Evaluating spatial patterns of influenza-like illness from Emergency Department visit data in Illinois

Introduction:  Federal incentive programs and national surveillance system applications have expanded the opportunity for public health jurisdictions to receive health data directly from Electronic Medical Records (EMR). In Illinois, Emergency Department (ED) visit surveillance has recently expanded as part of a state-wide implementation of syndromic surveillance reporting. During influenza seasons, visits classified as influenza-like illness (ILI) can provide additional information on the magnitude and distribution of illness.

Methods: Illinois submits ED visit data daily from two-thirds of acute-care hospitals, representing over 75% of the total ED visits in Illinois, to the National Syndromic Surveillance Program. During the 2014-2015 influenza season, all visits between September 28 and February 7 were queried for ILI chief complaint or diagnoses terms and aggregated to the county-level. GeoDa software was used to calculate global and local spatial autocorrelation using Moran I and Local G statistics. At the county level, ILI rates were compared to age and economic metrics, including proportion of the population under 5, proportions of the population over 64, median household income, and percent living in poverty, using ordinary least squares regression with applied weighting for spatial dependence. The strength of the linear relationship between ILI rate and  these variables was conducted using Spearman correlation coefficient.

Results: Through the peak of the 2014-2015 influenza season, there were 1.3 million ED visits reported to Illinois Department of Public Health, with 98% of counties having some ILI activity. The median ILI rate across counties was 21.4 per 1000 visits. Spatial autocorrelation indicated a significant cluster of low ILI activity in the eastern region and significant clusters of high ILI activity in western and southern counties. Results from the regression analysis were marginally significant for lower median household income (p=0.0736), but not for any other county-level demographic or economic indicators.

Discussion: State-wide ED visit surveillance can complement other influenza surveillance systems and provide an indicator of disease trends from milder illness presentations in geographic areas that may otherwise be underrepresented. The ecological nature of results at the county level did not indicate strong associations between ILI rates and population characteristics. As coverage levels increase and a historic data repository evolves, analysis at finer levels of spatial granularity may provide more conclusive results.