Bayesian latent variable modelling of multivariate spatio-temporal variation in cancer mortality

Area of Science:

• Biostatistics
• Epidemiology
• Spatial Analysis

Background:

• Multivariate health data often exhibit complex spatial and temporal correlations.
• Analyzing multiple related diseases jointly requires flexible statistical frameworks.
• Understanding cancer risk trends necessitates methods that account for geographic and time-based variations.

Purpose of the Study:

• To develop and apply Bayesian hierarchical latent factor models for analyzing spatially and temporally correlated health data.
• To jointly analyze multiple diet-related cancers in Greece, estimating common and disease-specific risk trends.
• To uncover latent spatial and temporal patterns in cancer data potentially linked to the Greek population's habitual diet.

Main Methods:

• Description of three alternative Bayesian hierarchical latent factor models.
• Integration of factor analysis principles with space-time disease mapping techniques.
• Application to area-level mortality data for six diet-related cancers in Greece (1980-1999).

Main Results:

• The study successfully applied Bayesian models to analyze complex cancer data.
• Identified spatial and temporal patterns in diet-related cancer mortality across Greece.
• The models provided a framework for estimating latent factors influencing cancer risk.

Conclusions:

• Bayesian hierarchical latent factor models offer a flexible approach for joint analysis of multivariate, space-time health data.
• The identified patterns suggest a link between latent factors, potentially reflecting dietary habits, and cancer risk in Greece.
• Further research can utilize these models to explore environmental and lifestyle influences on disease patterns.