in communities with better healthcare conditions and environments to prevent T2DM actively. At the same time, screening efforts should be enhanced in communities with relatively poorer healthcare and environmental conditions. The findings from this study will provide evidence and recommendations to support the government and public health authorities in designing targeted education, prevention, and intervention programs that can control and reduce the increasing burden of T2DM.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCAD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCanadian Dollar\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDissemination Areas\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGWR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGeographically Weighted Regression\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGWRF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGeographically Weighted Random Forest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHbA\u003csub\u003e1c\u003c/sub\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHemoglobin A\u003csub\u003e1C\u003c/sub\u003e\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMSE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMean Squared Error\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMonte Carlo\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNDVI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNormalized Difference Vegetation Index\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eOLS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eOrdinary Least Square\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eOOB\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eOut-of-Bag\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePDPs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePartial Dependency Plots\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePFI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePermutation Feature Importance\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eR\u003csup\u003e2\u003c/sup\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCoefficient of Determination\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSVI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eStreet View Imagery\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eT2DM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eType-2 Diabetes Mellitus\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eVIF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eVariance Inflation Factor\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDiabetes and healthy lifestyle datasets analyzed during the study are available from Environics Analytics, but restrictions apply to the availability of these data, which were used under license for the current study and are not publicly available. The Census and GIS are publicly available through the Canadian national, Ontario provincial, and the City of Toronto open data portals listed in the method section supporting this article's findings and conclusions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting of Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Social Sciences and Humanities Research Council of Canada [Grant Number: 430-2023-00116] and the Manchester-Melbourne-Toronto Research Fund.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors’ Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eH.G. conceived, designed, and implemented the experiments, analyzed the results, and wrote the manuscript. J.W. conceptualized the study, acquired funding, supervised the project, and contributed to the review and editing of the manuscript. D.Y.W. contributed to acquiring and processing street view data, improving the methods, interpreting the results, and conducting the manuscript discussion session. H.Z. contributed to acquiring and processing lifestyle and street view data. All authors have read and approved the published version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors extend their sincere appreciation to the anonymous reviewers and editors for providing valuable comments that significantly contributed to improving the quality of the manuscript. The authors also wish to thank Environics Analytics