Observing the spread of common illnesses through a community: Using geographic information systems (GIS) for surveillance

Abstract

Background: The recent implementation of electronic medical record systems allows for the development of systems to track common illness across a defined community. With the threats of bioterrorism and pandemic illness, syndromic surveillance methodologies have become an important area of study. There has been limited study of the application of syndromic surveillance techniques to communities for tracking common illnesses to improve health system resource allocation and inform communities. Methods: We analyzed visits from 26 primary care sites and one emergency department in a health system during a 13-month period in 2007 to 2008. Visits were coded for common respiratory and gastrointestinal illnesses. Using geographic information systems techniques, we plotted home addresses and developed criteria for census tract inclusion. The spatial distribution of the illnesses patterns was analyzed using Bayesian smoothing, Kriging and SaTScan (SaTScan, Boston, MA) statistical methods. Results: The study included 857,555 visits, 107,286 of which were in the emergency department and 750,269 in the primary care sites. Patient visits were plotted and then aggregated to census tracts. We determined that at least a median of 10 visits per week was required to provide sufficient volume in defining census tracts included in the study (109 census tracts). Weekly visit rates by census tract were plotted using nearest neighbor empirical Bayesian smoothing and Kriging to produce a continuous surface. To detect statistical clustering of weekly visit rates, we used SaTScan and identified 7 weeks with statistically significant clusters for respiratory illnesses and 8 weeks with statistically significant clusters for gastrointestinal illnesses (out of 56 weeks included in the study). After adjusting for population density, the visit rate remained consistent for respiratory illnesses (analysis of variance P = .937), but the visit rate for gastrointestinal illnesses increased in the fourth population density quartile (statistically different from quartiles 1, 2 and 3; analysis of variance P < .001 with Tukey multiple comparisons test), which included the highest population density areas in the study. Conclusions: We were able to use geographic information systems to assess visit rates for common illnesses in a defined community and identified spatial variability over time. Additional research is needed to help define parameters for implementation, but we believe this can have benefit for allocation of health resources and communicating with the community.