Modelling spatially correlated survival data for individuals with multiple cancers

Area of Science:

• Epidemiology
• Biostatistics
• Spatial Analysis
• Survival Analysis
• Cancer Research

Background:

• Spatial variation in disease patterns can indicate underlying risk factors influencing public health.
• Analyzing cancer survival data requires methods that account for geographical clustering and variations.
• The Surveillance Epidemiology and End Results (SEER) database offers a valuable resource for cancer survival research.

Purpose of the Study:

• To develop and apply Bayesian hierarchical survival models for capturing spatial patterns in cancer survival.
• To address the challenges of modeling survival data for multiple cancers and their spatial associations.
• To identify geographical regions with significant spatial variation in cancer survival for public health intervention.

Main Methods:

• Development of Bayesian hierarchical survival models within a proportional hazards framework.
• Introduction of county-cancer level frailties to represent spatial variation.
• Semiparametric modeling of the baseline hazard function using mixtures of beta distributions.
• Application and validation using data from the Surveillance Epidemiology and End Results (SEER) database.

Main Results:

• The developed models effectively capture spatial patterns and associations in cancer survival data.
• Model checking and comparison were performed to evaluate the performance of competing models.
• Implementation issues and practical considerations for applying these models were discussed.

Conclusions:

• Bayesian hierarchical survival models provide a robust framework for analyzing spatial effects in cancer survival.
• These models enhance the understanding of geographical disparities in cancer outcomes.
• The findings support public health efforts in targeted interventions for cancer control.