• No results found

More recent research on spatial analysis of access to dental services

Chapter 2 – Neighbourhoods, health and inequalities

2.6. Spatial inequalities in oral health

2.6.3. More recent research on spatial analysis of access to dental services

GIS and spatial analysis techniques have increased in scope since Taylor and

Carmichael’s (1980) work, and now cover a worldwide dental literature. Themes such as access to dental services have emerged as popular topics within the literature, and in the Australian context in particular. McGuire et al. (2011) used concentric circles to estimate patient distances to dental clinics in Victoria, while also demonstrating positive relationships between increased emergency service usage and areas of lower socio- economic status. Rocha et al. (2013) identified similar patterns in Melbourne, using geocoded patient data to show that those living over 10 kilometres from the Royal Dental Hospital experienced poorer access, while deprivation increased with distance from the hospital. Similar patterns were found in a study of public and private dentist locations in New South Wales (Willie-Stephens et al, 2014), demonstrating per capita income to be a good indicator of private dental service location, with comparable trends also seen in New Zealand (Kruger et al, 2012).

Kruger et al. (2011) used various geographical scales overlaid with graticule squares, combined with concentric circles, to show that 345,000 Western Australians fell outside 2 kilometre service buffers added to dental practices, with large disparities between urban and rural areas in practices per graticule. When compared to Queensland and Victoria, Western Australia was shown to have the fewest Census districts within 200 kilometres of a public dental clinic (Perera et al, 2010), while deprivation scores were also higher in these excluded populations. In line with this, Madan et al. (2010) demonstrated a trend for higher rates of public sector general anaesthetic in remote areas, compared to higher private rates in more accessible areas of the state which also correlated with ‘better’ socio-economic profiles. Again the findings suggested the need

57

to focus on more disadvantaged and remote communities. Similar methods to those above have also been used in Australia to examine water fluoridation coverage among schoolchildren (Desai et al, 2015). The use of buffers linked to socio-economic data demonstrated links between socio-economic deciles and the proportion of children in non-optimally fluoridated schools, as well as disparities between urban and rural areas in fluoride coverage. Given the vast space services in this country must cover, it is unsurprising that research has taken this focus. However, these techniques are often descriptive and offer little explanation as to why certain patterns occur.

Transportation can also act as a significant barrier to services. Borrell et al. (2006) used subway lines to demonstrate disparities in the locations of dental care for elderly residents of ethnic minorities in New York City. This showed that residents in North Manhattan may have been forced to use Columbia University’s dental clinics due to being negatively affected by transport limitations. Rail transportation formed the basis of work by Zainab et al. (2015), which used buffers, geo-coded locations and socio- economic data to demonstrate that, despite the city’s excellent urban transport planning, retirees and elders in Sydney had lower accessibility to train networks capable of getting them to the Sydney Dental Hospital. This study is typical of many within this literature, in that buffers are added to geo-located points, then compared with measures of socio- economic or demographic status usually taken from administrative data or surveys. Contrary to the results of previous studies measuring distance to services, Dumas and Polk (2015) demonstrated that distance to dental providers was not sufficient in

explaining barriers to services in urban areas when investigating dental clinic utilisation among Medicaid insured children in Pittsburgh, with children often travelling further than needed rather than visiting their closest dentist. A state-wide study from Iowa however suggests that Medicaid insured children faced more serious travel burdens than those with private insurance (McKernan et al, 2016), living further from their nearest dentist, and having to drive further to access their current service. Lower rates of ‘bypass’ were also seen in Medicaid and rural populations, most likely due to travel practicalities and resources. McKernan et al. (2015) identified dental service areas in the same state using small area analysis of patient origins and destination data, using this to create weighted dental visits per zip code for each dentist. Their analysis using these service areas again demonstrated urban-rural disparities, as well as racial disparities in geographical accessibility. The urban-rural disparities demonstrated in this study are

58

typical of the North American context (Emami et al, 2016). Nasseh et al. (2017) used road networks instead of straight line distance in combination with travel times, and a two-step floating catchment area method to define counties with dental shortages. This demonstrated that higher proportions of publicly insured children lived in service deficient areas in Missouri than Wisconsin, despite significant variation in both states. Despite the descriptive nature of some of the literature, novel geographical methods have also been used in spatial studies to analyse dental related trends. Feng et al. (2016) used geographically weighted regression, local indicators of spatial association, and spatial autocorrelation analysis to demonstrate a lack of overall association between the size of the dental workforce with the varying utilisation rates seen across the Appalachia region of the United States. Similar spatial statistics were applied in Jones et al. (2016a), with the addition of a three step floating catchment area method to calculate

neighbourhood level access to dental care and family physicians. This demonstrated that service types were concentrated close to other similar services, with dental services being more highly concentrated and located in the urban centre and important

commercial areas to the east of the centre, areas of greater socio-economic advantage. Spatial autocorrelation was also used by Meyer (2014) to analyse ‘hot spots’ of dental clustering in major urban areas of Ontario. Hot-spots of dental offices tended to be located in downtown areas, with ‘high-sale offices’ typically in suburban areas adjacent to neighbourhoods with deficits in dental services, also being characterised by higher income, population, growth and younger residents. Jager et al. (2016) used a novel approach involving ordinary least squares regression and geographically weighted regression of demographic data and data on dental workforce losses to estimate the dental workforce in 2030 in Northern Germany. Many areas were shown to be over- served, while some rural regions had no dental services at all. Use of Moran’s I also showed there was no compensation from over-served neighbouring areas.

Saman et al. (2011) utilised hot-spot analysis to identify clusters based on the

percentage of adults with 6 or more teeth removed due to tooth decay or gum disease, along with cartogram analysis to demonstrate oral health disparities by population and the ratio of dentists per 10,000. Counties with lower populations and dentist ratios were shown to have worse oral health. A fuller explanation of cartogram-based methods will be provided in Section 4.10. Horner et al. (2007) used GIS to identify zip-codes falling outside a 10-mile radius from dental practices, before using a ‘location set covering

59

problem’ (p.114) to allocate services to these areas. Some new practices were able to cover multiple zip-codes, as under-served areas tended to be clustered spatially. This is a rare example of using GIS for potential practical solutions, rather than descriptive analysis. Krause et al. (2012) highlight some of the general challenges in using GIS to analyse oral health disparities, in particular the collection of data at aggregated scales, making small area analysis difficult.