for providing access to the health and lifestyle factor data used in this study. Special thanks to Dr. Pierre Desrochers and Dr. Chunjiang Li for their unwavering support and valuable input, which played a crucial role in enhancing the overall quality of this research. Furthermore, the authors would like to acknowledge and express their gratitude to Mr. Xianmiao Zhao, Mr. Ryan Siu, and Mr. Bruce Huang for their invaluable assistance in configuring and troubleshooting the spatial computational workstation that contributes to the modelling process in this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors’ Information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDepartment of Geography and Planning, University of Toronto, 100 St. George Street, Toronto, ON, M5S 3G3, Canada\u003c/p\u003e\n\u003cp\u003eHaoxuan Ge, Jue Wang, and Devin Yongzhao Wu\u003c/p\u003e\n\u003cp\u003eDepartment of Geography, Geomatics and Environment, University of Toronto Mississauga, 3359 Mississauga Road, Mississauga, ON, L5L 1C6, Canada\u003c/p\u003e\n\u003cp\u003eHaoxuan Ge, Jue Wang, and Devin Yongzhao Wu\u003c/p\u003e\n\u003cp\u003eDepartment of Geography, Sustainability, Community, and Urban Studies, University of Connecticut, Storrs, CT, USA\u003c/p\u003e\n\u003cp\u003eHanlin Zhou\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUse of Generative AI and AI-assisted technologies\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn the preparation of this manuscript, generative AI (ChatGPT 4o) was used solely to enhance readability and improve language clarity during the writing process. The AI assistance was limited to refining grammar, sentence structure, logistics of paragraph flow, and overall coherence without influencing the originality, accuracy, or integrity of the content.\u003c/p\u003e\n\u003cp\u003eAll AI-generated outputs were carefully reviewed, examined, and edited under human oversight and control to ensure that the final text was accurate, complete, and free from biases. The substantive content, analysis, and conclusions presented in this manuscript remain entirely the responsibility of the authors.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eToronto Public Health. Health Surveillance Indicators: Diabetes [Internet]. 2017 [cited 2024 Jan 7]. Available from: https://www.toronto.ca/wp-content/uploads/2017/12/8c72-tph-hsi-diabetes-july18f.pdf\u003c/li\u003e\n\u003cli\u003eChristine PJ, Auchincloss AH, Bertoni AG, Carnethon MR, S\u0026aacute;nchez BN, Moore K, et al. Longitudinal Associations Between Neighborhood Physical and Social Environments and Incident Type 2 Diabetes Mellitus: The Multi-Ethnic Study of Atherosclerosis (MESA). JAMA Internal Medicine. 2015;175:1311\u0026ndash;20. \u003c/li\u003e\n\u003cli\u003eKahn SE, Hull RL, Utzschneider KM. Mechanisms linking obesity to insulin resistance and type 2 diabetes. Nature. 2006;444:840\u0026ndash;6. \u003c/li\u003e\n\u003cli\u003eAstell-Burt T, Feng X, Kolt GS. Is Neighborhood Green Space Associated With a Lower Risk of Type 2 Diabetes? Evidence From 267,072 Australians. Diabetes Care. 2013;37:197\u0026ndash;201. \u003c/li\u003e\n\u003cli\u003eBodicoat DH, O\u0026rsquo;Donovan G, Dalton AM, Gray LJ, Yates T, Edwardson C, et al. The association between neighbourhood greenspace and type 2 diabetes in a large cross-sectional study. BMJ Open. 2014;4:e006076. \u003c/li\u003e\n\u003cli\u003eCreatore MI, Glazier RH, Moineddin R, Fazli GS, Johns A, Gozdyra P, et al. Association of Neighborhood Walkability With Change in Overweight, Obesity, and Diabetes. JAMA. 2016;315:2211\u0026ndash;20. \u003c/li\u003e\n\u003cli\u003eDendup T, Feng X, Clingan S, Astell-Burt T. Environmental Risk Factors for Developing Type 2 Diabetes Mellitus: A Systematic Review. International Journal of Environmental Research and Public Health. 2018;15:78. \u003c/li\u003e\n\u003cli\u003eGlazier RH, Creatore MI, Weyman JT, Fazli G, Matheson FI, Gozdyra P, et al. Density, Destinations or Both? A Comparison of Measures of Walkability in Relation to Transportation Behaviors, Obesity and Diabetes in Toronto, Canada. PLOS ONE. 2014;9:e85295. \u003c/li\u003e\n\u003cli\u003eM\u0026uuml;ller-Riemenschneider F, Pereira G, Villanueva K, Christian H, Knuiman M, Giles-Corti B, et al. Neighborhood walkability and cardiometabolic risk factors in australian adults: an observational study. BMC Public Health. 