Modelling access to GPs relative to need in Australia: Geographic variation among First Nations and non-Indigenous populations

First introduced in 2014, the Australian Institute of Health and Welfare’s (AIHW) index of Access Relative to Need (the ‘ARN index’) estimates how local access to General Practitioners (GPs) relative to the need for primary health care varies across Australia for First Nations people and non-Indigenous Australians at the level of the Australian Bureau of Statistics’ (ABS) Statistical Areas Level 1 (SA1). The type of access to GPs considered in the ARN model is physical access and does not take other potentially important barriers such as affordability or cultural appropriateness into account. This report presents the recent refinements of the ARN methodology and discusses data and other methodological issues that have the potential to limit the accuracy of estimates of access to services in Australia.

Access

The model used to calculate the estimate of access that is part of the ARN index belongs to a family of models called floating catchment models. Instead of relying on services-to-population ratios within fixed areas or other similar measures, these models assume that access to services also depends on distance or travel time. The ARN index access model is a 3-step floating catchment model:

1. The first step estimates the number of people from each SA1 population who choose each available GP service location based on estimated drive times and GP capacity (full-time equivalent GPs). Services with relatively short drive times and high capacity attract more people than other services. The model assumes that all GP service locations within an hour’s drive can be accessed.
2. The second step estimates the GPs to need-adjusted population ratio for each service – that is the demand on each service relative to its capacity. The demand put on each service location by a population depends on the willingness to travel from each population to the service, the number of people from each population choosing to use the service and the per capita need of each population. The longer the drive time, the less willing people are to travel to their chosen service location.
3. The third step estimates the access of each SA1 population based on the GP to need-adjusted population ratio at each service used by the population as the sum of the ratios of all services within reach of each population adjusted for the drive time cost and the proportion of people attending each service adjusted by the willingness to travel.

Per capita need and drive times are key components of this model. The need estimates are based on socioeconomic and demographic characteristics of the First Nations, non-Indigenous and total SA1 populations. Analysis of the high-level associations between those characteristics and the amount of time that Australians spend with GPs in parts of Australia where access to services is relatively good has allowed us to calibrate the overall differences in per capita need between populations.

Drive times from local populations to service locations were estimated using the centroid (geographic midpoint) of each SA1 or, for SA1s larger than 3,000 km², multiple point locations based on where people live within the SA1s as captured by the ABS’s population grid.

Access relative to need

In addition to estimates of access produced by the 3-step floating catchment model, we also calculate an estimate of access relative to need (the composite ARN index). The rationale behind using a measure of access relative to need, rather than just a measure of access where per capita need is used to estimate demand on services, is that the impact on a population of a change in access will depend on its need. A population with a high per capita need is likely to be affected more severely by low access than a population with a low per capita need. Whereas calibration of the demand aspect of need can be done using data on the use of GP services, it is much more challenging to calibrate how need interacts with access to shape health outcomes over long periods. Because of this, the access relative to need estimates should always be considered alongside the access estimates.

Geographic variation in access

The ARN modelling shows that the proportion of people living in areas where physical access to GPs is relatively limited increases with increasing remoteness. For example, 3% of First Nations people and 1% of non-Indigenous Australians in Major cities live in areas with an access score below 7 (access equivalent to having 7 GPs per 10,000 people of average need if there are no drive time barriers). This increases to 71% and 58% respectively in Very remote areas. Whereas a higher proportion of First Nations people live in areas with relatively poor access in both Major cities and Very remote areas, that is not the case in Inner regional, Outer regional or Remote areas.

A higher proportion of First Nations people than non-Indigenous Australians live in regional and remote areas where residents are more likely to live in areas with poor access than in the major cities. This contributes to the overall difference in the proportion of First Nations people (17%) and non-Indigenous Australians (5.1%) living in areas with access scores below 7.