Using geographically weighted regression to validate approaches for modelling accessibility to primary health care

Abstract

This research explores local variation in accessibility to primary health care and relationships between travel time and New Zealand deprivation index in the rural Otago. The global relationship between travel time and NZDep2001 index was significantly negative with a t value of -6. 11. Suggesting that in general, areas with high travel time to PHC services have lower NZ Deprivation scores than areas with low travel time. Furthermore, there was a great deal of spatial variation in travel time and deprivation index which is not explained by the global regression framework. Geographically Weighted Regression (GWR) analysis was undertaken using an adaptively defined kernel with a bi-square function. The kernel bandwidth was determined by minimisation of the Akaike Information Criterion (AIC) value. The GWR outputs showed that some areas with high travel time were also areas of high NZ deprivation score. Therefore, the GWR results highlighted 'hot spot' areas in terms of poor accessibility to PHC facilities in rural Otago, New Zealand.