2013;13:755. \u003c/li\u003e\n\u003cli\u003ePoortinga W. Perceptions of the environment, physical activity, and obesity. Social Science \u0026amp; Medicine (1982). 2006;63:2835\u0026ndash;46. \u003c/li\u003e\n\u003cli\u003eSundquist K, Eriksson U, Mezuk B, Ohlsson H. Neighborhood walkability, deprivation and incidence of type 2 diabetes: A population-based study on 512,061 Swedish adults. Health \u0026amp; Place. 2015;31:24\u0026ndash;30. \u003c/li\u003e\n\u003cli\u003eStockdale SE, Wells KB, Tang L, Belin TR, Zhang L, Sherbourne CD. The importance of social context: Neighborhood stressors, stress-buffering mechanisms, and alcohol, drug, and mental health disorders. Social Science \u0026amp; Medicine. 2007;65:1867\u0026ndash;81. \u003c/li\u003e\n\u003cli\u003eAhern M, Brown C, Dukas S. A National Study of the Association Between Food Environments and County-Level Health Outcomes. The Journal of Rural Health. 2011;27:367\u0026ndash;79. \u003c/li\u003e\n\u003cli\u003eAndersen ZJ, Raaschou-Nielsen O, Ketzel M, Jensen SS, Hvidberg M, Loft S, et al. Diabetes Incidence and Long-Term Exposure to Air Pollution: A cohort study. Diabetes Care. 2011;35:92\u0026ndash;8. \u003c/li\u003e\n\u003cli\u003eDzhambov AM, Dimitrova DD. Exposures to road traffic, noise, and air pollution as risk factors for type 2 diabetes: A feasibility study in Bulgaria. Noise and Health. 2016;18:133. \u003c/li\u003e\n\u003cli\u003eGebreab SY, Hickson DA, Sims M, Wyatt SB, Davis SK, Correa A, et al. Neighborhood social and physical environments and type 2 diabetes mellitus in African Americans: The Jackson Heart Study. Health \u0026amp; Place. 2017;43:128\u0026ndash;37. \u003c/li\u003e\n\u003cli\u003eHill-Briggs F, Adler NE, Berkowitz SA, Chin MH, Gary-Webb TL, Navas-Acien A, et al. Social Determinants of Health and Diabetes: A Scientific Review. Diabetes Care. 2020;44:258\u0026ndash;79. \u003c/li\u003e\n\u003cli\u003eMezuk B, Li X, Cederin K, Rice K, Sundquist J, Sundquist K. Beyond Access: Characteristics of the Food Environment and Risk of Diabetes. American Journal of Epidemiology. 2016;183:1129\u0026ndash;37. \u003c/li\u003e\n\u003cli\u003ePolsky JY, Moineddin R, Glazier RH, Dunn JR, Booth GL. Relative and absolute availability of fast-food restaurants in relation to the development of diabetes: A population-based cohort study. Canadian Journal of Public Health. 2016;107:eS27\u0026ndash;33. \u003c/li\u003e\n\u003cli\u003eS\u0026oslash;rensen M, Andersen ZJ, Nordsborg RB, Becker T, Tj\u0026oslash;nneland A, Overvad K, et al. Long-Term Exposure to Road Traffic Noise and Incident Diabetes: A Cohort Study. Environmental Health Perspectives. 2013;121:217\u0026ndash;22. \u003c/li\u003e\n\u003cli\u003eAmerican Diabetes Association Professional Practice Committee. 1. Improving Care and Promoting Health in Populations: Standards of Care in Diabetes\u0026mdash;2025. Diabetes Care. 2024;48:S14\u0026ndash;26. \u003c/li\u003e\n\u003cli\u003eChowdhury MAB, Uddin MJ, Khan HMR, Haque MR. Type 2 diabetes and its correlates among adults in Bangladesh: a population based study. BMC Public Health. 2015;15:1070. \u003c/li\u003e\n\u003cli\u003eKolpak P, Wang L. Exploring the social and neighbourhood predictors of diabetes: a comparison between Toronto and Chicago. Primary Health Care Research \u0026amp; Development. 2017;18:291\u0026ndash;9. \u003c/li\u003e\n\u003cli\u003eInternational Diabetes Federation. IDF Diabetes Atlas 2021 [Internet]. 2021 [cited 2023 Oct 20]. Available from: https://diabetesatlas.org/atlas/tenth-edition/\u003c/li\u003e\n\u003cli\u003eAgardh E, Allebeck P, Hallqvist J, Moradi T, Sidorchuk A. Type 2 diabetes incidence and socio-economic position: a systematic review and meta-analysis. International Journal of Epidemiology. 2011;40:804\u0026ndash;18. \u003c/li\u003e\n\u003cli\u003eWu H, Meng X, Wild SH, Gasevic D, Jackson CA. Socioeconomic status and prevalence of type 2 diabetes in mainland China, Hong Kong and Taiwan: a systematic review. Journal of Global Health. 