Geospatial analysis of breast cancer incidence in Virginia 2000-2005: A methodological, analytical and policy perspective

Background. High breast cancer incidence and mortality rates remain of concern in Virginia. Virginia's cancer data remains under-analyzed using current geospatial techniques because of methodological issues regarding geocoding, under-reporting in certain areas of the state, and appropriate units of analysis. This study addresses the impediments to small area analysis by examining data accuracy and levels of aggregation, and provides a theoretical framework for understanding the reliability and consistency of various geospatial clustering techniques.

Objective/purpose. The study analyzes the spatial distribution of Virginia's breast cancer incidences from 2000-2005 by addressing the methodological problem of regional incidence under-reporting in rural areas of the state, and determines if data collection methodologies could be hindering reliable cluster detection.

Methods. Standard geospatial cluster techniques, including, SatScan, Geoda softwares and ArcGIS, were used to determine the Moran Index, Kriging, Kernel Density contours, and Hot Spots with Rendering. Data reliability at different levels of aggregation from Block Group, Census Tract, to ZIP Code was determined. Small area analysis techniques to reframe additional research questions and area cluster detection methodologies were also applied.

Results. The research has demonstrated the viability of using spatial analysis techniques to detect clusters in data with regional under-reporting. A convergence of clusters was also noted at all levels of aggregations.

Discussion/conclusions. Because areas of under-reporting have been detected, physician education can now be targeted to improve response rates. The study has reframed future inquiry, provided guidance for public health interventions, and the basis for community involvement in new models of care.