Poster Presentation COSA-IPOS Joint Scientific Meeting 2012

Using routinely collected cancer data to infer risk factor patterns (#760)

Susanna M Cramb 1 2 , Kerrie L Mengersen 2 , Peter D Baade 1 3 4
  1. Cancer Council Queensland, Fortitude Valley, Qld, Australia
  2. School of Mathematical Sciences, Queensland University of Technology, Brisbane, Queensland, Australia
  3. School of Public Health and Social Work, Queensland University of Technology, Kelvin Grove, Queensland, Australia
  4. Griffith Health Institute, Griffith University, Southport, Queensland, Australia

Cancer Registries collect reliable and detailed population-based data on cancer incidence and mortality, providing the foundation for effective cancer control at both the global and local level. In contrast, cancer risk factor information is often based on self-reported survey data, which has known limitations, such as the potential to introduce bias. The typically limited sample sizes also preclude examining small-area patterns. Recent attempts to overcome these restrictions have involved applying statistical shared component models to cancer data to obtain estimates for the cancers of interest and also the underlying, unmeasured risk factors.

The goal of this study was to use routinely collected, population-based lung cancer incidence data to estimate the spatio-temporal patterns in the underlying risk factors, of which tobacco smoking is a major component.

Using data from the Queensland Cancer Registry, male and female lung cancer incidence rates were jointly modelled using a Bayesian spatio-temporal shared component model across 477 small areas and 13 years (1997 to 2009). The spatial pattern and temporal trends in the shared component were identified, and were consistent with available data on tobacco smoking. Females had a higher risk than males in the south-east corner, but a reduced risk in some regional areas to the north of Brisbane, which may indicate differences in the smoking pattern. Known influences on smoking, such as socioeconomic status, remoteness and Indigenous population, were also included at the area level to investigate their impact on the patterns.

These results enable areas with higher underlying risk factors to be identified with greater precision and confidence. The methodology can also be applied to other groups of cancers to estimate spatio-temporal patterns for other risk factors of interest. Further work is required to isolate the effects of lag time and other risk factors so as to better quantify the area-level smoking prevalence.