Spatial and Spatio-Temporal Modelling of Cancer Data using Bayesian Approaches and Machine Learning Techniques

This thesis develops and applies Bayesian spatial and spatio-temporal models, together with machine learning approaches, to strengthen small-area estimation of cancer clinical quality registry outcomes in Victoria, Australia. Analyses of registry data identified marked geographic inequities in lung cancer care quality, diagnostic intervals, treatment initiation, and survival, with persistent disadvantage in socioeconomically deprived and remote populations. Joint modelling delineated overlapping spatial structures across cancers, indicating potential common etiologic and healthcare pathways. Machine learning algorithms demonstrated strong predictive validity for local lung cancer mortality. The findings yield statistically robust and clinically relevant evidence to inform equity-focused cancer control strategies.