smallareamapp

smallareamapp estimates small-area cancer risk and maps disease incidence at Community Health Service Area (CHSA) levels using Bayesian hierarchical models (BYM2) implemented with Integrated Nested Laplace Approximation (INLA).

Key Features:

• Bayesian Hierarchical Modeling (BYM2): The modified Besag, York, and Mollié model (BYM2) estimates cancer incidence counts and adjusts for spatial dependence across CHSAs to produce stabilized risk estimates.
• Integrated Nested Laplace Approximation (INLA): INLA is used to implement Bayesian models and approximate posterior distributions for computational efficiency.
• Indirect Age-Standardization and SIRs: Indirect age-standardization is applied to derive age-adjusted expected counts and standardized incidence ratios (SIRs).
• Posterior Median Relative Risks (RRs): The workflow reports posterior median RRs for cancers across CHSAs.
• Spatial Autocorrelation Assessment (Moran's I): Moran's I statistic is computed to evaluate the strength and direction of spatial autocorrelation.
• Exceedance Probability Analysis: Exceedance probabilities are calculated with a threshold RR = 1.1, and areas with probability ≥ 80% are identified as elevated risk.
• Adjustment for Small Populations (Smoothing): Extreme SIRs in areas with small populations are smoothed toward a null relative risk (RR = 1.0) to stabilize estimates.
• Predictive Posterior Checks: Model validity is assessed using predictive integral transformation (PIT) values and comparisons of observed versus fitted values.

Scientific Applications:

• Small-area cancer risk mapping: Produces maps of lung, female breast, cervical, and colorectal cancer incidence among British Columbia residents (2011–2018).
• Identification of elevated-risk CHSAs: Identifies CHSAs with elevated cancer risks based on exceedance probabilities.
• Hypothesis generation: Generates hypotheses about potential environmental or demographic factors driving spatial variation in cancer incidence.
• Examination of ecologic associations: Enables examination of ecologic associations while adjusting for spatial dependencies.
• Public health planning: Quantifies geographic disparities to inform public health strategies and resource allocation.

Methodology:

Obtain cancer incidence data from registries (e.g., the BC Cancer Registry); apply indirect age-standardization to derive age-adjusted expected counts and SIRs; fit Bayesian hierarchical models using a modified BYM2 implemented with INLA to estimate posterior median RRs across CHSAs; compute Moran's I, calculate exceedance probabilities (threshold RR = 1.1; report ≥ 80%), smooth SIRs toward RR = 1.0 for small populations, and perform predictive posterior checks using PIT values and observed versus fitted comparisons.