Exploring the intra-urban variations in the relationship among geographic accessibility to PHC services and socio-demographic factors

In this study, we investigate the intra-urban variations in the relationships among various socio-demographic factors and geographical accessibility to primary health care (PHC) services using a local regression model. Geographic accessibility to PHC services is calculated at a local scale for two Canadian urban centers (Calgary, AB and Toronto, ON) using a three-step floating catchment area (3SFCA) method. Socio-demographic factors were derived from 2006 Canada census data. The regression analysis was performed using two different methods: 1) a single regression model for both cities together, using a regional dummy variable, and 2) separate models for each city. A similar modeling procedure was applied for both methods: first, a best Ordinary Least Squares (OLS) regression model was determined using a forward step-wise approach in SPSS software. Next, to test the spatial non-stationarity in the regression residuals, the best OLS model was repeated in ArcGIS. Further, to explore whether or not regression coefficients vary across space, we applied the geographically weighted regression (GWR) method with an adaptive spatial kernel. The GWR results exhibit the intra-urban variations in the relationships between socio-demographic factors and the accessibility score. A comparison of the GWR models demonstrates the benefit of local spatial regression in disaggregating the relationships between socio-demographic variables and the geographical accessibility to PHC services at a local scale; however, our results suggest that a more careful modeling approach is required when analysing the data with spatial effects.