Assessing representativeness of a rural australian clinical database using a spatial modelling approach

It is generally recognized that people in rural Australia and world-wide do worse in terms of health outcomes compared to the urban population. Epidemiological studies rely on large datasets obtained through national surveys but efforts to survey rural populations usually result in small datasets. Hence small datasets are often disregarded even if they are the only source of health data available to study health outcomes at the local level. The main criticism is usually lack of representativeness of the general population. In this study, a spatial modelling approach was developed to assess the representativeness of a rural Australian clinical database. We compared two methods commonly used in health geography, namely Generalized Additive Models and the spatial scan statistic. Both methods were shown to have strengths that can be exploited to detect underrepresentation of a small health dataset. We concluded that our participant data are largely representative of the underlying population and highlight focus areas for further participant recruitment, allowing disease cluster mapping to with confidence, even on the small dataset.