2017;7:011103. \u003c/li\u003e\n\u003cli\u003eCheng YJ, Kanaya AM, Araneta MRG, Saydah SH, Kahn HS, Gregg EW, et al. Prevalence of Diabetes by Race and Ethnicity in the United States, 2011-2016. JAMA. 2019;322:2389\u0026ndash;98. \u003c/li\u003e\n\u003cli\u003eLedford CJW, Seehusen DA, Crawford PF. Geographic and Race/Ethnicity Differences in Patient Perceptions of Diabetes. Journal of Primary Care \u0026amp; Community Health. 2019;10:2150132719845819. \u003c/li\u003e\n\u003cli\u003eTenkorang EY. Early onset of type 2 diabetes among visible minority and immigrant populations in Canada. Ethnicity \u0026amp; Health. 2017;22:266\u0026ndash;84. \u003c/li\u003e\n\u003cli\u003eKnott C, Bell S, Britton A. Alcohol Consumption and the Risk of Type 2 Diabetes: A Systematic Review and Dose-Response Meta-analysis of More Than 1.9 Million Individuals From 38 Observational Studies. Diabetes Care. 2015;38:1804\u0026ndash;12. \u003c/li\u003e\n\u003cli\u003eŚliwińska-Mossoń M, Milnerowicz H. The impact of smoking on the development of diabetes and its complications. Diabetes and Vascular Disease Research. 2017;14:265\u0026ndash;76. \u003c/li\u003e\n\u003cli\u003eVerma S, Hussain ME. Obesity and diabetes: An update. Diabetes \u0026amp; Metabolic Syndrome: Clinical Research \u0026amp; Reviews. 2017;11:73\u0026ndash;9. \u003c/li\u003e\n\u003cli\u003eGarrett C, Doherty A. Diabetes and mental health. Clinical Medicine. 2014;14:669\u0026ndash;72. \u003c/li\u003e\n\u003cli\u003eRobinson DJ, Coons M, Haensel H, Vallis M, Yale J-F. Diabetes and Mental Health. Canadian Journal of Diabetes. 2018;42:S130\u0026ndash;41. \u003c/li\u003e\n\u003cli\u003eFord KJ, Robitaille A. How sweet is your love? Disentangling the role of marital status and quality on average glycemic levels among adults 50 years and older in the English Longitudinal Study of Ageing. BMJ Open Diabetes Research \u0026amp; Care. 2023;11. \u003c/li\u003e\n\u003cli\u003eZhang X, Bullard KM, Gregg EW, Beckles GL, Williams DE, Barker LE, et al. Access to Health Care and Control of ABCs of Diabetes. Diabetes Care. 2012;35:1566\u0026ndash;71. \u003c/li\u003e\n\u003cli\u003eCuadros DF, Li J, Musuka G, Awad SF. Spatial epidemiology of diabetes: Methods and insights. World Journal of Diabetes. 2021;12:1042\u0026ndash;56. \u003c/li\u003e\n\u003cli\u003eKwan M-P. The stationarity bias in research on the environmental determinants of health. Health \u0026amp; Place. 2021;70:102609. \u003c/li\u003e\n\u003cli\u003eGe H, Wang J. Spatial Non-Stationarity Effects of Unhealthy Food Environments and Green Spaces for Type-2 Diabetes in Toronto. Sustainability. 2023;15:1762. \u003c/li\u003e\n\u003cli\u003eKauhl B, Schweikart J, Krafft T, Keste A, Moskwyn M. Do the risk factors for type 2 diabetes mellitus vary by location? A spatial analysis of health insurance claims in Northeastern Germany using kernel density estimation and geographically weighted regression. International Journal of Health Geographics. 2016;15:38. \u003c/li\u003e\n\u003cli\u003eGeorganos S, Grippa T, Niang Gadiaga A, Linard C, Lennert M, Vanhuysse S, et al. Geographical random forests: a spatial extension of the random forest algorithm to address spatial heterogeneity in remote sensing and population modelling. Geocarto International. 2021;36:121\u0026ndash;36. \u003c/li\u003e\n\u003cli\u003eGeorganos S, Kalogirou S. A Forest of Forests: A Spatially Weighted and Computationally Efficient Formulation of Geographical Random Forests. ISPRS International Journal of Geo-Information. 2022;11:471. \u003c/li\u003e\n\u003cli\u003eQui\u0026ntilde;ones S, Goyal A, Ahmed ZU. Geographically weighted machine learning model for untangling spatial heterogeneity of type 2 diabetes mellitus (T2D) prevalence in the USA. Scientific Reports. 2021;11:6955. \u003c/li\u003e\n\u003cli\u003eLotfata A, Moosazadeh M, Helbich M, Hoseini B. Socioeconomic and environmental determinants of asthma prevalence: a cross-sectional study at the U.S. County level using geographically weighted random forests. International Journal of Health Geographics. 2023;22:18. \u003c/li\u003e\n\u003cli\u003eLotfata A, Grekousis G, Wang R. Using geographical random forest models to explore spatial patterns in the neighborhood determinants of hypertension prevalence across chicago, illinois, USA. Environment and Planning B: Urban Analytics and City Science. 2023;50:2376\u0026ndash;93. \u003c/li\u003e\n\u003cli\u003eWu D, Zhang Y, Xiang Q. Geographically weighted random forests for macro-level crash frequency prediction. Accident Analysis \u0026amp; Prevention. 