Relative survival is a measure of survival corrected for the effect of causes of death other than the cancer of interest. This variable is provided by the population-based cancer registry and is an indicator of the quality of cancer patient management.
Geographical variations in relative cancer survival might reflect differences in health care effectiveness. Studies comparing geographical variations use estimates of relative survival obtained independently in each of the corresponding geographical units. The statistical stability of survival estimates can be seriously compromised when the population of these geographical units is small.
The main purpose of this study was to evaluate the geographical variation in the relative survival rates of women with breast cancer in the Girona Health Region.
Firstly, we proposed smoothing out the estimates of relative survival using a (complete) hierarchical Bayesian model. Secondly, we evaluated which geographical units would provide more stable estimates and, therefore, be used for the cancer survival analysis as an indicator of the quality of the health services.
We found that the best model to control extra-variability when estimating the relative survival of women with breast cancer in the Girona Health Region was a (complete) hierarchical Bayesian model incorporating two random effects, one for heterogeneity and one for spatial dependence.
The most stable relative survival estimates were obtained when the basic health areas were analyzed as geographical units. However, we did not find a statistically significant association between the rate of socioeconomic deprivation and the relative survival of breast cancer in women.