Managing distance and covariate information with point-based clustering.

Area of Science:

• Spatial epidemiology
• Geographic information systems (GIS)
• Public health research

Background:

• Geographic analysis of disease often relies on point-based data and assessing spatial patterns like clustering.
• Understanding spatial disease patterns requires methods that handle covariates and spatial clustering within finite locations.
• Previous methods lacked rigorous approaches for constrained spatial clustering and bias in geographic distance measures.

Purpose of the Study:

• To develop and present a rigorous method for analyzing spatial clustering of point-based observations within a finite set of locations.
• To apply this method to investigate the clustering of deliberate self-harm (DSH) cases.
• To account for spatial covariates like socio-economic deprivation and biases in distance measurements.

Main Methods:

• A Monte-Carlo simulation approach based on Ripley's K function was employed.
• A rotated Minkowski L1 distance metric was used to assess variations in physical distance and clustering.
• Data from 136 emergency hospital presentations for DSH in a New Zealand town were analyzed, with the study area defined by residential land parcels.

Main Results:

• Spatially correlated area-based deprivation was identified.
• After accounting for deprivation and distance bias, evidence of DSH clustering was found at spatial scales up to 500 meters.
• The findings suggest a potential role for social contagion in the observed clustering of DSH within the urban cohort.

Conclusions:

• Estimating distance-based clustering at multiple scales in geographic space necessitates robust methods.
• A Monte-Carlo approach to Ripley's K, incorporating covariates and distance bias models, is vital for health-related clustering assessments.
• Neighborhood social network structures and spatially clustered covariates like deprivation are important factors in understanding point-based clustering of health outcomes such as DSH.