2024;194:107370. \u003c/li\u003e\n\u003cli\u003eKang Y, Zhang F, Gao S, Lin H, Liu Y. A review of urban physical environment sensing using street view imagery in public health studies. Annals of GIS. 2020;26:261\u0026ndash;75. \u003c/li\u003e\n\u003cli\u003eNguyen QC, Khanna S, Dwivedi P, Huang D, Huang Y, Tasdizen T, et al. Using Google Street View to examine associations between built environment characteristics and U.S. health outcomes. Preventive Medicine Reports. 2019;14:100859. \u003c/li\u003e\n\u003cli\u003eNguyen QC, Keralis JM, Dwivedi P, Ng AE, Javanmardi M, Khanna S, et al. Leveraging 31 Million Google Street View Images to Characterize Built Environments and Examine County Health Outcomes. Public Health Reports. 2021;136:201\u0026ndash;11. \u003c/li\u003e\n\u003cli\u003eWei J, Yue W, Li M, Gao J. Mapping human perception of urban landscape from street-view images: A deep-learning approach. International Journal of Applied Earth Observation and Geoinformation. 2022;112:102886. \u003c/li\u003e\n\u003cli\u003eHe J, Zhang J, Yao Y, Li X. Extracting human perceptions from street view images for better assessing urban renewal potential. Cities. 2023;134:104189. \u003c/li\u003e\n\u003cli\u003eNguyen QC, Belnap T, Dwivedi P, Deligani AHN, Kumar A, Li D, et al. Google Street View Images as Predictors of Patient Health Outcomes, 2017\u0026ndash;2019. Big Data and Cognitive Computing. 2022;6:15. \u003c/li\u003e\n\u003cli\u003eNguyen TT, Nguyen QC, Rubinsky AD, Tasdizen T, Deligani AHN, Dwivedi P, et al. Google Street View-Derived Neighborhood Characteristics in California Associated with Coronary Heart Disease, Hypertension, Diabetes. International Journal of Environmental Research and Public Health. 2021;18:10428. \u003c/li\u003e\n\u003cli\u003eKeralis JM, Javanmardi M, Khanna S, Dwivedi P, Huang D, Tasdizen T, et al. Health and the built environment in United States cities: measuring associations using Google Street View-derived indicators of the built environment. BMC Public Health. 2020;20:215. \u003c/li\u003e\n\u003cli\u003ePainter K. The influence of street lighting improvements on crime, fear and pedestrian street use, after dark. Landscape and Urban Planning. 1996;35:193\u0026ndash;201. \u003c/li\u003e\n\u003cli\u003eVlahov D, Freudenberg N, Proietti F, Ompad D, Quinn A, Nandi V, et al. Urban as a Determinant of Health. Journal of Urban Health. 2007;84:16\u0026ndash;26. \u003c/li\u003e\n\u003cli\u003eSeresinhe CI, Preis T, Moat HS. Quantifying the Impact of Scenic Environments on Health. Scientific Reports. 2015;5:16899. \u003c/li\u003e\n\u003cli\u003eZhang F, Zhou B, Liu L, Liu Y, Fung HH, Lin H, et al. Measuring human perceptions of a large-scale urban region using machine learning. Landscape and Urban Planning. 2018;180:148\u0026ndash;60. \u003c/li\u003e\n\u003cli\u003eGalea S, Ahern J, Rudenstine S, Wallace Z, Vlahov D. Urban built environment and depression: a multilevel analysis. Journal of Epidemiology \u0026amp; Community Health. 2005;59:822\u0026ndash;7. \u003c/li\u003e\n\u003cli\u003eWang R, Helbich M, Yao Y, Zhang J, Liu P, Yuan Y, et al. Urban greenery and mental wellbeing in adults: Cross-sectional mediation analyses on multiple pathways across different greenery measures. Environmental Research. 2019;176:108535. \u003c/li\u003e\n\u003cli\u003eWon J, Lee C, Forjuoh SN, Ory MG. Neighborhood safety factors associated with older adults\u0026rsquo; health-related outcomes: A systematic literature review. Social Science \u0026amp; Medicine. 2016;165:177\u0026ndash;86. \u003c/li\u003e\n\u003cli\u003eWu J, Li Q, Feng Y, Bhuyan SS, Tarimo CS, Zeng X, et al. Active commuting and the risk of obesity, hypertension and diabetes: a systematic review and meta-analysis of observational studies. BMJ Global Health. 2021;6. \u003c/li\u003e\n\u003cli\u003eZhou H, Wang J, Widener M, Wilson K. Examining the relationship between active transport and exposure to streetscape diversity during travel: A study using GPS data and street view imagery. Computers, Environment and Urban Systems. 2024;110:102105. \u003c/li\u003e\n\u003cli\u003eZhou H, Wang J, Wilson K. Impacts of perceived safety and beauty of park environments on time spent in parks: Examining the potential of street view imagery and phone-based GPS data. International Journal of Applied Earth Observation and Geoinformation. 