2014.ISDS.Abstracts.Final.pdf


ISDS Annual Conference Proceedings 2014. This is an Open Access article distributed under the terms of the Creative Commons Attribution-
Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, 
and reproduction in any medium, provided the original work is properly cited.

ISDS 2014 Conference Abstracts

Comparison of Three Critical Syndrome Classifications: 
Louisiana vs. BioSense

Jenna Iberg Johnson*

Infectious Disease Epidemiology, Louisiana Office of Public Health, New Orleans, LA, USA

Objective
To compare the results of BioSense and Louisiana syndrome 

classifications for influenza-like-illness, gastrointestinal, and upper 
respiratory infections applied to Louisiana emergency department 
data.

Introduction
The Louisiana Office of Public Health (OPH) Infectious Disease 

Epidemiology Section (IDEpi) conducts emergency department (ED) 
syndromic surveillance using the Louisiana Early Event Detection 
System (LEEDS). IDEpi has the capability to define and change 
syndrome definitions in LEEDS based on surveillance needs and 
quality assurance activities. IDEpi submits all of the ED data to 
BioSense, which uses different syndrome definitions than LEEDS. 
Both BioSense and LEEDS use text and ICD code searches in any 
available chief complaint, admit reason and diagnosis data. The results 
of LEEDS and BioSense syndrome classifications for influenza-like-
illness (ILI), gastrointestinal (GI), and upper respiratory infections 
(URI) applied to Louisiana’s ED data were compared to examine if 
the different syndrome definitions yield similar results when applied 
to the same data.

Methods
Daily electronic ED data is imported to both the LEEDS and 

BioSense databases and processed for syndrome classification. 
IDEpi queried the LEEDS database and the BioSense front-end 
application to pull weekly visits classified as influenza-like-illness, 
gastrointestinal, and upper respiratory infections for the period 
of CDC week 1327 through week 1426 (6/30/13-6/28/14). The 
syndrome percentage means of BioSense and LEEDS syndrome pairs 
were compared with paired t-tests. The linear relationship between 
BioSense and LEEDS syndrome pairs were measured with Pearson 
correlation coefficients. The syndrome results were also split into the 
age groups used by the BioSense front-end application and Pearson 
correlation coefficients were calculated for each syndrome age group 
pair. The Early Aberration Reporting System (EARS) C2 method was 
applied to all syndrome results to examine if alerts were generated 
during corresponding weeks for each syndrome pair. Weekly data 
were exported from LEEDS and BioSense and analyzed in R statistics 
package.

Results
The syndrome percentage means of BioSense ILI and LEEDS ILI 

were significantly different (paired t-test, p<0.000). The correlation 
coefficient for BioSense ILI and LEEDS ILI was 0.98 and age 
group correlation coefficients ranged from 0.83 to 0.99 (Pearson’s 
correlation, p<0.000). C2 generated eleven alarms for BioSense ILI 
and twelve for LEEDS ILI, of which nine occurred on corresponding 
weeks.

The syndrome percentage means of BioSense GI and LEEDS GI 
were significantly different (paired t-test, p<0.000). The correlation 
coefficient for BioSense GI and LEEDS GI was 0.90 and age 
group correlation coefficients ranged from 0.69 to 0.96 (Pearson’s 

correlation, p<0.000). C2 generated two alarms for BioSense GI and 
one for LEEDS GI, of which one occurred on a corresponding week.

The syndrome percentage means of BioSense URI and LEEDS URI 
were significantly different (paired t-test, p<0.000). The correlation 
coefficient for BioSense URI and LEEDS URI was 0.96 and age 
group correlation coefficients ranged from 0.81 to 0.97 (Pearson’s 
correlation, p<0.000). C2 generated six alarms for BioSense URI 
and seven for LEEDS URI, of which six occurred on corresponding 
weeks.

Conclusions
The results of BioSense and LEEDS syndrome classifications 

for influenza-like-illness, gastrointestinal, and upper respiratory 
infections applied to Louisiana emergency department syndromic 
surveillance data were highly correlated for each syndrome however 
the syndrome percentage means were significantly different for 
each syndrome pair. Therefore, while percentages of total visits 
attributed to a syndrome as a measurement of syndrome burden may 
not be comparable, trends over time are comparable. In addition, the 
majority of C2 alerts were generated on corresponding weeks for 
each syndrome pair, providing confidence in the use of C2 applied 
to current syndrome definition results as a means of aberration 
detection. As public health jurisdictions work towards developing 
common syndrome classifications to increase data comparability 
across jurisdictions, this analysis provides evidence that the current 
differences in syndrome definitions between jurisdictions may not 
hinder comparability of trends over time.

Keywords
syndromic surveillance; syndrome classification; syndrome 
definition; BioSense

*Jenna Iberg Johnson

E-mail: jenna.ibergjohnson@la.gov    

Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org * 7(1):30, 2015