Communicating the impact of the modifiable areal unit problem when estimating cancer burden at larger areal levels in Australia. — The Association Specialists
  1. Schedule
  2. Oral Presentations - Cancer causes and control
  3. Communicating the impact of the modifiable areal unit problem when estimating cancer burden at larger areal levels in Australia.

Communicating the impact of the modifiable areal unit problem when estimating cancer burden at larger areal levels in Australia. (23930)

Jeyamalar T Thurai Rathnam 1 2 , Eugene Slaughter 3 , Darren Wraith 4 , Kerrie Mengersen 5 , Peter Baade 1 , Jessica Cameron 1 2
  1. Viertel Cancer Research Centre, Cancer Council Queensland, 553 Gregory Terrace, Fortitude Valley, Queensland, Australia
  2. Centre for Data Science, Queensland University of Technology, Brisbane, Queensland, Australia
  3. Centre for Children's Health Research (CCHR), 62 Graham Street, South Brisbane, Queensland, Australia
  4. Faculty of Health, School of Public Health & Social Work, Queensland University of Technology, Brisbane, Queensland, Australia
  5. Faculty of Science, School of Mathematical Sciences, Queensland University of Technology, Brisbane, Queensland, Australia

Background

Spatial modelling is done at higher resolutions to ensure precision of estimates and minimise the risk of bias like the modifiable areal unit problem. The Australian Cancer Atlas provides cancer burden estimates using Statistical Area 2 (SA2) boundaries, an areal structure designed by the Australian Bureau of Statistics. The SA2s are designed to group together communities that interact socially and economically and have been shown to minimise the risk of bias. However, in practice, health resource management and interventions are often carried out by Health Service networks like the broader Primary Health Network (PHN) area structure. This study aimed to obtain PHN-level cancer incidence estimates that capture both a point estimate describing the average for the PHN with corresponding uncertainty and the variability within PHNs.

Methods

The standardised incidence ratios (SIRs) for 31 PHNs were modelled based on a Bayesian hierarchical framework, using SA2-level estimates obtained from the Australian Cancer Atlas. We trialled three different models and two distinct weighting strategies: one based on population distributions and the other based on expected count distributions.

Results

There is substantial variation in incidence rates within PHNs for all invasive cancers combined, resulting in all PHNs having similar incidence rates to the Australian average. The Western Queensland PHN had the highest variability in SIRs, reflecting greater disparities among its constituent SA2s.

Conclusion

PHN-level estimates of SIRs and the variability within PHNs will provide the evidence necessary for targeted interventions and efficient use of health resources. These data could be used to guide planning and resource allocation to improve cancer outcomes across Australia.