2022;115:103078. \u003c/li\u003e\n\u003cli\u003eXu J, Liu Y, Liu Y, An R, Tong Z. Integrating street view images and deep learning to explore the association between human perceptions of the built environment and cardiovascular disease in older adults. Social Science \u0026amp; Medicine. 2023;338:116304. \u003c/li\u003e\n\u003cli\u003eLiu J, Yang L, Xiao L, Tao Z. Perceived Neighborhood Environment Impacts on Health Behavior, Multi-Dimensional Health, and Life Satisfaction. Frontiers in Public Health. 2022;10. \u003c/li\u003e\n\u003cli\u003eBennett GG, McNeill LH, Wolin KY, Duncan DT, Puleo E, Emmons KM. Safe To Walk? Neighborhood Safety and Physical Activity Among Public Housing Residents. PLOS Medicine. 2007;4:e306. \u003c/li\u003e\n\u003cli\u003eStatistics Canada. Census Profile, 2021 Census of Population [Internet]. 2022 [cited 2024 Jan 6]. Available from: https://www12.statcan.gc.ca/census-recensement/2021/dp-pd/prof/index.cfm?Lang=E\u003c/li\u003e\n\u003cli\u003eEnvironics Analytics [Internet]. 2024 [cited 2024 Oct 31]. Available from: https://environicsanalytics.com/en-ca/data/health-databases/communityhealth\u003c/li\u003e\n\u003cli\u003eGovernment of Canada SC. Canadian Community Health Survey (CCHS) - 2018 [Internet]. 2018 [cited 2025 Jan 18]. Available from: https://www23.statcan.gc.ca/imdb/p3Instr.pl?Function=assembleInstr\u0026amp;a=1\u0026amp;\u0026amp;lang=en\u0026amp;Item_Id=839130\u003c/li\u003e\n\u003cli\u003eWu DY, Wang J. Analyzing Urban Crime Through Street View Imagery: Insights from Urban Micro Built Environment and Perceptions. Urban Science. 2024;8:247. \u003c/li\u003e\n\u003cli\u003eDubey A, Naik N, Parikh D, Raskar R, Hidalgo CA. Deep Learning the City : Quantifying Urban Perception At A Global Scale [Internet]. arXiv; 2016 [cited 2023 Oct 21]. Available from: http://arxiv.org/abs/1608.01769\u003c/li\u003e\n\u003cli\u003eGoogle Maps. Google Maps Street View [Internet]. Google Maps Street View. 2024 [cited 2024 Feb 21]. Available from: https://www.google.com/streetview/\u003c/li\u003e\n\u003cli\u003eZhao H, Shi J, Qi X, Wang X, Jia J. Pyramid Scene Parsing Network [Internet]. arXiv; 2017 [cited 2023 Oct 21]. Available from: http://arxiv.org/abs/1612.01105\u003c/li\u003e\n\u003cli\u003eCordts M, Omran M, Ramos S, Rehfeld T, Enzweiler M, Benenson R, et al. The Cityscapes Dataset for Semantic Urban Scene Understanding [Internet]. arXiv; 2016 [cited 2023 Oct 21]. Available from: http://arxiv.org/abs/1604.01685\u003c/li\u003e\n\u003cli\u003eGuo J, He H, He T, Lausen L, Li M, Lin H, et al. GluonCV and GluonNLP: Deep Learning in Computer Vision and Natural Language Processing [Internet]. arXiv; 2020 [cited 2023 Oct 21]. Available from: http://arxiv.org/abs/1907.04433\u003c/li\u003e\n\u003cli\u003eLandsat 8 | U.S. Geological Survey [Internet]. 2025 [cited 2025 Jan 18]. Available from: https://www.usgs.gov/landsat-missions/landsat-8\u003c/li\u003e\n\u003cli\u003eCanada NR. Lakes, Rivers and Glaciers in Canada - CanVec Series - Hydrographic Features - Open Government Portal [Internet]. 2019 [cited 2025 Jan 18]. Available from: https://open.canada.ca/data/en/dataset/9d96e8c9-22fe-4ad2-b5e8-94a6991b744b\u003c/li\u003e\n\u003cli\u003eOntario Road Network [Internet]. 2022 [cited 2025 Jan 18]. Available from: https://geohub.lio.gov.on.ca/datasets/mnrf::ontario-road-network-orn-road-net-element/about\u003c/li\u003e\n\u003cli\u003eToronto C of. Sidewalk Inventory [Internet]. City of Toronto. City of Toronto; 2017 [cited 2025 Jan 18]. Available from: https://www.toronto.ca/services-payments/streets-parking-transportation/sidewalk-tours-wayfinding/sidewalks-in-toronto/sidewalk-inventory/\u003c/li\u003e\n\u003cli\u003eKwan M-P. The Uncertain Geographic Context Problem. Annals of the Association of American Geographers. 2012;102:958\u0026ndash;68. \u003c/li\u003e\n\u003cli\u003eVall\u0026eacute;e J, Cadot E, Grillo F, Parizot I, Chauvin P. The combined effects of activity space and neighbourhood of residence on participation in preventive health-care activities: The case of cervical screening in the Paris metropolitan area (France). Health \u0026amp; Place. 