Individual level covariate adjusted conditional autoregressive (indiCAR) model for disease mapping.

Area of Science:

• Biostatistics
• Spatial Epidemiology
• Public Health

Background:

• Disease mapping visualizes geographical variations, aiding hypothesis generation.
• Traditional conditional autoregressive (CAR) models face memory constraints with large sample sizes.
• Existing methods are often infeasible for large-scale disease mapping applications.

Purpose of the Study:

• To introduce indiCAR, a novel algorithm for fitting CAR models in disease mapping.
• To accommodate individual and group-level covariates while adjusting for spatial correlation.
• To provide a scalable solution for large datasets where other methods fail.

Main Methods:

• Development of a new algorithm, indiCAR, for conditional autoregressive (CAR) model fitting.
• Incorporation of individual and group-level covariates.
• Adjustment for spatial correlation in disease rates.
• Application in a distributed computing framework for Big Data.

Main Results:

• Simulation studies confirm indiCAR's reliable estimation of regression and random effect parameters.
• Analysis of neutropenia admissions in NSW identified associations with age, gender, and socioeconomic factors.
• Significant spatial dependence indicates variations in cancer patient management across NSW.

Conclusions:

• Incorporating individual covariate data enhances parameter estimation in disease mapping.
• indiCAR is beneficial for health registries with individual and area-level data.
• The scalability of indiCAR is advantageous for Big Data applications in disease mapping.