This project analyzes the prevalence and distribution of Type 2 Diabetes (T2D) across Canada using spatial data science techniques. T2D affects over 3 million Canadians and accounts for 90% of all diabetes cases, with significant health and economic consequences. While public health interventions often use a population-wide approach, this study focuses on regional and socio-economic disparities. Using data from the Canadian Community Health Survey (2009–2018), we explore the spatial and social determinants of T2D to inform more targeted public health strategies.
Existing studies tend to apply global models that treat the Canadian population as homogenous, ignoring local disparities in health outcomes. This project aims to investigate regional variations in T2D prevalence by incorporating spatial analysis techniques and exploring socio-economic and lifestyle-related determinants. The goal is to identify high-risk regions and support data-driven public health planning.
Our objective was to explore spatial patterns of T2D across Canada, analyze the relationship between T2D and socio-economic factors, identify regional disparities in T2D prevalence and inform region-specific public health interventions.
The analysis is based on the Canadian Community Health Survey (CCHS) data from 2009 to 2018 with variables on self-reported T2D status, socioeconomic and demographic indicators, health behaviours (e.g., BMI, physical activity) and regional information for spatial mapping. Canadian Community Health Survey (CCHS) cycles (2009–2018) were combined using the cchsflow R package to standardize variable names and formats across years, and the Type 2 Diabetes (T2D) variable was derived using the Ng-Dasgupta-Johnson (NDJ) algorithm with ~95% accuracy. Missing data was handled using MICE (Multivariate Imputation by Chained Equations) for numeric, binary, and categorical data, and geographic codes were adjusted and aligned for matching with Statistics Canada's health region shapefiles.
A binary logistic regression model was built using survey-weighted data to assess the impact of the following variables: Age, Sex, Ethnicity, Education, Income, Marital Status, BMI, Smoking, and Activity. The glmulti R package (AIC-based) was used for the mode,l and ANOVA for validation. The results showed that Age and BMI had the strongest effects, where Adults aged 65+ were 127x more likely to have T2D than those aged 18–1,9, and Obese individuals were 5.4x more likely to have T2D than those underweight. Other findings included:
- Males are more likely than females (OR = 1.38).
- Higher education/income reduced T2D risk.
- Non-white immigrants had a higher likelihood.
ArcGIS Pro was used to prepare spatial data by aggregating survey data to the health region level. Getis-Ord Gi* and Moran’s I were used to identify spatial clustering (hot/cold spots of T2D prevalence). The results showed that the prevalence of T2D varied geographically, ranging from 3.5% to 12.7%. Hot spots of high T2D prevalence were Atlantic Canada, Northern Ontario, and parts of Eastern Quebec, while the cold spots of low prevalence were British Columbia and Southern Alberta. Geographically Weighted Regression (GWR) was performed to analyze regional effects on diabetes prevalence, and AIC and R² were used to compare model performance across regions and validate spatial heterogeneity. Residuals appeared randomly distributed, indicating a well-normalized model.
- R
- ArcGIS Online