2010;16:838\u0026ndash;52. \u003c/li\u003e\n\u003cli\u003eMoreno C, Allam Z, Chabaud D, Gall C, Pratlong F. Introducing the \u0026ldquo;15-Minute City\u0026rdquo;: Sustainability, Resilience and Place Identity in Future Post-Pandemic Cities. Smart Cities. 2021;4:93\u0026ndash;111. \u003c/li\u003e\n\u003cli\u003eChen X, Ye X, Widener MJ, Delmelle E, Kwan M-P, Shannon J, et al. A systematic review of the modifiable areal unit problem (MAUP) in community food environmental research. Urban Informatics. 2022;1:22. \u003c/li\u003e\n\u003cli\u003eTompkins JW, Luginaah IN, Booth GL, Harris SB. The Geography of Diabetes in London, Canada: The Need for Local Level Policy for Prevention and Management. International Journal of Environmental Research and Public Health. 2010;7:2407\u0026ndash;22. \u003c/li\u003e\n\u003cli\u003eLuo Y, Yan J, McClure S. Distribution of the environmental and socioeconomic risk factors on COVID-19 death rate across continental USA: a spatial nonlinear analysis. Environmental Science and Pollution Research. 2021;28:6587\u0026ndash;99. \u003c/li\u003e\n\u003cli\u003eGreenwell BM. pdp: An R Package for Constructing Partial Dependence Plots. The R Journal. 2017;9:421\u0026ndash;36. \u003c/li\u003e\n\u003cli\u003eYousefzadeh Barri E, Farber S, Kramer A, Jahanshahi H, Allen J, Beyazit E. Can transit investments in low-income neighbourhoods increase transit use? Exploring the nexus of income, car-ownership, and transit accessibility in Toronto. Transportation Research Part D: Transport and Environment. 2021;95:102849. \u003c/li\u003e\n\u003cli\u003eChaix B, Kestens Y, Duncan S, Merrien C, Thierry B, Pannier B, et al. Active transportation and public transportation use to achieve physical activity recommendations? A combined GPS, accelerometer, and mobility survey study. Int J Behav Nutr Phys Act. 2014;11:124. \u003c/li\u003e\n\u003cli\u003eChu A, Shah BR, Rashid M, Booth GL, Fazli GS, Tu K, et al. Trends in glucose testing among individuals without diabetes in Ontario between 2010 and 2017: a population-based cohort study. Canadian Medical Association Open Access Journal. 2022;10:E772\u0026ndash;80. \u003c/li\u003e\n\u003cli\u003eWilson SE, Rosella LC, Lipscombe LL, Manuel DG. The effectiveness and efficiency of diabetes screening in Ontario, Canada: a population-based cohort study. BMC Public Health. 2010;10:506. \u003c/li\u003e\n\u003cli\u003eLorraine L. Lipscombe, Fatema M. Ali, Krista Lamb, Laura C. Rosella, Baiju R. Shah, Ian S. Zenlea. Current State of Type 2 Diabetes in the Peel Region [Internet]. Novo Nordisk Network for Healthy Populations; 2024. Available from: https://healthypopulationsnetwork.utoronto.ca/sites/default/files/assets/files/nhp-reportcurrent-state-type-2-diabetes-peel-region.pdf\u003c/li\u003e\n\u003cli\u003ePasala SK, Rao AA, Sridhar GR. Built environment and diabetes. Int J Diabetes Dev Ctries. 2010;30:63\u0026ndash;8. \u003c/li\u003e\n\u003cli\u003ede Miguel-Yanes JM, Shrader P, Pencina MJ, Fox CS, Manning AK, Grant RW, et al. Genetic Risk Reclassification for Type 2 Diabetes by Age Below or Above 50 Years Using 40 Type 2 Diabetes Risk Single Nucleotide Polymorphisms. Diabetes Care. 2010;34:121\u0026ndash;5. \u003c/li\u003e\n\u003cli\u003eKirkman MS, Briscoe VJ, Clark N, Florez H, Haas LB, Halter JB, et al. Diabetes in Older Adults. Diabetes Care. 2012;35:2650\u0026ndash;64. \u003c/li\u003e\n\u003cli\u003eBerggreen-Clausen A, Pha SH, Alvesson HM, Andersson A, Daivadanam M. Food environment interactions after migration: a scoping review on low- and middle-income country immigrants in high-income countries. Public Health Nutrition. 2022;25:136\u0026ndash;58. \u003c/li\u003e\n\u003cli\u003ePandey M, Kamrul R, Michaels CR, McCarron M. Identifying Barriers to Healthcare Access for New Immigrants: A Qualitative Study in Regina, Saskatchewan, Canada. Journal of Immigrant and Minority Health. 2022;24:188\u0026ndash;98. \u003c/li\u003e\n\u003cli\u003eSundareswaran M, Martignetti L, Purkey E. Barriers to primary care among immigrants and refugees in Peterborough, Ontario: a qualitative study of provider perspectives. BMC Primary Care. 2024;25:199. \u003c/li\u003e\n\u003cli\u003eBajgain BB, Bajgain KT, Badal S, Aghajafari F, Jackson J, Santana M-J. Patient-Reported Experiences in Accessing Primary Healthcare among Immigrant Population in Canada: A Rapid Literature Review. International Journal of Environmental Research and Public Health. 