Emergency department visit data for rapid detection and monitoring of norovirus activity, United States

Abstract

Noroviruses are the leading cause of gastroenteritis in the United States, but timely measures of disease are lacking. BioSense, a national-level electronic surveillance system, assigns data on chief complaints (patient symptoms) collected during emergency department (ED) visits to 78 subsyndromes in near real-time. In a series of linear regression models, BioSense visits mapped by chief complaints of diarrhea and nausea/vomiting subsyndromes as a monthly proportion of all visits correlated strongly with reported norovirus outbreaks from 6 states during 2007-2010. Higher correlations were seen for diarrhea (R = 0.828-0.926) than for nausea/vomiting (R = 0.729-0.866) across multiple age groups. Diarrhea ED visit proportions exhibited winter seasonality attributable to norovirus; rotavirus contributed substantially for children <5 years of age. Diarrhea ED visit data estimated the onset, peak, and end of norovirus season within 4 weeks of observed dates and could be reliable, timely indicators of norovirus activity.

Keywords: United States; biosurveillance; chief complaint; detection; emergency department; enteric infections; gastroenteritis; norovirus; outbreaks; rotavirus; surveillance; tracking; viruses.

Figure 1

Proportion of BioSense emergency department (ED) visits for diarrhea subsyndrome (A) and norovirus and rotavirus surveillance data (B), United States, January 2007–April 2010. The proportion of ED visits mapped by chief complaint to diarrhea subsyndrome in the 6 states analyzed (Georgia, Missouri, Ohio, Pennsylvania, Tennessee, and Wyoming) and reported norovirus outbreak data displayed seasonal peaks in the winter months. This seasonal pattern was observed for all 5 age groups (0–4, 5–17, 18–64, >65 years, and all ages); a higher proportion of ED visits for diarrhea was seen among children <5 years of age. Rotavirus activity, as measured by the proportion of positive antigen tests, also showed winter seasonality, with peaks that lagged behind those of norovirus.

Figure 2

Model attribution of the proportion of BioSense emergency department (ED) visits mapped by chief complaint to diarrhea subsyndrome compared with estimates of norovirus and rotavirus infections, United States, January 2007–April 2010. A) Patients 0–4 years of age; B) patients of all ages. Predicted norovirus largely accounted for the observed seasonal variations in the proportion of diarrhea visits in the all-ages group (17.5% of predicted total, January 2007–April 2010), with rotavirus making a smaller contribution (4.7%); norovirus (13.6%, August 2007–April 2010) and predicted rotavirus (13.6%) equally accounted for the seasonality in the 0–4-year age group. All other etiologies captured by the background and secular increase (baseline) did not contribute to the observed winter seasonality.

Figure 3

Correlation between the proportion of BioSense emergency department (ED) visits mapped by chief complaint to diarrhea subsyndrome and norovirus outbreaks as a function of total BioSense ED visits per month using state-specific data for the 6 states analyzed, United States, January 2007–April 2010. Correlation coefficients for each state are plotted by corresponding total ED visits/month on a logarithmic scale. Models tended to perform better in states with greater total ED visits. Higher correlation (R>0.60) was observed for states with >5,000 BioSense ED visits/month. State number labels on data points correspond to those in Table 2.

Figure 4

Estimation of norovirus season time markers using BioSense data on emergency department (ED) visits mapped by chief complaint to diarrhea subsyndrome, United States, 2009–2011. Observed season time markers (solid vertical lines) as defined by norovirus outbreak data are labeled as follows: season onset (O), season peak (P), and season end (E). Applying these rules yielded estimates for each season marker (dotted vertical lines) within 2 weeks of observed dates for the 2009–2010 season and within 4 weeks of observed dates for the 2010–2011 season.