17: The Impact of Specific Surveillance System Methodology on Influenza-Like Illness Prevalence Rates

Abstract

Influenza surveillance allows the Centers for Disease Control and Prevention (CDC) and local public health departments (PHD) to monitor for an influenza outbreak and track related illness and deaths. The true prevalence rate of influenza-like illness (ILI) is determined by various reporting methods utilized by hospitals and outpatient clinics. The sensitivity of a specific surveillance system, as well as the amount and quality of data access and the system's ability to handle free text, can greatly affect reported ILI rates. The objective of this study is to compare ILI rates for the 2009-2010 flu season using GUARDIAN (Geographic Utilization of Artificial intelligence in Real-Time for Disease Identification and Alert Notification), an advanced syndromic surveillance system, and a standard electronic medical record (EMR) reporting system to overall ILI rates reported by local public health departments and CDC regional and national rates. A retrospective, cross-sectional study was conducted from 2009-Week 36 to 2010-Week 12 in the emergency department (ED) of a large, urban, academic medical center (AMC). All ED patient information (n= 29,037) was processed through both GUARDIAN and EMR reporting systems and assigned an ILI status (Yes/No). EMR data reflected a chief complaint report of fever and cough and/or sore throat. GUARDIAN data reflected a report of actual vital signs and utilized a sophisticated natural language processor (NLP) review of chief complaint and physician notes (free text data). Weekly ILI rates based on the above two surveillance systems were generated and compared to the ILI rates reported by local, regional, and national PHDs. ANOVA analysis was performed to compare ILI rates. It was observed that the profiles (shape of the curve) of the weekly ILI rates were similar, but the magnitudes differed. Standard EMR reporting using AMC data was statistically similar to that of local PHDs, CDC region, 5 and national ILI reporting (see Table). GUARDIAN ILI rates were, on average, higher by 8% to 10% and demonstrated a statistically significant difference compared to EMR reporting and national, regional and local ILI rates.Tabled 1 GUARDIAN utilizes all available patient chart data with a sophisticated NLP which allows it to be able to identify additional ILI cases that are not captured through traditional approaches, such as standard EMR reporting. In addition, GUARDIAN predictions matched clinical validation 97.3% of the time. Thus, the ILI rate detected by GUARDIAN in the AMC is consistently higher than EMR reporting, which is typically used by most health care facilities to report to local PHDs and the CDC.