2020;17:8724. \u003c/li\u003e\n\u003cli\u003eBopp M, Gayah VV, Campbell ME. Examining the Link Between Public Transit Use and Active Commuting. International Journal of Environmental Research and Public Health. 2015;12:4256\u0026ndash;74. \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Type-2 Diabetes, Street View Image, Environmental Health, Geographically Weighted Random Forest, Spatial Non-Stationarity","lastPublishedDoi":"10.21203/rs.3.rs-5984185/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5984185/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eNumerous studies have shown that the environment can affect the prevalence of Type-2 Diabetes Mellitus (T2DM) by encouraging healthy lifestyle behaviours. Alongside traditional medical prevention strategies, social determinants have also been prioritized as a top consideration in T2DM clinical treatment and care. However, limited research has explored the association between neighbourhood perceptions of aesthetics and safety and T2DM prevalence, potentially through indirect pathways influencing behavioural response. Combining the effects, this research has two main objectives: (1) to identify the relationships between street view environmental, socioeconomic, and lifestyle factors with T2DM prevalence rates and (2) to determine how these associations vary spatially across different regions of Toronto.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis study applied a Geographically Weighted Random Forest regression to analyze the spatial non-stationarity associations and account for potential confounding factors in the relationship between 27 variables with T2DM prevalence across Toronto\u0026rsquo;s 3,800 Dissemination Areas. After modelling, the study examined local variations in feature effects using partial dependency plots and permutation-based feature importance maps to assess how variable associations on T2DM prevalence various around Toronto.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe model achieved an R\u003csup\u003e2\u003c/sup\u003e of 77%. Having regular healthcare, age, smoking rate, and obesity prevalence have the strongest positive correlation with T2DM prevalence. Beauty and safety perception, NDVI, and mental issues have a weak positive association with T2DM prevalence. In the downtown financial districts, immigration rates and drinking rates were identified as negatively associated with T2DM prevalence. Meanwhile, marital status, obesity, life dissatisfaction, and commuting by walking or cycling were found to have positive or negative spatial non-stationary associations across different geographical regions in Toronto.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eStreet view-derived environmental perceptions show spatially non-stationary associations with the prevalence of T2DM in Toronto. Higher T2DM rates are observed in dissemination areas with better street-view environments and access to healthcare providers. This may reflect underdiagnosis in areas with poorer perceived environments and less frequent access to healthcare providers. Residents living in better-perceived environments may not necessarily engage in more physical activity or active transportation. The findings offer valuable insights to assist government and public health authorities design targeted prevention and intervention programs in Toronto.\u003c/p\u003e","manuscriptTitle":"A Geographically Weighted Random Forest Analysis of Spatial Non-Stationarity Association of Street View Environmental, Socioeconomic, and Lifestyle Factors with Type-2 Diabetes Prevalence in Toronto","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-11 12:14:42","doi":"10.21203/rs.3.rs-5984185/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3ed8e649-95b1-4396-b073-1ccecab28c1a","owner":[],"postedDate":"April 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-04-29T21:23:10+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-11 12:14:42","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5984185","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5984185","identity":"rs-5984185","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}